Index: vendor/libarchive/dist/cpio/test/main.c =================================================================== --- vendor/libarchive/dist/cpio/test/main.c (revision 231296) +++ vendor/libarchive/dist/cpio/test/main.c (revision 231297) @@ -1,2612 +1,2612 @@ /* * Copyright (c) 2003-2009 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" #ifdef HAVE_SYS_TIME_H #include #endif #include #ifdef HAVE_ICONV_H #include #endif #include #include #ifdef HAVE_SIGNAL_H #include #endif #include #include /* * This same file is used pretty much verbatim for all test harnesses. * * The next few lines are the only differences. * TODO: Move this into a separate configuration header, have all test * suites share one copy of this file. */ __FBSDID("$FreeBSD: src/usr.bin/cpio/test/main.c,v 1.3 2008/08/24 04:58:22 kientzle Exp $"); #define KNOWNREF "test_option_f.cpio.uu" #define ENVBASE "BSDCPIO" /* Prefix for environment variables. */ #define PROGRAM "bsdcpio" /* Name of program being tested. */ #define PROGRAM_ALIAS "cpio" /* Generic alias for program */ #undef LIBRARY /* Not testing a library. */ #undef EXTRA_DUMP /* How to dump extra data */ #undef EXTRA_ERRNO /* How to dump errno */ /* How to generate extra version info. */ #define EXTRA_VERSION (systemf("%s --version", testprog) ? "" : "") /* * * Windows support routines * * Note: Configuration is a tricky issue. Using HAVE_* feature macros * in the test harness is dangerous because they cover up * configuration errors. The classic example of this is omitting a * configure check. If libarchive and libarchive_test both look for * the same feature macro, such errors are hard to detect. Platform * macros (e.g., _WIN32 or __GNUC__) are a little better, but can * easily lead to very messy code. It's best to limit yourself * to only the most generic programming techniques in the test harness * and thus avoid conditionals altogether. Where that's not possible, * try to minimize conditionals by grouping platform-specific tests in * one place (e.g., test_acl_freebsd) or by adding new assert() * functions (e.g., assertMakeHardlink()) to cover up platform * differences. Platform-specific coding in libarchive_test is often * a symptom that some capability is missing from libarchive itself. */ #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #ifndef F_OK #define F_OK (0) #endif #ifndef S_ISDIR #define S_ISDIR(m) ((m) & _S_IFDIR) #endif #ifndef S_ISREG #define S_ISREG(m) ((m) & _S_IFREG) #endif #if !defined(__BORLANDC__) #define access _access #undef chdir #define chdir _chdir #endif #ifndef fileno #define fileno _fileno #endif /*#define fstat _fstat64*/ #if !defined(__BORLANDC__) #define getcwd _getcwd #endif #define lstat stat /*#define lstat _stat64*/ /*#define stat _stat64*/ #define rmdir _rmdir #if !defined(__BORLANDC__) #define strdup _strdup #define umask _umask #endif #define int64_t __int64 #endif #if defined(HAVE__CrtSetReportMode) # include #endif #if defined(_WIN32) && !defined(__CYGWIN__) void *GetFunctionKernel32(const char *name) { static HINSTANCE lib; static int set; if (!set) { set = 1; lib = LoadLibrary("kernel32.dll"); } if (lib == NULL) { fprintf(stderr, "Can't load kernel32.dll?!\n"); exit(1); } return (void *)GetProcAddress(lib, name); } static int my_CreateSymbolicLinkA(const char *linkname, const char *target, int flags) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, DWORD); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateSymbolicLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, flags); } static int my_CreateHardLinkA(const char *linkname, const char *target) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, LPSECURITY_ATTRIBUTES); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateHardLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, NULL); } int my_GetFileInformationByName(const char *path, BY_HANDLE_FILE_INFORMATION *bhfi) { HANDLE h; int r; memset(bhfi, 0, sizeof(*bhfi)); h = CreateFile(path, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) return (0); r = GetFileInformationByHandle(h, bhfi); CloseHandle(h); return (r); } #endif #if defined(HAVE__CrtSetReportMode) static void invalid_parameter_handler(const wchar_t * expression, const wchar_t * function, const wchar_t * file, unsigned int line, uintptr_t pReserved) { /* nop */ } #endif /* * * OPTIONS FLAGS * */ /* Enable core dump on failure. */ static int dump_on_failure = 0; /* Default is to remove temp dirs and log data for successful tests. */ static int keep_temp_files = 0; /* Default is to run the specified tests once and report errors. */ static int until_failure = 0; /* Default is to just report pass/fail for each test. */ static int verbosity = 0; #define VERBOSITY_SUMMARY_ONLY -1 /* -q */ #define VERBOSITY_PASSFAIL 0 /* Default */ #define VERBOSITY_LIGHT_REPORT 1 /* -v */ #define VERBOSITY_FULL 2 /* -vv */ /* A few places generate even more output for verbosity > VERBOSITY_FULL, * mostly for debugging the test harness itself. */ /* Cumulative count of assertion failures. */ static int failures = 0; /* Cumulative count of reported skips. */ static int skips = 0; /* Cumulative count of assertions checked. */ static int assertions = 0; /* Directory where uuencoded reference files can be found. */ static const char *refdir; /* * Report log information selectively to console and/or disk log. */ static int log_console = 0; static FILE *logfile; static void vlogprintf(const char *fmt, va_list ap) { #ifdef va_copy va_list lfap; va_copy(lfap, ap); #endif if (log_console) vfprintf(stdout, fmt, ap); if (logfile != NULL) #ifdef va_copy vfprintf(logfile, fmt, lfap); va_end(lfap); #else vfprintf(logfile, fmt, ap); #endif } static void logprintf(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vlogprintf(fmt, ap); va_end(ap); } /* Set up a message to display only if next assertion fails. */ static char msgbuff[4096]; static const char *msg, *nextmsg; void failure(const char *fmt, ...) { va_list ap; if (fmt == NULL) { nextmsg = NULL; } else { va_start(ap, fmt); vsprintf(msgbuff, fmt, ap); va_end(ap); nextmsg = msgbuff; } } /* * Copy arguments into file-local variables. * This was added to permit vararg assert() functions without needing * variadic wrapper macros. Turns out that the vararg capability is almost * never used, so almost all of the vararg assertions can be simplified * by removing the vararg capability and reworking the wrapper macro to * pass __FILE__, __LINE__ directly into the function instead of using * this hook. I suspect this machinery is used so rarely that we * would be better off just removing it entirely. That would simplify * the code here noticeably. */ static const char *skipping_filename; static int skipping_line; void skipping_setup(const char *filename, int line) { skipping_filename = filename; skipping_line = line; } /* Called at the beginning of each assert() function. */ static void assertion_count(const char *file, int line) { (void)file; /* UNUSED */ (void)line; /* UNUSED */ ++assertions; /* Proper handling of "failure()" message. */ msg = nextmsg; nextmsg = NULL; /* Uncomment to print file:line after every assertion. * Verbose, but occasionally useful in tracking down crashes. */ /* printf("Checked %s:%d\n", file, line); */ } /* * For each test source file, we remember how many times each * assertion was reported. Cleared before each new test, * used by test_summarize(). */ static struct line { int count; int skip; } failed_lines[10000]; const char *failed_filename; /* Count this failure, setup up log destination and handle initial report. */ static void failure_start(const char *filename, int line, const char *fmt, ...) { va_list ap; /* Record another failure for this line. */ ++failures; failed_filename = filename; failed_lines[line].count++; /* Determine whether to log header to console. */ switch (verbosity) { case VERBOSITY_LIGHT_REPORT: log_console = (failed_lines[line].count < 2); break; default: log_console = (verbosity >= VERBOSITY_FULL); } /* Log file:line header for this failure */ va_start(ap, fmt); #if _MSC_VER logprintf("%s(%d): ", filename, line); #else logprintf("%s:%d: ", filename, line); #endif vlogprintf(fmt, ap); va_end(ap); logprintf("\n"); if (msg != NULL && msg[0] != '\0') { logprintf(" Description: %s\n", msg); msg = NULL; } /* Determine whether to log details to console. */ if (verbosity == VERBOSITY_LIGHT_REPORT) log_console = 0; } /* Complete reporting of failed tests. */ /* * The 'extra' hook here is used by libarchive to include libarchive * error messages with assertion failures. It could also be used * to add strerror() output, for example. Just define the EXTRA_DUMP() * macro appropriately. */ static void failure_finish(void *extra) { (void)extra; /* UNUSED (maybe) */ #ifdef EXTRA_DUMP if (extra != NULL) { logprintf(" errno: %d\n", EXTRA_ERRNO(extra)); logprintf(" detail: %s\n", EXTRA_DUMP(extra)); } #endif if (dump_on_failure) { fprintf(stderr, " *** forcing core dump so failure can be debugged ***\n"); abort(); exit(1); } } /* Inform user that we're skipping some checks. */ void test_skipping(const char *fmt, ...) { char buff[1024]; va_list ap; va_start(ap, fmt); vsprintf(buff, fmt, ap); va_end(ap); /* Use failure() message if set. */ msg = nextmsg; nextmsg = NULL; /* failure_start() isn't quite right, but is awfully convenient. */ failure_start(skipping_filename, skipping_line, "SKIPPING: %s", buff); --failures; /* Undo failures++ in failure_start() */ /* Don't failure_finish() here. */ /* Mark as skip, so doesn't count as failed test. */ failed_lines[skipping_line].skip = 1; ++skips; } /* * * ASSERTIONS * */ /* Generic assert() just displays the failed condition. */ int assertion_assert(const char *file, int line, int value, const char *condition, void *extra) { assertion_count(file, line); if (!value) { failure_start(file, line, "Assertion failed: %s", condition); failure_finish(extra); } return (value); } /* chdir() and report any errors */ int assertion_chdir(const char *file, int line, const char *pathname) { assertion_count(file, line); if (chdir(pathname) == 0) return (1); failure_start(file, line, "chdir(\"%s\")", pathname); failure_finish(NULL); return (0); } /* Verify two integers are equal. */ int assertion_equal_int(const char *file, int line, long long v1, const char *e1, long long v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e1, v1, v1, v1); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e2, v2, v2, v2); failure_finish(extra); return (0); } /* * Utility to convert a single UTF-8 sequence. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch; int cnt; uint32_t wc; *pwc = 0; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if (n < (size_t)cnt) return (-1); /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) return (-1);/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if (n < 4) return (-1); if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); if ((s[3] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) return (-1);/* Overlong sequence. */ break; default: return (-1); } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > 0x10FFFF) return (-1); /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); } static void strdump(const char *e, const char *p, int ewidth, int utf8) { const char *q = p; logprintf(" %*s = ", ewidth, e); if (p == NULL) { logprintf("NULL\n"); return; } logprintf("\""); while (*p != '\0') { unsigned int c = 0xff & *p++; switch (c) { case '\a': printf("\a"); break; case '\b': printf("\b"); break; case '\n': printf("\n"); break; case '\r': printf("\r"); break; default: if (c >= 32 && c < 127) logprintf("%c", c); else logprintf("\\x%02X", c); } } logprintf("\""); logprintf(" (length %d)", q == NULL ? -1 : (int)strlen(q)); /* * If the current string is UTF-8, dump its code points. */ if (utf8) { size_t len; uint32_t uc; int n; int cnt = 0; p = q; len = strlen(p); logprintf(" ["); while ((n = _utf8_to_unicode(&uc, p, len)) > 0) { if (p != q) logprintf(" "); logprintf("%04X", uc); p += n; len -= n; cnt++; } logprintf("]"); logprintf(" (count %d", cnt); if (n < 0) { logprintf(",unknown %d bytes", len); } logprintf(")"); } logprintf("\n"); } /* Verify two strings are equal, dump them if not. */ int assertion_equal_string(const char *file, int line, const char *v1, const char *e1, const char *v2, const char *e2, void *extra, int utf8) { int l1, l2; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && strcmp(v1, v2) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); l1 = strlen(e1); l2 = strlen(e2); if (l1 < l2) l1 = l2; strdump(e1, v1, l1, utf8); strdump(e2, v2, l1, utf8); failure_finish(extra); return (0); } static void wcsdump(const char *e, const wchar_t *w) { logprintf(" %s = ", e); if (w == NULL) { logprintf("(null)"); return; } logprintf("\""); while (*w != L'\0') { unsigned int c = *w++; if (c >= 32 && c < 127) logprintf("%c", c); else if (c < 256) logprintf("\\x%02X", c); else if (c < 0x10000) logprintf("\\u%04X", c); else logprintf("\\U%08X", c); } logprintf("\"\n"); } #ifndef HAVE_WCSCMP static int wcscmp(const wchar_t *s1, const wchar_t *s2) { while (*s1 == *s2++) { if (*s1++ == L'\0') return 0; } if (*s1 > *--s2) return 1; else return -1; } #endif /* Verify that two wide strings are equal, dump them if not. */ int assertion_equal_wstring(const char *file, int line, const wchar_t *v1, const char *e1, const wchar_t *v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); if (v1 != NULL && v2 != NULL && wcscmp(v1, v2) == 0) return (1); failure_start(file, line, "%s != %s", e1, e2); wcsdump(e1, v1); wcsdump(e2, v2); failure_finish(extra); return (0); } /* * Pretty standard hexdump routine. As a bonus, if ref != NULL, then * any bytes in p that differ from ref will be highlighted with '_' * before and after the hex value. */ static void hexdump(const char *p, const char *ref, size_t l, size_t offset) { size_t i, j; char sep; if (p == NULL) { logprintf("(null)\n"); return; } for(i=0; i < l; i+=16) { logprintf("%04x", (unsigned)(i + offset)); sep = ' '; for (j = 0; j < 16 && i + j < l; j++) { if (ref != NULL && p[i + j] != ref[i + j]) sep = '_'; logprintf("%c%02x", sep, 0xff & (int)p[i+j]); if (ref != NULL && p[i + j] == ref[i + j]) sep = ' '; } for (; j < 16; j++) { logprintf("%c ", sep); sep = ' '; } logprintf("%c", sep); for (j=0; j < 16 && i + j < l; j++) { int c = p[i + j]; if (c >= ' ' && c <= 126) logprintf("%c", c); else logprintf("."); } logprintf("\n"); } } /* Verify that two blocks of memory are the same, display the first * block of differences if they're not. */ int assertion_equal_mem(const char *file, int line, const void *_v1, const char *e1, const void *_v2, const char *e2, size_t l, const char *ld, void *extra) { const char *v1 = (const char *)_v1; const char *v2 = (const char *)_v2; size_t offset; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && memcmp(v1, v2, l) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" size %s = %d\n", ld, (int)l); /* Dump 48 bytes (3 lines) so that the first difference is * in the second line. */ offset = 0; while (l > 64 && memcmp(v1, v2, 32) == 0) { /* Two lines agree, so step forward one line. */ v1 += 16; v2 += 16; l -= 16; offset += 16; } logprintf(" Dump of %s\n", e1); hexdump(v1, v2, l < 128 ? l : 128, offset); logprintf(" Dump of %s\n", e2); hexdump(v2, v1, l < 128 ? l : 128, offset); logprintf("\n"); failure_finish(extra); return (0); } /* Verify that the named file exists and is empty. */ int assertion_empty_file(const char *filename, int line, const char *f1) { char buff[1024]; struct stat st; ssize_t s; FILE *f; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) return (1); failure_start(filename, line, "File should be empty: %s", f1); logprintf(" File size: %d\n", (int)st.st_size); logprintf(" Contents:\n"); f = fopen(f1, "rb"); if (f == NULL) { logprintf(" Unable to open %s\n", f1); } else { s = ((off_t)sizeof(buff) < st.st_size) ? (ssize_t)sizeof(buff) : (ssize_t)st.st_size; s = fread(buff, 1, s, f); hexdump(buff, NULL, s, 0); fclose(f); } failure_finish(NULL); return (0); } /* Verify that the named file exists and is not empty. */ int assertion_non_empty_file(const char *filename, int line, const char *f1) { struct stat st; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) { failure_start(filename, line, "File empty: %s", f1); failure_finish(NULL); return (0); } return (1); } /* Verify that two files have the same contents. */ /* TODO: hexdump the first bytes that actually differ. */ int assertion_equal_file(const char *filename, int line, const char *fn1, const char *fn2) { char buff1[1024]; char buff2[1024]; FILE *f1, *f2; int n1, n2; assertion_count(filename, line); f1 = fopen(fn1, "rb"); f2 = fopen(fn2, "rb"); for (;;) { n1 = fread(buff1, 1, sizeof(buff1), f1); n2 = fread(buff2, 1, sizeof(buff2), f2); if (n1 != n2) break; if (n1 == 0 && n2 == 0) { fclose(f1); fclose(f2); return (1); } if (memcmp(buff1, buff2, n1) != 0) break; } fclose(f1); fclose(f2); failure_start(filename, line, "Files not identical"); logprintf(" file1=\"%s\"\n", fn1); logprintf(" file2=\"%s\"\n", fn2); failure_finish(NULL); return (0); } /* Verify that the named file does exist. */ int assertion_file_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (!_access(f, 0)) return (1); #else if (!access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should exist: %s", f); failure_finish(NULL); return (0); } /* Verify that the named file doesn't exist. */ int assertion_file_not_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (_access(f, 0)) return (1); #else if (access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should not exist: %s", f); failure_finish(NULL); return (0); } /* Compare the contents of a file to a block of memory. */ int assertion_file_contents(const char *filename, int line, const void *buff, int s, const char *fn) { char *contents; FILE *f; int n; assertion_count(filename, line); f = fopen(fn, "rb"); if (f == NULL) { failure_start(filename, line, "File should exist: %s", fn); failure_finish(NULL); return (0); } contents = malloc(s * 2); n = fread(contents, 1, s * 2, f); fclose(f); if (n == s && memcmp(buff, contents, s) == 0) { free(contents); return (1); } failure_start(filename, line, "File contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) hexdump(contents, buff, n > 512 ? 512 : n, 0); else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s > 512 ? 512 : s, 0); } failure_finish(NULL); free(contents); return (0); } /* Check the contents of a text file, being tolerant of line endings. */ int assertion_text_file_contents(const char *filename, int line, const char *buff, const char *fn) { char *contents; const char *btxt, *ftxt; FILE *f; int n, s; assertion_count(filename, line); f = fopen(fn, "r"); if (f == NULL) { failure_start(filename, line, "File doesn't exist: %s", fn); failure_finish(NULL); return (0); } s = strlen(buff); contents = malloc(s * 2 + 128); n = fread(contents, 1, s * 2 + 128 - 1, f); if (n >= 0) contents[n] = '\0'; fclose(f); /* Compare texts. */ btxt = buff; ftxt = (const char *)contents; while (*btxt != '\0' && *ftxt != '\0') { if (*btxt == *ftxt) { ++btxt; ++ftxt; continue; } if (btxt[0] == '\n' && ftxt[0] == '\r' && ftxt[1] == '\n') { /* Pass over different new line characters. */ ++btxt; ftxt += 2; continue; } break; } if (*btxt == '\0' && *ftxt == '\0') { free(contents); return (1); } failure_start(filename, line, "Contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) { hexdump(contents, buff, n, 0); logprintf(" expected\n", fn); hexdump(buff, contents, s, 0); } else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s, 0); } failure_finish(NULL); free(contents); return (0); } /* Verify that a text file contains the specified lines, regardless of order */ /* This could be more efficient if we sorted both sets of lines, etc, but * since this is used only for testing and only ever deals with a dozen or so * lines at a time, this relatively crude approach is just fine. */ int assertion_file_contains_lines_any_order(const char *file, int line, const char *pathname, const char *lines[]) { char *buff; size_t buff_size; size_t expected_count, actual_count, i, j; char **expected; char *p, **actual; char c; int expected_failure = 0, actual_failure = 0; assertion_count(file, line); buff = slurpfile(&buff_size, "%s", pathname); if (buff == NULL) { failure_start(pathname, line, "Can't read file: %s", pathname); failure_finish(NULL); return (0); } /* Make a copy of the provided lines and count up the expected file size. */ expected_count = 0; for (i = 0; lines[i] != NULL; ++i) { } expected_count = i; expected = malloc(sizeof(char *) * expected_count); for (i = 0; lines[i] != NULL; ++i) { expected[i] = strdup(lines[i]); } /* Break the file into lines */ actual_count = 0; for (c = '\0', p = buff; p < buff + buff_size; ++p) { if (*p == '\x0d' || *p == '\x0a') *p = '\0'; if (c == '\0' && *p != '\0') ++actual_count; c = *p; } actual = malloc(sizeof(char *) * actual_count); for (j = 0, p = buff; p < buff + buff_size; p += 1 + strlen(p)) { if (*p != '\0') { actual[j] = p; ++j; } } /* Erase matching lines from both lists */ for (i = 0; i < expected_count; ++i) { if (expected[i] == NULL) continue; for (j = 0; j < actual_count; ++j) { if (actual[j] == NULL) continue; if (strcmp(expected[i], actual[j]) == 0) { free(expected[i]); expected[i] = NULL; actual[j] = NULL; break; } } } /* If there's anything left, it's a failure */ for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) ++expected_failure; } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) ++actual_failure; } if (expected_failure == 0 && actual_failure == 0) { free(buff); free(expected); free(actual); return (1); } failure_start(file, line, "File doesn't match: %s", pathname); for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) { logprintf(" Expected but not present: %s\n", expected[i]); free(expected[i]); } } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) logprintf(" Present but not expected: %s\n", actual[j]); } failure_finish(NULL); free(buff); free(expected); free(actual); return (0); } /* Test that two paths point to the same file. */ /* As a side-effect, asserts that both files exist. */ static int is_hardlink(const char *file, int line, const char *path1, const char *path2) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi1, bhfi2; int r; assertion_count(file, line); r = my_GetFileInformationByName(path1, &bhfi1); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path1); failure_finish(NULL); return (0); } r = my_GetFileInformationByName(path2, &bhfi2); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path2); failure_finish(NULL); return (0); } return (bhfi1.dwVolumeSerialNumber == bhfi2.dwVolumeSerialNumber && bhfi1.nFileIndexHigh == bhfi2.nFileIndexHigh && bhfi1.nFileIndexLow == bhfi2.nFileIndexLow); #else struct stat st1, st2; int r; assertion_count(file, line); r = lstat(path1, &st1); if (r != 0) { failure_start(file, line, "File should exist: %s", path1); failure_finish(NULL); return (0); } r = lstat(path2, &st2); if (r != 0) { failure_start(file, line, "File should exist: %s", path2); failure_finish(NULL); return (0); } return (st1.st_ino == st2.st_ino && st1.st_dev == st2.st_dev); #endif } int assertion_is_hardlink(const char *file, int line, const char *path1, const char *path2) { if (is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s are not hardlinked", path1, path2); failure_finish(NULL); return (0); } int assertion_is_not_hardlink(const char *file, int line, const char *path1, const char *path2) { if (!is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s should not be hardlinked", path1, path2); failure_finish(NULL); return (0); } /* Verify a/b/mtime of 'pathname'. */ /* If 'recent', verify that it's within last 10 seconds. */ static int assertion_file_time(const char *file, int line, const char *pathname, long t, long nsec, char type, int recent) { long long filet, filet_nsec; int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define EPOC_TIME (116444736000000000ULL) FILETIME ftime, fbirthtime, fatime, fmtime; ULARGE_INTEGER wintm; HANDLE h; ftime.dwLowDateTime = 0; ftime.dwHighDateTime = 0; assertion_count(file, line); /* Note: FILE_FLAG_BACKUP_SEMANTICS applies to open * a directory file. If not, CreateFile() will fail when * the pathname is a directory. */ h = CreateFile(pathname, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } r = GetFileTime(h, &fbirthtime, &fatime, &fmtime); switch (type) { case 'a': ftime = fatime; break; case 'b': ftime = fbirthtime; break; case 'm': ftime = fmtime; break; } CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't GetFileTime %s\n", pathname); failure_finish(NULL); return (0); } wintm.LowPart = ftime.dwLowDateTime; wintm.HighPart = ftime.dwHighDateTime; filet = (wintm.QuadPart - EPOC_TIME) / 10000000; filet_nsec = ((wintm.QuadPart - EPOC_TIME) % 10000000) * 100; nsec = (nsec / 100) * 100; /* Round the request */ #else struct stat st; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } switch (type) { case 'a': filet = st.st_atime; break; case 'm': filet = st.st_mtime; break; case 'b': filet = 0; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } #if defined(__FreeBSD__) switch (type) { case 'a': filet_nsec = st.st_atimespec.tv_nsec; break; case 'b': filet = st.st_birthtime; filet_nsec = st.st_birthtimespec.tv_nsec; break; case 'm': filet_nsec = st.st_mtimespec.tv_nsec; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } /* FreeBSD generally only stores to microsecond res, so round. */ filet_nsec = (filet_nsec / 1000) * 1000; nsec = (nsec / 1000) * 1000; #else filet_nsec = nsec = 0; /* Generic POSIX only has whole seconds. */ if (type == 'b') return (1); /* Generic POSIX doesn't have birthtime */ #if defined(__HAIKU__) if (type == 'a') return (1); /* Haiku doesn't have atime. */ #endif #endif #endif if (recent) { /* Check that requested time is up-to-date. */ time_t now = time(NULL); if (filet < now - 10 || filet > now + 1) { failure_start(file, line, "File %s has %ctime %lld, %lld seconds ago\n", pathname, type, filet, now - filet); failure_finish(NULL); return (0); } } else if (filet != t || filet_nsec != nsec) { failure_start(file, line, "File %s has %ctime %lld.%09lld, expected %lld.%09lld", pathname, type, filet, filet_nsec, t, nsec); failure_finish(NULL); return (0); } return (1); } /* Verify atime of 'pathname'. */ int assertion_file_atime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'a', 0); } /* Verify atime of 'pathname' is up-to-date. */ int assertion_file_atime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'a', 1); } /* Verify birthtime of 'pathname'. */ int assertion_file_birthtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'b', 0); } /* Verify birthtime of 'pathname' is up-to-date. */ int assertion_file_birthtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'b', 1); } /* Verify mtime of 'pathname'. */ int assertion_file_mtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'm', 0); } /* Verify mtime of 'pathname' is up-to-date. */ int assertion_file_mtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'm', 1); } /* Verify number of links to 'pathname'. */ int assertion_file_nlinks(const char *file, int line, const char *pathname, int nlinks) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; assertion_count(file, line); r = my_GetFileInformationByName(pathname, &bhfi); if (r != 0 && bhfi.nNumberOfLinks == (DWORD)nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, bhfi.nNumberOfLinks, nlinks); failure_finish(NULL); return (0); #else struct stat st; int r; assertion_count(file, line); r = lstat(pathname, &st); - if (r == 0 && st.st_nlink == nlinks) + if (r == 0 && (int)st.st_nlink == nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, st.st_nlink, nlinks); failure_finish(NULL); return (0); #endif } /* Verify size of 'pathname'. */ int assertion_file_size(const char *file, int line, const char *pathname, long size) { int64_t filesize; int r; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) { BY_HANDLE_FILE_INFORMATION bhfi; r = !my_GetFileInformationByName(pathname, &bhfi); filesize = ((int64_t)bhfi.nFileSizeHigh << 32) + bhfi.nFileSizeLow; } #else { struct stat st; r = lstat(pathname, &st); filesize = st.st_size; } #endif if (r == 0 && filesize == size) return (1); failure_start(file, line, "File %s has size %ld, expected %ld", pathname, (long)filesize, (long)size); failure_finish(NULL); return (0); } /* Assert that 'pathname' is a dir. If mode >= 0, verify that too. */ int assertion_is_dir(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Dir should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISDIR(st.st_mode)) { failure_start(file, line, "%s is not a dir", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "Dir %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Verify that 'pathname' is a regular file. If 'mode' is >= 0, * verify that too. */ int assertion_is_reg(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0 || !S_ISREG(st.st_mode)) { failure_start(file, line, "File should exist: %s", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "File %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Check whether 'pathname' is a symbolic link. If 'contents' is * non-NULL, verify that the symlink has those contents. */ static int is_symlink(const char *file, int line, const char *pathname, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) (void)pathname; /* UNUSED */ (void)contents; /* UNUSED */ assertion_count(file, line); /* Windows sort-of has real symlinks, but they're only usable * by privileged users and are crippled even then, so there's * really not much point in bothering with this. */ return (0); #else char buff[300]; struct stat st; ssize_t linklen; int r; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Symlink should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISLNK(st.st_mode)) return (0); if (contents == NULL) return (1); linklen = readlink(pathname, buff, sizeof(buff)); if (linklen < 0) { failure_start(file, line, "Can't read symlink %s", pathname); failure_finish(NULL); return (0); } buff[linklen] = '\0'; if (strcmp(buff, contents) != 0) return (0); return (1); #endif } /* Assert that path is a symlink that (optionally) contains contents. */ int assertion_is_symlink(const char *file, int line, const char *path, const char *contents) { if (is_symlink(file, line, path, contents)) return (1); if (contents) failure_start(file, line, "File %s is not a symlink to %s", path, contents); else failure_start(file, line, "File %s is not a symlink", path); failure_finish(NULL); return (0); } /* Create a directory and report any errors. */ int assertion_make_dir(const char *file, int line, const char *dirname, int mode) { assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ if (0 == _mkdir(dirname)) return (1); #else if (0 == mkdir(dirname, mode)) return (1); #endif failure_start(file, line, "Could not create directory %s", dirname); failure_finish(NULL); return(0); } /* Create a file with the specified contents and report any failures. */ int assertion_make_file(const char *file, int line, const char *path, int mode, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) /* TODO: Rework this to set file mode as well. */ FILE *f; (void)mode; /* UNUSED */ assertion_count(file, line); f = fopen(path, "wb"); if (f == NULL) { failure_start(file, line, "Could not create file %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if (strlen(contents) != fwrite(contents, 1, strlen(contents), f)) { fclose(f); failure_start(file, line, "Could not write file %s", path); failure_finish(NULL); return (0); } } fclose(f); return (1); #else int fd; assertion_count(file, line); fd = open(path, O_CREAT | O_WRONLY, mode >= 0 ? mode : 0644); if (fd < 0) { failure_start(file, line, "Could not create %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if ((ssize_t)strlen(contents) != write(fd, contents, strlen(contents))) { close(fd); failure_start(file, line, "Could not write to %s", path); failure_finish(NULL); return (0); } } close(fd); return (1); #endif } /* Create a hardlink and report any failures. */ int assertion_make_hardlink(const char *file, int line, const char *newpath, const char *linkto) { int succeeded; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) succeeded = my_CreateHardLinkA(newpath, linkto); #elif HAVE_LINK succeeded = !link(linkto, newpath); #else succeeded = 0; #endif if (succeeded) return (1); failure_start(file, line, "Could not create hardlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Create a symlink and report any failures. */ int assertion_make_symlink(const char *file, int line, const char *newpath, const char *linkto) { #if defined(_WIN32) && !defined(__CYGWIN__) int targetIsDir = 0; /* TODO: Fix this */ assertion_count(file, line); if (my_CreateSymbolicLinkA(newpath, linkto, targetIsDir)) return (1); #elif HAVE_SYMLINK assertion_count(file, line); if (0 == symlink(linkto, newpath)) return (1); #endif failure_start(file, line, "Could not create symlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Set umask, report failures. */ int assertion_umask(const char *file, int line, int mask) { assertion_count(file, line); (void)file; /* UNUSED */ (void)line; /* UNUSED */ umask(mask); return (1); } /* Set times, report failures. */ int assertion_utimes(const char *file, int line, const char *pathname, long at, long at_nsec, long mt, long mt_nsec) { int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define WINTIME(sec, nsec) ((Int32x32To64(sec, 10000000) + EPOC_TIME)\ + (((nsec)/1000)*10)) HANDLE h; ULARGE_INTEGER wintm; FILETIME fatime, fmtime; FILETIME *pat, *pmt; assertion_count(file, line); h = CreateFileA(pathname,GENERIC_READ | GENERIC_WRITE, FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } if (at > 0 || at_nsec > 0) { wintm.QuadPart = WINTIME(at, at_nsec); fatime.dwLowDateTime = wintm.LowPart; fatime.dwHighDateTime = wintm.HighPart; pat = &fatime; } else pat = NULL; if (mt > 0 || mt_nsec > 0) { wintm.QuadPart = WINTIME(mt, mt_nsec); fmtime.dwLowDateTime = wintm.LowPart; fmtime.dwHighDateTime = wintm.HighPart; pmt = &fmtime; } else pmt = NULL; if (pat != NULL || pmt != NULL) r = SetFileTime(h, NULL, pat, pmt); else r = 1; CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't SetFileTime %s\n", pathname); failure_finish(NULL); return (0); } return (1); #else /* defined(_WIN32) && !defined(__CYGWIN__) */ struct stat st; struct timeval times[2]; #if !defined(__FreeBSD__) mt_nsec = at_nsec = 0; /* Generic POSIX only has whole seconds. */ #endif if (mt == 0 && mt_nsec == 0 && at == 0 && at_nsec == 0) return (1); r = lstat(pathname, &st); if (r < 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } if (mt == 0 && mt_nsec == 0) { mt = st.st_mtime; #if defined(__FreeBSD__) mt_nsec = st.st_mtimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ mt_nsec = (mt_nsec / 1000) * 1000; #endif } if (at == 0 && at_nsec == 0) { at = st.st_atime; #if defined(__FreeBSD__) at_nsec = st.st_atimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ at_nsec = (at_nsec / 1000) * 1000; #endif } times[1].tv_sec = mt; times[1].tv_usec = mt_nsec / 1000; times[0].tv_sec = at; times[0].tv_usec = at_nsec / 1000; #ifdef HAVE_LUTIMES r = lutimes(pathname, times); #else r = utimes(pathname, times); #endif if (r < 0) { failure_start(file, line, "Can't utimes %s\n", pathname); failure_finish(NULL); return (0); } return (1); #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ } /* * * UTILITIES for use by tests. * */ /* * Check whether platform supports symlinks. This is intended * for tests to use in deciding whether to bother testing symlink * support; if the platform doesn't support symlinks, there's no point * in checking whether the program being tested can create them. * * Note that the first time this test is called, we actually go out to * disk to create and verify a symlink. This is necessary because * symlink support is actually a property of a particular filesystem * and can thus vary between directories on a single system. After * the first call, this returns the cached result from memory, so it's * safe to call it as often as you wish. */ int canSymlink(void) { /* Remember the test result */ static int value = 0, tested = 0; if (tested) return (value); ++tested; assertion_make_file(__FILE__, __LINE__, "canSymlink.0", 0644, "a"); /* Note: Cygwin has its own symlink() emulation that does not * use the Win32 CreateSymbolicLink() function. */ #if defined(_WIN32) && !defined(__CYGWIN__) value = my_CreateSymbolicLinkA("canSymlink.1", "canSymlink.0", 0) && is_symlink(__FILE__, __LINE__, "canSymlink.1", "canSymlink.0"); #elif HAVE_SYMLINK value = (0 == symlink("canSymlink.0", "canSymlink.1")) && is_symlink(__FILE__, __LINE__, "canSymlink.1","canSymlink.0"); #endif return (value); } /* * Can this platform run the gzip program? */ /* Platform-dependent options for hiding the output of a subcommand. */ #if defined(_WIN32) && !defined(__CYGWIN__) static const char *redirectArgs = ">NUL 2>NUL"; /* Win32 cmd.exe */ #else static const char *redirectArgs = ">/dev/null 2>/dev/null"; /* POSIX 'sh' */ #endif int canGzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the gunzip program? */ int canGunzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gunzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Sleep as needed; useful for verifying disk timestamp changes by * ensuring that the wall-clock time has actually changed before we * go back to re-read something from disk. */ void sleepUntilAfter(time_t t) { while (t >= time(NULL)) #if defined(_WIN32) && !defined(__CYGWIN__) Sleep(500); #else sleep(1); #endif } /* * Call standard system() call, but build up the command line using * sprintf() conventions. */ int systemf(const char *fmt, ...) { char buff[8192]; va_list ap; int r; va_start(ap, fmt); vsprintf(buff, fmt, ap); if (verbosity > VERBOSITY_FULL) logprintf("Cmd: %s\n", buff); r = system(buff); va_end(ap); return (r); } /* * Slurp a file into memory for ease of comparison and testing. * Returns size of file in 'sizep' if non-NULL, null-terminates * data in memory for ease of use. */ char * slurpfile(size_t * sizep, const char *fmt, ...) { char filename[8192]; struct stat st; va_list ap; char *p; ssize_t bytes_read; FILE *f; int r; va_start(ap, fmt); vsprintf(filename, fmt, ap); va_end(ap); f = fopen(filename, "rb"); if (f == NULL) { /* Note: No error; non-existent file is okay here. */ return (NULL); } r = fstat(fileno(f), &st); if (r != 0) { logprintf("Can't stat file %s\n", filename); fclose(f); return (NULL); } p = malloc((size_t)st.st_size + 1); if (p == NULL) { logprintf("Can't allocate %ld bytes of memory to read file %s\n", (long int)st.st_size, filename); fclose(f); return (NULL); } bytes_read = fread(p, 1, (size_t)st.st_size, f); if (bytes_read < st.st_size) { logprintf("Can't read file %s\n", filename); fclose(f); free(p); return (NULL); } p[st.st_size] = '\0'; if (sizep != NULL) *sizep = (size_t)st.st_size; fclose(f); return (p); } /* Read a uuencoded file from the reference directory, decode, and * write the result into the current directory. */ #define UUDECODE(c) (((c) - 0x20) & 0x3f) void extract_reference_file(const char *name) { char buff[1024]; FILE *in, *out; sprintf(buff, "%s/%s.uu", refdir, name); in = fopen(buff, "r"); failure("Couldn't open reference file %s", buff); assert(in != NULL); if (in == NULL) return; /* Read up to and including the 'begin' line. */ for (;;) { if (fgets(buff, sizeof(buff), in) == NULL) { /* TODO: This is a failure. */ return; } if (memcmp(buff, "begin ", 6) == 0) break; } /* Now, decode the rest and write it. */ /* Not a lot of error checking here; the input better be right. */ out = fopen(name, "wb"); while (fgets(buff, sizeof(buff), in) != NULL) { char *p = buff; int bytes; if (memcmp(buff, "end", 3) == 0) break; bytes = UUDECODE(*p++); while (bytes > 0) { int n = 0; /* Write out 1-3 bytes from that. */ if (bytes > 0) { n = UUDECODE(*p++) << 18; n |= UUDECODE(*p++) << 12; fputc(n >> 16, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++) << 6; fputc((n >> 8) & 0xFF, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++); fputc(n & 0xFF, out); --bytes; } } } fclose(out); fclose(in); } int is_LargeInode(const char *file) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; r = my_GetFileInformationByName(file, &bhfi); if (r != 0) return (0); return (bhfi.nFileIndexHigh & 0x0000FFFFUL); #else struct stat st; int64_t ino; if (stat(file, &st) < 0) return (0); ino = (int64_t)st.st_ino; return (ino > 0xffffffff); #endif } /* * * TEST management * */ /* * "list.h" is simply created by "grep DEFINE_TEST test_*.c"; it has * a line like * DEFINE_TEST(test_function) * for each test. */ /* Use "list.h" to declare all of the test functions. */ #undef DEFINE_TEST #define DEFINE_TEST(name) void name(void); #include "list.h" /* Use "list.h" to create a list of all tests (functions and names). */ #undef DEFINE_TEST #define DEFINE_TEST(n) { n, #n, 0 }, struct { void (*func)(void); const char *name; int failures; } tests[] = { #include "list.h" }; /* * Summarize repeated failures in the just-completed test. */ static void test_summarize(int failed) { unsigned int i; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: printf(failed ? "E" : "."); fflush(stdout); break; case VERBOSITY_PASSFAIL: printf(failed ? "FAIL\n" : "ok\n"); break; } log_console = (verbosity == VERBOSITY_LIGHT_REPORT); for (i = 0; i < sizeof(failed_lines)/sizeof(failed_lines[0]); i++) { if (failed_lines[i].count > 1 && !failed_lines[i].skip) logprintf("%s:%d: Summary: Failed %d times\n", failed_filename, i, failed_lines[i].count); } /* Clear the failure history for the next file. */ failed_filename = NULL; memset(failed_lines, 0, sizeof(failed_lines)); } /* * Actually run a single test, with appropriate setup and cleanup. */ static int test_run(int i, const char *tmpdir) { char workdir[1024]; char logfilename[64]; int failures_before = failures; int oldumask; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: /* No per-test reports at all */ break; case VERBOSITY_PASSFAIL: /* rest of line will include ok/FAIL marker */ printf("%3d: %-50s", i, tests[i].name); fflush(stdout); break; default: /* Title of test, details will follow */ printf("%3d: %s\n", i, tests[i].name); } /* Chdir to the top-level work directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Can't chdir to top work dir %s\n", tmpdir); exit(1); } /* Create a log file for this test. */ sprintf(logfilename, "%s.log", tests[i].name); logfile = fopen(logfilename, "w"); fprintf(logfile, "%s\n\n", tests[i].name); /* Chdir() to a work dir for this specific test. */ snprintf(workdir, sizeof(workdir), "%s/%s", tmpdir, tests[i].name); testworkdir = workdir; if (!assertMakeDir(testworkdir, 0755) || !assertChdir(testworkdir)) { fprintf(stderr, "ERROR: Can't chdir to work dir %s\n", testworkdir); exit(1); } /* Explicitly reset the locale before each test. */ setlocale(LC_ALL, "C"); /* Record the umask before we run the test. */ umask(oldumask = umask(0)); /* * Run the actual test. */ (*tests[i].func)(); /* * Clean up and report afterwards. */ testworkdir = NULL; /* Restore umask */ umask(oldumask); /* Reset locale. */ setlocale(LC_ALL, "C"); /* Reset directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Couldn't chdir to temp dir %s\n", tmpdir); exit(1); } /* Report per-test summaries. */ tests[i].failures = failures - failures_before; test_summarize(tests[i].failures); /* Close the per-test log file. */ fclose(logfile); logfile = NULL; /* If there were no failures, we can remove the work dir and logfile. */ if (tests[i].failures == 0) { if (!keep_temp_files && assertChdir(tmpdir)) { #if defined(_WIN32) && !defined(__CYGWIN__) /* Make sure not to leave empty directories. * Sometimes a processing of closing files used by tests * is not done, then rmdir will be failed and it will * leave a empty test directory. So we should wait a few * seconds and retry rmdir. */ int r, t; for (t = 0; t < 10; t++) { if (t > 0) Sleep(1000); r = systemf("rmdir /S /Q %s", tests[i].name); if (r == 0) break; } systemf("del %s", logfilename); #else systemf("rm -rf %s", tests[i].name); systemf("rm %s", logfilename); #endif } } /* Return appropriate status. */ return (tests[i].failures); } /* * * * MAIN and support routines. * * */ static void usage(const char *program) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i; printf("Usage: %s [options] ...\n", program); printf("Default is to run all tests.\n"); printf("Otherwise, specify the numbers of the tests you wish to run.\n"); printf("Options:\n"); printf(" -d Dump core after any failure, for debugging.\n"); printf(" -k Keep all temp files.\n"); printf(" Default: temp files for successful tests deleted.\n"); #ifdef PROGRAM printf(" -p Path to executable to be tested.\n"); printf(" Default: path taken from " ENVBASE " environment variable.\n"); #endif printf(" -q Quiet.\n"); printf(" -r Path to dir containing reference files.\n"); printf(" Default: Current directory.\n"); printf(" -u Keep running specifies tests until one fails.\n"); printf(" -v Verbose.\n"); printf("Available tests:\n"); for (i = 0; i < limit; i++) printf(" %d: %s\n", i, tests[i].name); exit(1); } static char * get_refdir(const char *d) { char tried[512] = { '\0' }; char buff[128]; char *pwd, *p; /* If a dir was specified, try that */ if (d != NULL) { pwd = NULL; snprintf(buff, sizeof(buff), "%s", d); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); goto failure; } /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; /* Look for a known file. */ snprintf(buff, sizeof(buff), "%s", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(LIBRARY) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, LIBRARY); #else snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM); #endif p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(PROGRAM_ALIAS) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM_ALIAS); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #endif if (memcmp(pwd, "/usr/obj", 8) == 0) { snprintf(buff, sizeof(buff), "%s", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); } failure: printf("Unable to locate known reference file %s\n", KNOWNREF); printf(" Checked following directories:\n%s\n", tried); #if defined(_WIN32) && !defined(__CYGWIN__) && defined(_DEBUG) DebugBreak(); #endif exit(1); success: free(p); free(pwd); return strdup(buff); } int main(int argc, char **argv) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i = 0, j = 0, start, end, tests_run = 0, tests_failed = 0, option; time_t now; char *refdir_alloc = NULL; const char *progname; char **saved_argv; const char *tmp, *option_arg, *p; char tmpdir[256], *pwd, *testprogdir, *tmp2 = NULL; char tmpdir_timestamp[256]; (void)argc; /* UNUSED */ /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; #if defined(HAVE__CrtSetReportMode) /* To stop to run the default invalid parameter handler. */ _set_invalid_parameter_handler(invalid_parameter_handler); /* Disable annoying assertion message box. */ _CrtSetReportMode(_CRT_ASSERT, 0); #endif /* * Name of this program, used to build root of our temp directory * tree. */ progname = p = argv[0]; if ((testprogdir = (char *)malloc(strlen(progname) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir, progname); while (*p != '\0') { /* Support \ or / dir separators for Windows compat. */ if (*p == '/' || *p == '\\') { progname = p + 1; i = j; } ++p; j++; } testprogdir[i] = '\0'; if (testprogdir[0] != '/') { /* Fixup path for relative directories. */ if ((testprogdir = (char *)realloc(testprogdir, strlen(pwd) + 1 + strlen(testprogdir) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir + strlen(pwd) + 1, testprogdir); strcpy(testprogdir, pwd); testprogdir[strlen(pwd)] = '/'; } #ifdef PROGRAM /* Get the target program from environment, if available. */ testprogfile = getenv(ENVBASE); #endif if (getenv("TMPDIR") != NULL) tmp = getenv("TMPDIR"); else if (getenv("TMP") != NULL) tmp = getenv("TMP"); else if (getenv("TEMP") != NULL) tmp = getenv("TEMP"); else if (getenv("TEMPDIR") != NULL) tmp = getenv("TEMPDIR"); else tmp = "/tmp"; /* Allow -d to be controlled through the environment. */ if (getenv(ENVBASE "_DEBUG") != NULL) dump_on_failure = 1; /* Get the directory holding test files from environment. */ refdir = getenv(ENVBASE "_TEST_FILES"); /* * Parse options, without using getopt(), which isn't available * on all platforms. */ ++argv; /* Skip program name */ while (*argv != NULL) { if (**argv != '-') break; p = *argv++; ++p; /* Skip '-' */ while (*p != '\0') { option = *p++; option_arg = NULL; /* If 'opt' takes an argument, parse that. */ if (option == 'p' || option == 'r') { if (*p != '\0') option_arg = p; else if (*argv == NULL) { fprintf(stderr, "Option -%c requires argument.\n", option); usage(progname); } else option_arg = *argv++; p = ""; /* End of this option word. */ } /* Now, handle the option. */ switch (option) { case 'd': dump_on_failure = 1; break; case 'k': keep_temp_files = 1; break; case 'p': #ifdef PROGRAM testprogfile = option_arg; #else fprintf(stderr, "-p option not permitted\n"); usage(progname); #endif break; case 'q': verbosity--; break; case 'r': refdir = option_arg; break; case 'u': until_failure++; break; case 'v': verbosity++; break; default: fprintf(stderr, "Unrecognized option '%c'\n", option); usage(progname); } } } /* * Sanity-check that our options make sense. */ #ifdef PROGRAM if (testprogfile == NULL) { if ((tmp2 = (char *)malloc(strlen(testprogdir) + 1 + strlen(PROGRAM) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(tmp2, testprogdir); strcat(tmp2, "/"); strcat(tmp2, PROGRAM); testprogfile = tmp2; } { char *testprg; #if defined(_WIN32) && !defined(__CYGWIN__) /* Command.com sometimes rejects '/' separators. */ testprg = strdup(testprogfile); for (i = 0; testprg[i] != '\0'; i++) { if (testprg[i] == '/') testprg[i] = '\\'; } testprogfile = testprg; #endif /* Quote the name that gets put into shell command lines. */ testprg = malloc(strlen(testprogfile) + 3); strcpy(testprg, "\""); strcat(testprg, testprogfile); strcat(testprg, "\""); testprog = testprg; } #endif #if !defined(_WIN32) && defined(SIGPIPE) { /* Ignore SIGPIPE signals */ struct sigaction sa; sa.sa_handler = SIG_IGN; sigemptyset(&sa.sa_mask); sa.sa_flags = 0; sigaction(SIGPIPE, &sa, NULL); } #endif /* * Create a temp directory for the following tests. * Include the time the tests started as part of the name, * to make it easier to track the results of multiple tests. */ now = time(NULL); for (i = 0; ; i++) { strftime(tmpdir_timestamp, sizeof(tmpdir_timestamp), "%Y-%m-%dT%H.%M.%S", localtime(&now)); sprintf(tmpdir, "%s/%s.%s-%03d", tmp, progname, tmpdir_timestamp, i); if (assertMakeDir(tmpdir,0755)) break; if (i >= 999) { fprintf(stderr, "ERROR: Unable to create temp directory %s\n", tmpdir); exit(1); } } /* * If the user didn't specify a directory for locating * reference files, try to find the reference files in * the "usual places." */ refdir = refdir_alloc = get_refdir(refdir); /* * Banner with basic information. */ printf("\n"); printf("If tests fail or crash, details will be in:\n"); printf(" %s\n", tmpdir); printf("\n"); if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("Reference files will be read from: %s\n", refdir); #ifdef PROGRAM printf("Running tests on: %s\n", testprog); #endif printf("Exercising: "); fflush(stdout); printf("%s\n", EXTRA_VERSION); } else { printf("Running "); fflush(stdout); } /* * Run some or all of the individual tests. */ saved_argv = argv; do { argv = saved_argv; if (*argv == NULL) { /* Default: Run all tests. */ for (i = 0; i < limit; i++) { tests_run++; if (test_run(i, tmpdir)) { tests_failed++; if (until_failure) goto finish; } } } else { while (*(argv) != NULL) { if (**argv >= '0' && **argv <= '9') { char *vp = *argv; start = 0; while (*vp >= '0' && *vp <= '9') { start *= 10; start += *vp - '0'; ++vp; } if (*vp == '\0') { end = start; } else if (*vp == '-') { ++vp; if (*vp == '\0') { end = limit - 1; } else { end = 0; while (*vp >= '0' && *vp <= '9') { end *= 10; end += *vp - '0'; ++vp; } } } else { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } if (start < 0 || end >= limit || start > end) { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } } else { for (start = 0; start < limit; ++start) { if (strcmp(*argv, tests[start].name) == 0) break; } end = start; if (start >= limit) { printf("*** INVALID Test ``%s''\n", *argv); free(refdir_alloc); usage(progname); /* usage() never returns */ } } while (start <= end) { tests_run++; if (test_run(start, tmpdir)) { tests_failed++; if (until_failure) goto finish; } ++start; } argv++; } } } while (until_failure); finish: /* Must be freed after all tests run */ free(tmp2); free(testprogdir); free(pwd); /* * Report summary statistics. */ if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("\n"); printf("Totals:\n"); printf(" Tests run: %8d\n", tests_run); printf(" Tests failed: %8d\n", tests_failed); printf(" Assertions checked:%8d\n", assertions); printf(" Assertions failed: %8d\n", failures); printf(" Skips reported: %8d\n", skips); } if (failures) { printf("\n"); printf("Failing tests:\n"); for (i = 0; i < limit; ++i) { if (tests[i].failures) printf(" %d: %s (%d failures)\n", i, tests[i].name, tests[i].failures); } printf("\n"); printf("Details for failing tests: %s\n", tmpdir); printf("\n"); } else { if (verbosity == VERBOSITY_SUMMARY_ONLY) printf("\n"); printf("%d tests passed, no failures\n", tests_run); } free(refdir_alloc); /* If the final tmpdir is empty, we can remove it. */ /* This should be the usual case when all tests succeed. */ assertChdir(".."); rmdir(tmpdir); return (tests_failed ? 1 : 0); } Index: vendor/libarchive/dist/libarchive/archive_rb.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_rb.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_rb.c (revision 231297) @@ -1,701 +1,701 @@ /*- * Copyright (c) 2001 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Matt Thomas . * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * Based on: NetBSD: rb.c,v 1.6 2010/04/30 13:58:09 joerg Exp */ #include "archive_platform.h" #include #include "archive_rb.h" /* Keep in sync with archive_rb.h */ #define RB_DIR_LEFT 0 #define RB_DIR_RIGHT 1 #define RB_DIR_OTHER 1 #define rb_left rb_nodes[RB_DIR_LEFT] #define rb_right rb_nodes[RB_DIR_RIGHT] #define RB_FLAG_POSITION 0x2 #define RB_FLAG_RED 0x1 #define RB_FLAG_MASK (RB_FLAG_POSITION|RB_FLAG_RED) #define RB_FATHER(rb) \ ((struct archive_rb_node *)((rb)->rb_info & ~RB_FLAG_MASK)) #define RB_SET_FATHER(rb, father) \ ((void)((rb)->rb_info = (uintptr_t)(father)|((rb)->rb_info & RB_FLAG_MASK))) #define RB_SENTINEL_P(rb) ((rb) == NULL) #define RB_LEFT_SENTINEL_P(rb) RB_SENTINEL_P((rb)->rb_left) #define RB_RIGHT_SENTINEL_P(rb) RB_SENTINEL_P((rb)->rb_right) #define RB_FATHER_SENTINEL_P(rb) RB_SENTINEL_P(RB_FATHER((rb))) #define RB_CHILDLESS_P(rb) \ (RB_SENTINEL_P(rb) || (RB_LEFT_SENTINEL_P(rb) && RB_RIGHT_SENTINEL_P(rb))) #define RB_TWOCHILDREN_P(rb) \ (!RB_SENTINEL_P(rb) && !RB_LEFT_SENTINEL_P(rb) && !RB_RIGHT_SENTINEL_P(rb)) #define RB_POSITION(rb) \ (((rb)->rb_info & RB_FLAG_POSITION) ? RB_DIR_RIGHT : RB_DIR_LEFT) #define RB_RIGHT_P(rb) (RB_POSITION(rb) == RB_DIR_RIGHT) #define RB_LEFT_P(rb) (RB_POSITION(rb) == RB_DIR_LEFT) #define RB_RED_P(rb) (!RB_SENTINEL_P(rb) && ((rb)->rb_info & RB_FLAG_RED) != 0) #define RB_BLACK_P(rb) (RB_SENTINEL_P(rb) || ((rb)->rb_info & RB_FLAG_RED) == 0) #define RB_MARK_RED(rb) ((void)((rb)->rb_info |= RB_FLAG_RED)) #define RB_MARK_BLACK(rb) ((void)((rb)->rb_info &= ~RB_FLAG_RED)) #define RB_INVERT_COLOR(rb) ((void)((rb)->rb_info ^= RB_FLAG_RED)) #define RB_ROOT_P(rbt, rb) ((rbt)->rbt_root == (rb)) #define RB_SET_POSITION(rb, position) \ ((void)((position) ? ((rb)->rb_info |= RB_FLAG_POSITION) : \ ((rb)->rb_info &= ~RB_FLAG_POSITION))) #define RB_ZERO_PROPERTIES(rb) ((void)((rb)->rb_info &= ~RB_FLAG_MASK)) #define RB_COPY_PROPERTIES(dst, src) \ ((void)((dst)->rb_info ^= ((dst)->rb_info ^ (src)->rb_info) & RB_FLAG_MASK)) #define RB_SWAP_PROPERTIES(a, b) do { \ uintptr_t xorinfo = ((a)->rb_info ^ (b)->rb_info) & RB_FLAG_MASK; \ (a)->rb_info ^= xorinfo; \ (b)->rb_info ^= xorinfo; \ } while (/*CONSTCOND*/ 0) static void __archive_rb_tree_insert_rebalance(struct archive_rb_tree *, struct archive_rb_node *); static void __archive_rb_tree_removal_rebalance(struct archive_rb_tree *, struct archive_rb_node *, unsigned int); #define RB_SENTINEL_NODE NULL #define T 1 #define F 0 void __archive_rb_tree_init(struct archive_rb_tree *rbt, const struct archive_rb_tree_ops *ops) { rbt->rbt_ops = ops; - *((const struct archive_rb_node **)&rbt->rbt_root) = RB_SENTINEL_NODE; + *((struct archive_rb_node **)&rbt->rbt_root) = RB_SENTINEL_NODE; } struct archive_rb_node * __archive_rb_tree_find_node(struct archive_rb_tree *rbt, const void *key) { archive_rbto_compare_key_fn compare_key = rbt->rbt_ops->rbto_compare_key; struct archive_rb_node *parent = rbt->rbt_root; while (!RB_SENTINEL_P(parent)) { const signed int diff = (*compare_key)(parent, key); if (diff == 0) return parent; parent = parent->rb_nodes[diff > 0]; } return NULL; } struct archive_rb_node * __archive_rb_tree_find_node_geq(struct archive_rb_tree *rbt, const void *key) { archive_rbto_compare_key_fn compare_key = rbt->rbt_ops->rbto_compare_key; struct archive_rb_node *parent = rbt->rbt_root; struct archive_rb_node *last = NULL; while (!RB_SENTINEL_P(parent)) { const signed int diff = (*compare_key)(parent, key); if (diff == 0) return parent; if (diff < 0) last = parent; parent = parent->rb_nodes[diff > 0]; } return last; } struct archive_rb_node * __archive_rb_tree_find_node_leq(struct archive_rb_tree *rbt, const void *key) { archive_rbto_compare_key_fn compare_key = rbt->rbt_ops->rbto_compare_key; struct archive_rb_node *parent = rbt->rbt_root; struct archive_rb_node *last = NULL; while (!RB_SENTINEL_P(parent)) { const signed int diff = (*compare_key)(parent, key); if (diff == 0) return parent; if (diff > 0) last = parent; parent = parent->rb_nodes[diff > 0]; } return last; } int __archive_rb_tree_insert_node(struct archive_rb_tree *rbt, struct archive_rb_node *self) { archive_rbto_compare_nodes_fn compare_nodes = rbt->rbt_ops->rbto_compare_nodes; struct archive_rb_node *parent, *tmp; unsigned int position; int rebalance; tmp = rbt->rbt_root; /* * This is a hack. Because rbt->rbt_root is just a * struct archive_rb_node *, just like rb_node->rb_nodes[RB_DIR_LEFT], * we can use this fact to avoid a lot of tests for root and know * that even at root, updating * RB_FATHER(rb_node)->rb_nodes[RB_POSITION(rb_node)] will * update rbt->rbt_root. */ parent = (struct archive_rb_node *)(void *)&rbt->rbt_root; position = RB_DIR_LEFT; /* * Find out where to place this new leaf. */ while (!RB_SENTINEL_P(tmp)) { const signed int diff = (*compare_nodes)(tmp, self); if (diff == 0) { /* * Node already exists; don't insert. */ return F; } parent = tmp; position = (diff > 0); tmp = parent->rb_nodes[position]; } /* * Initialize the node and insert as a leaf into the tree. */ RB_SET_FATHER(self, parent); RB_SET_POSITION(self, position); if (parent == (struct archive_rb_node *)(void *)&rbt->rbt_root) { RB_MARK_BLACK(self); /* root is always black */ rebalance = F; } else { /* * All new nodes are colored red. We only need to rebalance * if our parent is also red. */ RB_MARK_RED(self); rebalance = RB_RED_P(parent); } self->rb_left = parent->rb_nodes[position]; self->rb_right = parent->rb_nodes[position]; parent->rb_nodes[position] = self; /* * Rebalance tree after insertion */ if (rebalance) __archive_rb_tree_insert_rebalance(rbt, self); return T; } /* * Swap the location and colors of 'self' and its child @ which. The child * can not be a sentinel node. This is our rotation function. However, * since it preserves coloring, it great simplifies both insertion and * removal since rotation almost always involves the exchanging of colors * as a separate step. */ /*ARGSUSED*/ static void __archive_rb_tree_reparent_nodes( struct archive_rb_node *old_father, const unsigned int which) { const unsigned int other = which ^ RB_DIR_OTHER; struct archive_rb_node * const grandpa = RB_FATHER(old_father); struct archive_rb_node * const old_child = old_father->rb_nodes[which]; struct archive_rb_node * const new_father = old_child; struct archive_rb_node * const new_child = old_father; /* * Exchange descendant linkages. */ grandpa->rb_nodes[RB_POSITION(old_father)] = new_father; new_child->rb_nodes[which] = old_child->rb_nodes[other]; new_father->rb_nodes[other] = new_child; /* * Update ancestor linkages */ RB_SET_FATHER(new_father, grandpa); RB_SET_FATHER(new_child, new_father); /* * Exchange properties between new_father and new_child. The only * change is that new_child's position is now on the other side. */ RB_SWAP_PROPERTIES(new_father, new_child); RB_SET_POSITION(new_child, other); /* * Make sure to reparent the new child to ourself. */ if (!RB_SENTINEL_P(new_child->rb_nodes[which])) { RB_SET_FATHER(new_child->rb_nodes[which], new_child); RB_SET_POSITION(new_child->rb_nodes[which], which); } } static void __archive_rb_tree_insert_rebalance(struct archive_rb_tree *rbt, struct archive_rb_node *self) { struct archive_rb_node * father = RB_FATHER(self); struct archive_rb_node * grandpa; struct archive_rb_node * uncle; unsigned int which; unsigned int other; for (;;) { /* * We are red and our parent is red, therefore we must have a * grandfather and he must be black. */ grandpa = RB_FATHER(father); which = (father == grandpa->rb_right); other = which ^ RB_DIR_OTHER; uncle = grandpa->rb_nodes[other]; if (RB_BLACK_P(uncle)) break; /* * Case 1: our uncle is red * Simply invert the colors of our parent and * uncle and make our grandparent red. And * then solve the problem up at his level. */ RB_MARK_BLACK(uncle); RB_MARK_BLACK(father); if (RB_ROOT_P(rbt, grandpa)) { /* * If our grandpa is root, don't bother * setting him to red, just return. */ return; } RB_MARK_RED(grandpa); self = grandpa; father = RB_FATHER(self); if (RB_BLACK_P(father)) { /* * If our greatgrandpa is black, we're done. */ return; } } /* * Case 2&3: our uncle is black. */ if (self == father->rb_nodes[other]) { /* * Case 2: we are on the same side as our uncle * Swap ourselves with our parent so this case * becomes case 3. Basically our parent becomes our * child. */ __archive_rb_tree_reparent_nodes(father, other); } /* * Case 3: we are opposite a child of a black uncle. * Swap our parent and grandparent. Since our grandfather * is black, our father will become black and our new sibling * (former grandparent) will become red. */ __archive_rb_tree_reparent_nodes(grandpa, which); /* * Final step: Set the root to black. */ RB_MARK_BLACK(rbt->rbt_root); } static void __archive_rb_tree_prune_node(struct archive_rb_tree *rbt, struct archive_rb_node *self, int rebalance) { const unsigned int which = RB_POSITION(self); struct archive_rb_node *father = RB_FATHER(self); /* * Since we are childless, we know that self->rb_left is pointing * to the sentinel node. */ father->rb_nodes[which] = self->rb_left; /* * Rebalance if requested. */ if (rebalance) __archive_rb_tree_removal_rebalance(rbt, father, which); } /* * When deleting an interior node */ static void __archive_rb_tree_swap_prune_and_rebalance(struct archive_rb_tree *rbt, struct archive_rb_node *self, struct archive_rb_node *standin) { const unsigned int standin_which = RB_POSITION(standin); unsigned int standin_other = standin_which ^ RB_DIR_OTHER; struct archive_rb_node *standin_son; struct archive_rb_node *standin_father = RB_FATHER(standin); int rebalance = RB_BLACK_P(standin); if (standin_father == self) { /* * As a child of self, any children would be opposite of * our parent. */ standin_son = standin->rb_nodes[standin_which]; } else { /* * Since we aren't a child of self, any children would be * on the same side as our parent. */ standin_son = standin->rb_nodes[standin_other]; } if (RB_RED_P(standin_son)) { /* * We know we have a red child so if we flip it to black * we don't have to rebalance. */ RB_MARK_BLACK(standin_son); rebalance = F; if (standin_father != self) { /* * Change the son's parentage to point to his grandpa. */ RB_SET_FATHER(standin_son, standin_father); RB_SET_POSITION(standin_son, standin_which); } } if (standin_father == self) { /* * If we are about to delete the standin's father, then when * we call rebalance, we need to use ourselves as our father. * Otherwise remember our original father. Also, since we are * our standin's father we only need to reparent the standin's * brother. * * | R --> S | * | Q S --> Q T | * | t --> | * * Have our son/standin adopt his brother as his new son. */ standin_father = standin; } else { /* * | R --> S . | * | / \ | T --> / \ | / | * | ..... | S --> ..... | T | * * Sever standin's connection to his father. */ standin_father->rb_nodes[standin_which] = standin_son; /* * Adopt the far son. */ standin->rb_nodes[standin_other] = self->rb_nodes[standin_other]; RB_SET_FATHER(standin->rb_nodes[standin_other], standin); /* * Use standin_other because we need to preserve standin_which * for the removal_rebalance. */ standin_other = standin_which; } /* * Move the only remaining son to our standin. If our standin is our * son, this will be the only son needed to be moved. */ standin->rb_nodes[standin_other] = self->rb_nodes[standin_other]; RB_SET_FATHER(standin->rb_nodes[standin_other], standin); /* * Now copy the result of self to standin and then replace * self with standin in the tree. */ RB_COPY_PROPERTIES(standin, self); RB_SET_FATHER(standin, RB_FATHER(self)); RB_FATHER(standin)->rb_nodes[RB_POSITION(standin)] = standin; if (rebalance) __archive_rb_tree_removal_rebalance(rbt, standin_father, standin_which); } /* * We could do this by doing * __archive_rb_tree_node_swap(rbt, self, which); * __archive_rb_tree_prune_node(rbt, self, F); * * But it's more efficient to just evaluate and recolor the child. */ static void __archive_rb_tree_prune_blackred_branch( struct archive_rb_node *self, unsigned int which) { struct archive_rb_node *father = RB_FATHER(self); struct archive_rb_node *son = self->rb_nodes[which]; /* * Remove ourselves from the tree and give our former child our * properties (position, color, root). */ RB_COPY_PROPERTIES(son, self); father->rb_nodes[RB_POSITION(son)] = son; RB_SET_FATHER(son, father); } /* * */ void __archive_rb_tree_remove_node(struct archive_rb_tree *rbt, struct archive_rb_node *self) { struct archive_rb_node *standin; unsigned int which; /* * In the following diagrams, we (the node to be removed) are S. Red * nodes are lowercase. T could be either red or black. * * Remember the major axiom of the red-black tree: the number of * black nodes from the root to each leaf is constant across all * leaves, only the number of red nodes varies. * * Thus removing a red leaf doesn't require any other changes to a * red-black tree. So if we must remove a node, attempt to rearrange * the tree so we can remove a red node. * * The simplest case is a childless red node or a childless root node: * * | T --> T | or | R --> * | * | s --> * | */ if (RB_CHILDLESS_P(self)) { const int rebalance = RB_BLACK_P(self) && !RB_ROOT_P(rbt, self); __archive_rb_tree_prune_node(rbt, self, rebalance); return; } if (!RB_TWOCHILDREN_P(self)) { /* * The next simplest case is the node we are deleting is * black and has one red child. * * | T --> T --> T | * | S --> R --> R | * | r --> s --> * | */ which = RB_LEFT_SENTINEL_P(self) ? RB_DIR_RIGHT : RB_DIR_LEFT; __archive_rb_tree_prune_blackred_branch(self, which); return; } /* * We invert these because we prefer to remove from the inside of * the tree. */ which = RB_POSITION(self) ^ RB_DIR_OTHER; /* * Let's find the node closes to us opposite of our parent * Now swap it with ourself, "prune" it, and rebalance, if needed. */ standin = __archive_rb_tree_iterate(rbt, self, which); __archive_rb_tree_swap_prune_and_rebalance(rbt, self, standin); } static void __archive_rb_tree_removal_rebalance(struct archive_rb_tree *rbt, struct archive_rb_node *parent, unsigned int which) { while (RB_BLACK_P(parent->rb_nodes[which])) { unsigned int other = which ^ RB_DIR_OTHER; struct archive_rb_node *brother = parent->rb_nodes[other]; /* * For cases 1, 2a, and 2b, our brother's children must * be black and our father must be black */ if (RB_BLACK_P(parent) && RB_BLACK_P(brother->rb_left) && RB_BLACK_P(brother->rb_right)) { if (RB_RED_P(brother)) { /* * Case 1: Our brother is red, swap its * position (and colors) with our parent. * This should now be case 2b (unless C or E * has a red child which is case 3; thus no * explicit branch to case 2b). * * B -> D * A d -> b E * C E -> A C */ __archive_rb_tree_reparent_nodes(parent, other); brother = parent->rb_nodes[other]; } else { /* * Both our parent and brother are black. * Change our brother to red, advance up rank * and go through the loop again. * * B -> *B * *A D -> A d * C E -> C E */ RB_MARK_RED(brother); if (RB_ROOT_P(rbt, parent)) return; /* root == parent == black */ which = RB_POSITION(parent); parent = RB_FATHER(parent); continue; } } /* * Avoid an else here so that case 2a above can hit either * case 2b, 3, or 4. */ if (RB_RED_P(parent) && RB_BLACK_P(brother) && RB_BLACK_P(brother->rb_left) && RB_BLACK_P(brother->rb_right)) { /* * We are black, our father is red, our brother and * both nephews are black. Simply invert/exchange the * colors of our father and brother (to black and red * respectively). * * | f --> F | * | * B --> * b | * | N N --> N N | */ RB_MARK_BLACK(parent); RB_MARK_RED(brother); break; /* We're done! */ } else { /* * Our brother must be black and have at least one * red child (it may have two). */ if (RB_BLACK_P(brother->rb_nodes[other])) { /* * Case 3: our brother is black, our near * nephew is red, and our far nephew is black. * Swap our brother with our near nephew. * This result in a tree that matches case 4. * (Our father could be red or black). * * | F --> F | * | x B --> x B | * | n --> n | */ __archive_rb_tree_reparent_nodes(brother, which); brother = parent->rb_nodes[other]; } /* * Case 4: our brother is black and our far nephew * is red. Swap our father and brother locations and * change our far nephew to black. (these can be * done in either order so we change the color first). * The result is a valid red-black tree and is a * terminal case. (again we don't care about the * father's color) * * If the father is red, we will get a red-black-black * tree: * | f -> f --> b | * | B -> B --> F N | * | n -> N --> | * * If the father is black, we will get an all black * tree: * | F -> F --> B | * | B -> B --> F N | * | n -> N --> | * * If we had two red nephews, then after the swap, * our former father would have a red grandson. */ RB_MARK_BLACK(brother->rb_nodes[other]); __archive_rb_tree_reparent_nodes(parent, other); break; /* We're done! */ } } } struct archive_rb_node * __archive_rb_tree_iterate(struct archive_rb_tree *rbt, struct archive_rb_node *self, const unsigned int direction) { const unsigned int other = direction ^ RB_DIR_OTHER; if (self == NULL) { self = rbt->rbt_root; if (RB_SENTINEL_P(self)) return NULL; while (!RB_SENTINEL_P(self->rb_nodes[direction])) self = self->rb_nodes[direction]; return self; } /* * We can't go any further in this direction. We proceed up in the * opposite direction until our parent is in direction we want to go. */ if (RB_SENTINEL_P(self->rb_nodes[direction])) { while (!RB_ROOT_P(rbt, self)) { - if (other == RB_POSITION(self)) + if (other == (unsigned int)RB_POSITION(self)) return RB_FATHER(self); self = RB_FATHER(self); } return NULL; } /* * Advance down one in current direction and go down as far as possible * in the opposite direction. */ self = self->rb_nodes[direction]; while (!RB_SENTINEL_P(self->rb_nodes[other])) self = self->rb_nodes[other]; return self; } Index: vendor/libarchive/dist/libarchive/archive_read_disk_entry_from_file.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_read_disk_entry_from_file.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_read_disk_entry_from_file.c (revision 231297) @@ -1,1005 +1,1005 @@ /*- * Copyright (c) 2003-2009 Tim Kientzle * Copyright (c) 2010 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD: head/lib/libarchive/archive_read_disk_entry_from_file.c 201084 2009-12-28 02:14:09Z kientzle $"); /* This is the tree-walking code for POSIX systems. */ #if !defined(_WIN32) || defined(__CYGWIN__) #ifdef HAVE_SYS_TYPES_H /* Mac OSX requires sys/types.h before sys/acl.h. */ #include #endif #ifdef HAVE_SYS_ACL_H #include #endif #ifdef HAVE_SYS_EXTATTR_H #include #endif #ifdef HAVE_SYS_IOCTL_H #include #endif #ifdef HAVE_SYS_PARAM_H #include #endif #ifdef HAVE_SYS_STAT_H #include #endif #ifdef HAVE_SYS_XATTR_H #include #endif #ifdef HAVE_SYS_EA_H #include #endif #ifdef HAVE_ACL_LIBACL_H #include #endif #ifdef HAVE_ATTR_XATTR_H #include #endif #ifdef HAVE_COPYFILE_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_LINUX_FIEMAP_H #include #endif #ifdef HAVE_LINUX_FS_H #include #endif /* * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h. * As the include guards don't agree, the order of include is important. */ #ifdef HAVE_LINUX_EXT2_FS_H #include /* for Linux file flags */ #endif #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) #include /* Linux file flags, broken on Cygwin */ #endif #ifdef HAVE_PATHS_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #include "archive.h" #include "archive_entry.h" #include "archive_private.h" #include "archive_read_disk_private.h" /* * Linux and FreeBSD plug this obvious hole in POSIX.1e in * different ways. */ #if HAVE_ACL_GET_PERM #define ACL_GET_PERM acl_get_perm #elif HAVE_ACL_GET_PERM_NP #define ACL_GET_PERM acl_get_perm_np #endif static int setup_acls_posix1e(struct archive_read_disk *, struct archive_entry *, int fd); static int setup_mac_metadata(struct archive_read_disk *, struct archive_entry *, int fd); static int setup_xattrs(struct archive_read_disk *, struct archive_entry *, int fd); static int setup_sparse(struct archive_read_disk *, struct archive_entry *, int fd); int archive_read_disk_entry_from_file(struct archive *_a, struct archive_entry *entry, int fd, const struct stat *st) { struct archive_read_disk *a = (struct archive_read_disk *)_a; const char *path, *name; struct stat s; int initial_fd = fd; int r, r1; archive_clear_error(_a); path = archive_entry_sourcepath(entry); if (path == NULL) path = archive_entry_pathname(entry); if (a->tree == NULL) { if (st == NULL) { #if HAVE_FSTAT if (fd >= 0) { if (fstat(fd, &s) != 0) { archive_set_error(&a->archive, errno, "Can't fstat"); return (ARCHIVE_FAILED); } } else #endif #if HAVE_LSTAT if (!a->follow_symlinks) { if (lstat(path, &s) != 0) { archive_set_error(&a->archive, errno, "Can't lstat %s", path); return (ARCHIVE_FAILED); } } else #endif if (stat(path, &s) != 0) { archive_set_error(&a->archive, errno, "Can't stat %s", path); return (ARCHIVE_FAILED); } st = &s; } archive_entry_copy_stat(entry, st); } /* Lookup uname/gname */ name = archive_read_disk_uname(_a, archive_entry_uid(entry)); if (name != NULL) archive_entry_copy_uname(entry, name); name = archive_read_disk_gname(_a, archive_entry_gid(entry)); if (name != NULL) archive_entry_copy_gname(entry, name); #ifdef HAVE_STRUCT_STAT_ST_FLAGS /* On FreeBSD, we get flags for free with the stat. */ /* TODO: Does this belong in copy_stat()? */ if (st->st_flags != 0) archive_entry_set_fflags(entry, st->st_flags, 0); #endif #if defined(EXT2_IOC_GETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS) /* Linux requires an extra ioctl to pull the flags. Although * this is an extra step, it has a nice side-effect: We get an * open file descriptor which we can use in the subsequent lookups. */ if ((S_ISREG(st->st_mode) || S_ISDIR(st->st_mode))) { if (fd < 0) fd = open(path, O_RDONLY | O_NONBLOCK); if (fd >= 0) { unsigned long stflags; - int r = ioctl(fd, EXT2_IOC_GETFLAGS, &stflags); + r = ioctl(fd, EXT2_IOC_GETFLAGS, &stflags); if (r == 0 && stflags != 0) archive_entry_set_fflags(entry, stflags, 0); } } #endif #if defined(HAVE_READLINK) || defined(HAVE_READLINKAT) if (S_ISLNK(st->st_mode)) { size_t linkbuffer_len = st->st_size + 1; char *linkbuffer; int lnklen; linkbuffer = malloc(linkbuffer_len); if (linkbuffer == NULL) { archive_set_error(&a->archive, ENOMEM, "Couldn't read link data"); return (ARCHIVE_FAILED); } #ifdef HAVE_READLINKAT if (a->entry_wd_fd >= 0) lnklen = readlinkat(a->entry_wd_fd, path, linkbuffer, linkbuffer_len); else #endif /* HAVE_READLINKAT */ lnklen = readlink(path, linkbuffer, linkbuffer_len); if (lnklen < 0) { archive_set_error(&a->archive, errno, "Couldn't read link data"); free(linkbuffer); return (ARCHIVE_FAILED); } linkbuffer[lnklen] = 0; archive_entry_set_symlink(entry, linkbuffer); free(linkbuffer); } #endif /* HAVE_READLINK || HAVE_READLINKAT */ r = setup_acls_posix1e(a, entry, fd); r1 = setup_xattrs(a, entry, fd); if (r1 < r) r = r1; r1 = setup_mac_metadata(a, entry, fd); if (r1 < r) r = r1; r1 = setup_sparse(a, entry, fd); if (r1 < r) r = r1; /* If we opened the file earlier in this function, close it. */ if (initial_fd != fd) close(fd); return (r); } #if defined(__APPLE__) && defined(HAVE_COPYFILE_H) /* * The Mac OS "copyfile()" API copies the extended metadata for a * file into a separate file in AppleDouble format (see RFC 1740). * * Mac OS tar and cpio implementations store this extended * metadata as a separate entry just before the regular entry * with a "._" prefix added to the filename. * * Note that this is currently done unconditionally; the tar program has * an option to discard this information before the archive is written. * * TODO: If there's a failure, report it and return ARCHIVE_WARN. */ static int setup_mac_metadata(struct archive_read_disk *a, struct archive_entry *entry, int fd) { int tempfd = -1; int copyfile_flags = COPYFILE_NOFOLLOW | COPYFILE_ACL | COPYFILE_XATTR; struct stat copyfile_stat; int ret = ARCHIVE_OK; void *buff; int have_attrs; const char *name, *tempdir, *tempfile = NULL; name = archive_entry_sourcepath(entry); if (name == NULL) name = archive_entry_pathname(entry); if (name == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't open file to read extended attributes: No name"); return (ARCHIVE_WARN); } /* Short-circuit if there's nothing to do. */ have_attrs = copyfile(name, NULL, 0, copyfile_flags | COPYFILE_CHECK); if (have_attrs == -1) { archive_set_error(&a->archive, errno, "Could not check extended attributes"); return (ARCHIVE_WARN); } if (have_attrs == 0) return (ARCHIVE_OK); tempdir = NULL; if (issetugid() == 0) tempdir = getenv("TMPDIR"); if (tempdir == NULL) tempdir = _PATH_TMP; tempfile = tempnam(tempdir, "tar.md."); /* XXX I wish copyfile() could pack directly to a memory * buffer; that would avoid the temp file here. For that * matter, it would be nice if fcopyfile() actually worked, * that would reduce the many open/close races here. */ if (copyfile(name, tempfile, 0, copyfile_flags | COPYFILE_PACK)) { archive_set_error(&a->archive, errno, "Could not pack extended attributes"); ret = ARCHIVE_WARN; goto cleanup; } tempfd = open(tempfile, O_RDONLY); if (tempfd < 0) { archive_set_error(&a->archive, errno, "Could not open extended attribute file"); ret = ARCHIVE_WARN; goto cleanup; } if (fstat(tempfd, ©file_stat)) { archive_set_error(&a->archive, errno, "Could not check size of extended attributes"); ret = ARCHIVE_WARN; goto cleanup; } buff = malloc(copyfile_stat.st_size); if (buff == NULL) { archive_set_error(&a->archive, errno, "Could not allocate memory for extended attributes"); ret = ARCHIVE_WARN; goto cleanup; } if (copyfile_stat.st_size != read(tempfd, buff, copyfile_stat.st_size)) { archive_set_error(&a->archive, errno, "Could not read extended attributes into memory"); ret = ARCHIVE_WARN; goto cleanup; } archive_entry_copy_mac_metadata(entry, buff, copyfile_stat.st_size); cleanup: if (tempfd >= 0) close(tempfd); if (tempfile != NULL) unlink(tempfile); return (ret); } #else /* * Stub implementation for non-Mac systems. */ static int setup_mac_metadata(struct archive_read_disk *a, struct archive_entry *entry, int fd) { (void)a; /* UNUSED */ (void)entry; /* UNUSED */ (void)fd; /* UNUSED */ return (ARCHIVE_OK); } #endif #ifdef HAVE_POSIX_ACL static void setup_acl_posix1e(struct archive_read_disk *a, struct archive_entry *entry, acl_t acl, int archive_entry_acl_type); static int setup_acls_posix1e(struct archive_read_disk *a, struct archive_entry *entry, int fd) { const char *accpath; acl_t acl; accpath = archive_entry_sourcepath(entry); if (accpath == NULL) accpath = archive_entry_pathname(entry); archive_entry_acl_clear(entry); /* Retrieve access ACL from file. */ if (fd >= 0) acl = acl_get_fd(fd); #if HAVE_ACL_GET_LINK_NP else if (!a->follow_symlinks) acl = acl_get_link_np(accpath, ACL_TYPE_ACCESS); #else else if ((!a->follow_symlinks) && (archive_entry_filetype(entry) == AE_IFLNK)) /* We can't get the ACL of a symlink, so we assume it can't have one. */ acl = NULL; #endif else acl = acl_get_file(accpath, ACL_TYPE_ACCESS); if (acl != NULL) { setup_acl_posix1e(a, entry, acl, ARCHIVE_ENTRY_ACL_TYPE_ACCESS); acl_free(acl); } /* Only directories can have default ACLs. */ if (S_ISDIR(archive_entry_mode(entry))) { acl = acl_get_file(accpath, ACL_TYPE_DEFAULT); if (acl != NULL) { setup_acl_posix1e(a, entry, acl, ARCHIVE_ENTRY_ACL_TYPE_DEFAULT); acl_free(acl); } } return (ARCHIVE_OK); } /* * Translate POSIX.1e ACL into libarchive internal structure. */ static void setup_acl_posix1e(struct archive_read_disk *a, struct archive_entry *entry, acl_t acl, int archive_entry_acl_type) { acl_tag_t acl_tag; acl_entry_t acl_entry; acl_permset_t acl_permset; int s, ae_id, ae_tag, ae_perm; const char *ae_name; s = acl_get_entry(acl, ACL_FIRST_ENTRY, &acl_entry); while (s == 1) { ae_id = -1; ae_name = NULL; acl_get_tag_type(acl_entry, &acl_tag); if (acl_tag == ACL_USER) { ae_id = (int)*(uid_t *)acl_get_qualifier(acl_entry); ae_name = archive_read_disk_uname(&a->archive, ae_id); ae_tag = ARCHIVE_ENTRY_ACL_USER; } else if (acl_tag == ACL_GROUP) { ae_id = (int)*(gid_t *)acl_get_qualifier(acl_entry); ae_name = archive_read_disk_gname(&a->archive, ae_id); ae_tag = ARCHIVE_ENTRY_ACL_GROUP; } else if (acl_tag == ACL_MASK) { ae_tag = ARCHIVE_ENTRY_ACL_MASK; } else if (acl_tag == ACL_USER_OBJ) { ae_tag = ARCHIVE_ENTRY_ACL_USER_OBJ; } else if (acl_tag == ACL_GROUP_OBJ) { ae_tag = ARCHIVE_ENTRY_ACL_GROUP_OBJ; } else if (acl_tag == ACL_OTHER) { ae_tag = ARCHIVE_ENTRY_ACL_OTHER; } else { /* Skip types that libarchive can't support. */ continue; } acl_get_permset(acl_entry, &acl_permset); ae_perm = 0; /* * acl_get_perm() is spelled differently on different * platforms; see above. */ if (ACL_GET_PERM(acl_permset, ACL_EXECUTE)) ae_perm |= ARCHIVE_ENTRY_ACL_EXECUTE; if (ACL_GET_PERM(acl_permset, ACL_READ)) ae_perm |= ARCHIVE_ENTRY_ACL_READ; if (ACL_GET_PERM(acl_permset, ACL_WRITE)) ae_perm |= ARCHIVE_ENTRY_ACL_WRITE; archive_entry_acl_add_entry(entry, archive_entry_acl_type, ae_perm, ae_tag, ae_id, ae_name); s = acl_get_entry(acl, ACL_NEXT_ENTRY, &acl_entry); } } #else static int setup_acls_posix1e(struct archive_read_disk *a, struct archive_entry *entry, int fd) { (void)a; /* UNUSED */ (void)entry; /* UNUSED */ (void)fd; /* UNUSED */ return (ARCHIVE_OK); } #endif #if (HAVE_FGETXATTR && HAVE_FLISTXATTR && HAVE_LISTXATTR && \ HAVE_LLISTXATTR && HAVE_GETXATTR && HAVE_LGETXATTR) || \ (HAVE_FGETEA && HAVE_FLISTEA && HAVE_LISTEA) /* * Linux and AIX extended attribute support. * * TODO: By using a stack-allocated buffer for the first * call to getxattr(), we might be able to avoid the second * call entirely. We only need the second call if the * stack-allocated buffer is too small. But a modest buffer * of 1024 bytes or so will often be big enough. Same applies * to listxattr(). */ static int setup_xattr(struct archive_read_disk *a, struct archive_entry *entry, const char *name, int fd) { ssize_t size; void *value = NULL; const char *accpath; accpath = archive_entry_sourcepath(entry); if (accpath == NULL) accpath = archive_entry_pathname(entry); #if HAVE_FGETXATTR if (fd >= 0) size = fgetxattr(fd, name, NULL, 0); else if (!a->follow_symlinks) size = lgetxattr(accpath, name, NULL, 0); else size = getxattr(accpath, name, NULL, 0); #elif HAVE_FGETEA if (fd >= 0) size = fgetea(fd, name, NULL, 0); else if (!a->follow_symlinks) size = lgetea(accpath, name, NULL, 0); else size = getea(accpath, name, NULL, 0); #endif if (size == -1) { archive_set_error(&a->archive, errno, "Couldn't query extended attribute"); return (ARCHIVE_WARN); } if (size > 0 && (value = malloc(size)) == NULL) { archive_set_error(&a->archive, errno, "Out of memory"); return (ARCHIVE_FATAL); } #if HAVE_FGETXATTR if (fd >= 0) size = fgetxattr(fd, name, value, size); else if (!a->follow_symlinks) size = lgetxattr(accpath, name, value, size); else size = getxattr(accpath, name, value, size); #elif HAVE_FGETEA if (fd >= 0) size = fgetea(fd, name, value, size); else if (!a->follow_symlinks) size = lgetea(accpath, name, value, size); else size = getea(accpath, name, value, size); #endif if (size == -1) { archive_set_error(&a->archive, errno, "Couldn't read extended attribute"); return (ARCHIVE_WARN); } archive_entry_xattr_add_entry(entry, name, value, size); free(value); return (ARCHIVE_OK); } static int setup_xattrs(struct archive_read_disk *a, struct archive_entry *entry, int fd) { char *list, *p; const char *path; ssize_t list_size; path = archive_entry_sourcepath(entry); if (path == NULL) path = archive_entry_pathname(entry); #if HAVE_FLISTXATTR if (fd >= 0) list_size = flistxattr(fd, NULL, 0); else if (!a->follow_symlinks) list_size = llistxattr(path, NULL, 0); else list_size = listxattr(path, NULL, 0); #elif HAVE_FLISTEA if (fd >= 0) list_size = flistea(fd, NULL, 0); else if (!a->follow_symlinks) list_size = llistea(path, NULL, 0); else list_size = listea(path, NULL, 0); #endif if (list_size == -1) { if (errno == ENOTSUP || errno == ENOSYS) return (ARCHIVE_OK); archive_set_error(&a->archive, errno, "Couldn't list extended attributes"); return (ARCHIVE_WARN); } if (list_size == 0) return (ARCHIVE_OK); if ((list = malloc(list_size)) == NULL) { archive_set_error(&a->archive, errno, "Out of memory"); return (ARCHIVE_FATAL); } #if HAVE_FLISTXATTR if (fd >= 0) list_size = flistxattr(fd, list, list_size); else if (!a->follow_symlinks) list_size = llistxattr(path, list, list_size); else list_size = listxattr(path, list, list_size); #elif HAVE_FLISTEA if (fd >= 0) list_size = flistea(fd, list, list_size); else if (!a->follow_symlinks) list_size = llistea(path, list, list_size); else list_size = listea(path, list, list_size); #endif if (list_size == -1) { archive_set_error(&a->archive, errno, "Couldn't retrieve extended attributes"); free(list); return (ARCHIVE_WARN); } for (p = list; (p - list) < list_size; p += strlen(p) + 1) { if (strncmp(p, "system.", 7) == 0 || strncmp(p, "xfsroot.", 8) == 0) continue; setup_xattr(a, entry, p, fd); } free(list); return (ARCHIVE_OK); } #elif HAVE_EXTATTR_GET_FILE && HAVE_EXTATTR_LIST_FILE && \ HAVE_DECL_EXTATTR_NAMESPACE_USER /* * FreeBSD extattr interface. */ /* TODO: Implement this. Follow the Linux model above, but * with FreeBSD-specific system calls, of course. Be careful * to not include the system extattrs that hold ACLs; we handle * those separately. */ static int setup_xattr(struct archive_read_disk *a, struct archive_entry *entry, int namespace, const char *name, const char *fullname, int fd); static int setup_xattr(struct archive_read_disk *a, struct archive_entry *entry, int namespace, const char *name, const char *fullname, int fd) { ssize_t size; void *value = NULL; const char *accpath; accpath = archive_entry_sourcepath(entry); if (accpath == NULL) accpath = archive_entry_pathname(entry); if (fd >= 0) size = extattr_get_fd(fd, namespace, name, NULL, 0); else if (!a->follow_symlinks) size = extattr_get_link(accpath, namespace, name, NULL, 0); else size = extattr_get_file(accpath, namespace, name, NULL, 0); if (size == -1) { archive_set_error(&a->archive, errno, "Couldn't query extended attribute"); return (ARCHIVE_WARN); } if (size > 0 && (value = malloc(size)) == NULL) { archive_set_error(&a->archive, errno, "Out of memory"); return (ARCHIVE_FATAL); } if (fd >= 0) size = extattr_get_fd(fd, namespace, name, value, size); else if (!a->follow_symlinks) size = extattr_get_link(accpath, namespace, name, value, size); else size = extattr_get_file(accpath, namespace, name, value, size); if (size == -1) { archive_set_error(&a->archive, errno, "Couldn't read extended attribute"); return (ARCHIVE_WARN); } archive_entry_xattr_add_entry(entry, fullname, value, size); free(value); return (ARCHIVE_OK); } static int setup_xattrs(struct archive_read_disk *a, struct archive_entry *entry, int fd) { char buff[512]; char *list, *p; ssize_t list_size; const char *path; int namespace = EXTATTR_NAMESPACE_USER; path = archive_entry_sourcepath(entry); if (path == NULL) path = archive_entry_pathname(entry); if (fd >= 0) list_size = extattr_list_fd(fd, namespace, NULL, 0); else if (!a->follow_symlinks) list_size = extattr_list_link(path, namespace, NULL, 0); else list_size = extattr_list_file(path, namespace, NULL, 0); if (list_size == -1 && errno == EOPNOTSUPP) return (ARCHIVE_OK); if (list_size == -1) { archive_set_error(&a->archive, errno, "Couldn't list extended attributes"); return (ARCHIVE_WARN); } if (list_size == 0) return (ARCHIVE_OK); if ((list = malloc(list_size)) == NULL) { archive_set_error(&a->archive, errno, "Out of memory"); return (ARCHIVE_FATAL); } if (fd >= 0) list_size = extattr_list_fd(fd, namespace, list, list_size); else if (!a->follow_symlinks) list_size = extattr_list_link(path, namespace, list, list_size); else list_size = extattr_list_file(path, namespace, list, list_size); if (list_size == -1) { archive_set_error(&a->archive, errno, "Couldn't retrieve extended attributes"); free(list); return (ARCHIVE_WARN); } p = list; while ((p - list) < list_size) { size_t len = 255 & (int)*p; char *name; strcpy(buff, "user."); name = buff + strlen(buff); memcpy(name, p + 1, len); name[len] = '\0'; setup_xattr(a, entry, namespace, name, buff, fd); p += 1 + len; } free(list); return (ARCHIVE_OK); } #else /* * Generic (stub) extended attribute support. */ static int setup_xattrs(struct archive_read_disk *a, struct archive_entry *entry, int fd) { (void)a; /* UNUSED */ (void)entry; /* UNUSED */ (void)fd; /* UNUSED */ return (ARCHIVE_OK); } #endif #if defined(HAVE_LINUX_FIEMAP_H) /* * Linux sparse interface. * * The FIEMAP ioctl returns an "extent" for each physical allocation * on disk. We need to process those to generate a more compact list * of logical file blocks. We also need to be very careful to use * FIEMAP_FLAG_SYNC here, since there are reports that Linux sometimes * does not report allocations for newly-written data that hasn't * been synced to disk. * * It's important to return a minimal sparse file list because we want * to not trigger sparse file extensions if we don't have to, since * not all readers support them. */ static int setup_sparse(struct archive_read_disk *a, struct archive_entry *entry, int fd) { char buff[4096]; struct fiemap *fm; struct fiemap_extent *fe; int64_t size; int count, do_fiemap; int initial_fd = fd; int exit_sts = ARCHIVE_OK; if (archive_entry_filetype(entry) != AE_IFREG || archive_entry_size(entry) <= 0 || archive_entry_hardlink(entry) != NULL) return (ARCHIVE_OK); if (fd < 0) { const char *path; path = archive_entry_sourcepath(entry); if (path == NULL) path = archive_entry_pathname(entry); fd = open(path, O_RDONLY | O_NONBLOCK); if (fd < 0) { archive_set_error(&a->archive, errno, "Can't open `%s'", path); return (ARCHIVE_FAILED); } } count = (sizeof(buff) - sizeof(*fm))/sizeof(*fe); fm = (struct fiemap *)buff; fm->fm_start = 0; fm->fm_length = ~0ULL;; fm->fm_flags = FIEMAP_FLAG_SYNC; fm->fm_extent_count = count; do_fiemap = 1; size = archive_entry_size(entry); for (;;) { int i, r; r = ioctl(fd, FS_IOC_FIEMAP, fm); if (r < 0) { /* When errno is ENOTTY, it is better we should * return ARCHIVE_OK because an earlier version *(<2.6.28) cannot perfom FS_IOC_FIEMAP. * We should also check if errno is EOPNOTSUPP, * it means "Operation not supported". */ if (errno != ENOTTY && errno != EOPNOTSUPP) { archive_set_error(&a->archive, errno, "FIEMAP failed"); exit_sts = ARCHIVE_FAILED; } goto exit_setup_sparse; } if (fm->fm_mapped_extents == 0) break; fe = fm->fm_extents; - for (i = 0; i < fm->fm_mapped_extents; i++, fe++) { + for (i = 0; i < (int)fm->fm_mapped_extents; i++, fe++) { if (!(fe->fe_flags & FIEMAP_EXTENT_UNWRITTEN)) { /* The fe_length of the last block does not * adjust itself to its size files. */ int64_t length = fe->fe_length; - if (fe->fe_logical + length > size) + if (fe->fe_logical + length > (uint64_t)size) length -= fe->fe_logical + length - size; if (fe->fe_logical == 0 && length == size) { /* This is not sparse. */ do_fiemap = 0; break; } if (length > 0) archive_entry_sparse_add_entry(entry, fe->fe_logical, length); } if (fe->fe_flags & FIEMAP_EXTENT_LAST) do_fiemap = 0; } if (do_fiemap) { fe = fm->fm_extents + fm->fm_mapped_extents -1; fm->fm_start = fe->fe_logical + fe->fe_length; } else break; } exit_setup_sparse: if (initial_fd != fd) close(fd); return (exit_sts); } #elif defined(SEEK_HOLE) && defined(SEEK_DATA) && defined(_PC_MIN_HOLE_SIZE) /* * FreeBSD and Solaris sparse interface. */ static int setup_sparse(struct archive_read_disk *a, struct archive_entry *entry, int fd) { int64_t size; int initial_fd = fd; off_t initial_off; /* FreeBSD/Solaris only, so off_t okay here */ off_t off_s, off_e; /* FreeBSD/Solaris only, so off_t okay here */ int exit_sts = ARCHIVE_OK; if (archive_entry_filetype(entry) != AE_IFREG || archive_entry_size(entry) <= 0 || archive_entry_hardlink(entry) != NULL) return (ARCHIVE_OK); /* Does filesystem support the reporting of hole ? */ if (fd >= 0) { if (fpathconf(fd, _PC_MIN_HOLE_SIZE) <= 0) return (ARCHIVE_OK); initial_off = lseek(fd, 0, SEEK_CUR); if (initial_off != 0) lseek(fd, 0, SEEK_SET); } else { const char *path; path = archive_entry_sourcepath(entry); if (path == NULL) path = archive_entry_pathname(entry); if (pathconf(path, _PC_MIN_HOLE_SIZE) <= 0) return (ARCHIVE_OK); fd = open(path, O_RDONLY | O_NONBLOCK); if (fd < 0) { archive_set_error(&a->archive, errno, "Can't open `%s'", path); return (ARCHIVE_FAILED); } initial_off = 0; } off_s = 0; size = archive_entry_size(entry); while (off_s < size) { off_s = lseek(fd, off_s, SEEK_DATA); if (off_s == (off_t)-1) { if (errno == ENXIO) break;/* no more hole */ archive_set_error(&a->archive, errno, "lseek(SEEK_HOLE) failed"); exit_sts = ARCHIVE_FAILED; goto exit_setup_sparse; } off_e = lseek(fd, off_s, SEEK_HOLE); if (off_s == (off_t)-1) { if (errno == ENXIO) { off_e = lseek(fd, 0, SEEK_END); if (off_e != (off_t)-1) break;/* no more data */ } archive_set_error(&a->archive, errno, "lseek(SEEK_DATA) failed"); exit_sts = ARCHIVE_FAILED; goto exit_setup_sparse; } if (off_s == 0 && off_e == size) break;/* This is not spase. */ archive_entry_sparse_add_entry(entry, off_s, off_e - off_s); off_s = off_e; } exit_setup_sparse: if (initial_fd != fd) close(fd); else lseek(fd, initial_off, SEEK_SET); return (exit_sts); } #else /* * Generic (stub) sparse support. */ static int setup_sparse(struct archive_read_disk *a, struct archive_entry *entry, int fd) { (void)a; /* UNUSED */ (void)entry; /* UNUSED */ (void)fd; /* UNUSED */ return (ARCHIVE_OK); } #endif #endif /* !defined(_WIN32) || defined(__CYGWIN__) */ Index: vendor/libarchive/dist/libarchive/archive_read_disk_posix.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_read_disk_posix.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_read_disk_posix.c (revision 231297) @@ -1,2311 +1,2313 @@ /*- * Copyright (c) 2003-2009 Tim Kientzle * Copyright (c) 2010,2011 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ /* This is the tree-walking code for POSIX systems. */ #if !defined(_WIN32) || defined(__CYGWIN__) #include "archive_platform.h" __FBSDID("$FreeBSD$"); #ifdef HAVE_SYS_PARAM_H #include #endif #ifdef HAVE_SYS_MOUNT_H #include #endif #ifdef HAVE_SYS_STAT_H #include #endif #ifdef HAVE_SYS_STATFS_H #include #endif #ifdef HAVE_SYS_STATVFS_H #include #endif #ifdef HAVE_SYS_TIME_H #include #endif #ifdef HAVE_LINUX_MAGIC_H #include #endif #ifdef HAVE_DIRECT_H #include #endif #ifdef HAVE_DIRENT_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #include "archive.h" #include "archive_string.h" #include "archive_entry.h" #include "archive_private.h" #include "archive_read_disk_private.h" #ifndef HAVE_FCHDIR #error fchdir function required. #endif #ifndef O_BINARY #define O_BINARY 0 #endif /*- * This is a new directory-walking system that addresses a number * of problems I've had with fts(3). In particular, it has no * pathname-length limits (other than the size of 'int'), handles * deep logical traversals, uses considerably less memory, and has * an opaque interface (easier to modify in the future). * * Internally, it keeps a single list of "tree_entry" items that * represent filesystem objects that require further attention. * Non-directories are not kept in memory: they are pulled from * readdir(), returned to the client, then freed as soon as possible. * Any directory entry to be traversed gets pushed onto the stack. * * There is surprisingly little information that needs to be kept for * each item on the stack. Just the name, depth (represented here as the * string length of the parent directory's pathname), and some markers * indicating how to get back to the parent (via chdir("..") for a * regular dir or via fchdir(2) for a symlink). */ /* * TODO: * 1) Loop checking. * 3) Arbitrary logical traversals by closing/reopening intermediate fds. */ struct restore_time { const char *name; time_t mtime; long mtime_nsec; time_t atime; long atime_nsec; mode_t filetype; int noatime; }; struct tree_entry { int depth; struct tree_entry *next; struct tree_entry *parent; struct archive_string name; size_t dirname_length; int64_t dev; int64_t ino; int flags; int filesystem_id; /* How to return back to the parent of a symlink. */ int symlink_parent_fd; /* How to restore time of a directory. */ struct restore_time restore_time; }; struct filesystem { int64_t dev; int synthetic; int remote; int noatime; #if defined(HAVE_READDIR_R) size_t name_max; #endif long incr_xfer_size; long max_xfer_size; long min_xfer_size; long xfer_align; /* * Buffer used for reading file contents. */ /* Exactly allocated memory pointer. */ unsigned char *allocation_ptr; /* Pointer adjusted to the filesystem alignment . */ unsigned char *buff; size_t buff_size; }; /* Definitions for tree_entry.flags bitmap. */ #define isDir 1 /* This entry is a regular directory. */ #define isDirLink 2 /* This entry is a symbolic link to a directory. */ #define needsFirstVisit 4 /* This is an initial entry. */ #define needsDescent 8 /* This entry needs to be previsited. */ #define needsOpen 16 /* This is a directory that needs to be opened. */ #define needsAscent 32 /* This entry needs to be postvisited. */ /* * Local data for this package. */ struct tree { struct tree_entry *stack; struct tree_entry *current; DIR *d; #define INVALID_DIR_HANDLE NULL struct dirent *de; #if defined(HAVE_READDIR_R) struct dirent *dirent; size_t dirent_allocated; #endif int flags; int visit_type; /* Error code from last failed operation. */ int tree_errno; /* Dynamically-sized buffer for holding path */ struct archive_string path; /* Last path element */ const char *basename; /* Leading dir length */ size_t dirname_length; int depth; int openCount; int maxOpenCount; int initial_dir_fd; int working_dir_fd; struct stat lst; struct stat st; int descend; int nlink; /* How to restore time of a file. */ struct restore_time restore_time; struct entry_sparse { int64_t length; int64_t offset; } *sparse_list, *current_sparse; int sparse_count; int sparse_list_size; char initial_symlink_mode; char symlink_mode; struct filesystem *current_filesystem; struct filesystem *filesystem_table; int current_filesystem_id; int max_filesystem_id; int allocated_filesytem; int entry_fd; int entry_eof; int64_t entry_remaining_bytes; int64_t entry_total; unsigned char *entry_buff; size_t entry_buff_size; }; /* Definitions for tree.flags bitmap. */ #define hasStat 16 /* The st entry is valid. */ #define hasLstat 32 /* The lst entry is valid. */ #define onWorkingDir 64 /* We are on the working dir where we are * reading directory entry at this time. */ #define needsRestoreTimes 128 static int tree_dir_next_posix(struct tree *t); #ifdef HAVE_DIRENT_D_NAMLEN /* BSD extension; avoids need for a strlen() call. */ #define D_NAMELEN(dp) (dp)->d_namlen #else #define D_NAMELEN(dp) (strlen((dp)->d_name)) #endif /* Initiate/terminate a tree traversal. */ static struct tree *tree_open(const char *, int, int); static struct tree *tree_reopen(struct tree *, const char *, int); static void tree_close(struct tree *); static void tree_free(struct tree *); static void tree_push(struct tree *, const char *, int, int64_t, int64_t, struct restore_time *); static int tree_enter_initial_dir(struct tree *); static int tree_enter_working_dir(struct tree *); static int tree_current_dir_fd(struct tree *); /* * tree_next() returns Zero if there is no next entry, non-zero if * there is. Note that directories are visited three times. * Directories are always visited first as part of enumerating their * parent; that is a "regular" visit. If tree_descend() is invoked at * that time, the directory is added to a work list and will * subsequently be visited two more times: once just after descending * into the directory ("postdescent") and again just after ascending * back to the parent ("postascent"). * * TREE_ERROR_DIR is returned if the descent failed (because the * directory couldn't be opened, for instance). This is returned * instead of TREE_POSTDESCENT/TREE_POSTASCENT. TREE_ERROR_DIR is not a * fatal error, but it does imply that the relevant subtree won't be * visited. TREE_ERROR_FATAL is returned for an error that left the * traversal completely hosed. Right now, this is only returned for * chdir() failures during ascent. */ #define TREE_REGULAR 1 #define TREE_POSTDESCENT 2 #define TREE_POSTASCENT 3 #define TREE_ERROR_DIR -1 #define TREE_ERROR_FATAL -2 static int tree_next(struct tree *); /* * Return information about the current entry. */ /* * The current full pathname, length of the full pathname, and a name * that can be used to access the file. Because tree does use chdir * extensively, the access path is almost never the same as the full * current path. * * TODO: On platforms that support it, use openat()-style operations * to eliminate the chdir() operations entirely while still supporting * arbitrarily deep traversals. This makes access_path troublesome to * support, of course, which means we'll need a rich enough interface * that clients can function without it. (In particular, we'll need * tree_current_open() that returns an open file descriptor.) * */ static const char *tree_current_path(struct tree *); static const char *tree_current_access_path(struct tree *); /* * Request the lstat() or stat() data for the current path. Since the * tree package needs to do some of this anyway, and caches the * results, you should take advantage of it here if you need it rather * than make a redundant stat() or lstat() call of your own. */ static const struct stat *tree_current_stat(struct tree *); static const struct stat *tree_current_lstat(struct tree *); static int tree_current_is_symblic_link_target(struct tree *); /* The following functions use tricks to avoid a certain number of * stat()/lstat() calls. */ /* "is_physical_dir" is equivalent to S_ISDIR(tree_current_lstat()->st_mode) */ static int tree_current_is_physical_dir(struct tree *); /* "is_dir" is equivalent to S_ISDIR(tree_current_stat()->st_mode) */ static int tree_current_is_dir(struct tree *); static int update_current_filesystem(struct archive_read_disk *a, int64_t dev); static int setup_current_filesystem(struct archive_read_disk *); static int tree_target_is_same_as_parent(struct tree *, const struct stat *); static int _archive_read_disk_open(struct archive *, const char *); static int _archive_read_free(struct archive *); static int _archive_read_close(struct archive *); static int _archive_read_data_block(struct archive *, const void **, size_t *, int64_t *); static int _archive_read_next_header2(struct archive *, struct archive_entry *); static const char *trivial_lookup_gname(void *, int64_t gid); static const char *trivial_lookup_uname(void *, int64_t uid); static int setup_sparse(struct archive_read_disk *, struct archive_entry *); static int close_and_restore_time(int fd, struct tree *, struct restore_time *); static struct archive_vtable * archive_read_disk_vtable(void) { static struct archive_vtable av; static int inited = 0; if (!inited) { av.archive_free = _archive_read_free; av.archive_close = _archive_read_close; av.archive_read_data_block = _archive_read_data_block; av.archive_read_next_header2 = _archive_read_next_header2; inited = 1; } return (&av); } const char * archive_read_disk_gname(struct archive *_a, int64_t gid) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_gname")) return (NULL); if (a->lookup_gname == NULL) return (NULL); return ((*a->lookup_gname)(a->lookup_gname_data, gid)); } const char * archive_read_disk_uname(struct archive *_a, int64_t uid) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_uname")) return (NULL); if (a->lookup_uname == NULL) return (NULL); return ((*a->lookup_uname)(a->lookup_uname_data, uid)); } int archive_read_disk_set_gname_lookup(struct archive *_a, void *private_data, const char * (*lookup_gname)(void *private, int64_t gid), void (*cleanup_gname)(void *private)) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_gname_lookup"); if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) (a->cleanup_gname)(a->lookup_gname_data); a->lookup_gname = lookup_gname; a->cleanup_gname = cleanup_gname; a->lookup_gname_data = private_data; return (ARCHIVE_OK); } int archive_read_disk_set_uname_lookup(struct archive *_a, void *private_data, const char * (*lookup_uname)(void *private, int64_t uid), void (*cleanup_uname)(void *private)) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_uname_lookup"); if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) (a->cleanup_uname)(a->lookup_uname_data); a->lookup_uname = lookup_uname; a->cleanup_uname = cleanup_uname; a->lookup_uname_data = private_data; return (ARCHIVE_OK); } /* * Create a new archive_read_disk object and initialize it with global state. */ struct archive * archive_read_disk_new(void) { struct archive_read_disk *a; a = (struct archive_read_disk *)malloc(sizeof(*a)); if (a == NULL) return (NULL); memset(a, 0, sizeof(*a)); a->archive.magic = ARCHIVE_READ_DISK_MAGIC; a->archive.state = ARCHIVE_STATE_NEW; a->archive.vtable = archive_read_disk_vtable(); a->lookup_uname = trivial_lookup_uname; a->lookup_gname = trivial_lookup_gname; a->entry_wd_fd = -1; return (&a->archive); } static int _archive_read_free(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; int r; if (_a == NULL) return (ARCHIVE_OK); archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free"); if (a->archive.state != ARCHIVE_STATE_CLOSED) r = _archive_read_close(&a->archive); else r = ARCHIVE_OK; tree_free(a->tree); if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) (a->cleanup_gname)(a->lookup_gname_data); if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) (a->cleanup_uname)(a->lookup_uname_data); archive_string_free(&a->archive.error_string); a->archive.magic = 0; __archive_clean(&a->archive); free(a); return (r); } static int _archive_read_close(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close"); if (a->archive.state != ARCHIVE_STATE_FATAL) a->archive.state = ARCHIVE_STATE_CLOSED; tree_close(a->tree); return (ARCHIVE_OK); } static void setup_symlink_mode(struct archive_read_disk *a, char symlink_mode, int follow_symlinks) { a->symlink_mode = symlink_mode; a->follow_symlinks = follow_symlinks; if (a->tree != NULL) { a->tree->initial_symlink_mode = a->symlink_mode; a->tree->symlink_mode = a->symlink_mode; } } int archive_read_disk_set_symlink_logical(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical"); setup_symlink_mode(a, 'L', 1); return (ARCHIVE_OK); } int archive_read_disk_set_symlink_physical(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical"); setup_symlink_mode(a, 'P', 0); return (ARCHIVE_OK); } int archive_read_disk_set_symlink_hybrid(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid"); setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */ return (ARCHIVE_OK); } int archive_read_disk_set_atime_restored(struct archive *_a) { #ifndef HAVE_UTIMES static int warning_done = 0; #endif struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime"); #ifdef HAVE_UTIMES a->restore_time = 1; if (a->tree != NULL) a->tree->flags |= needsRestoreTimes; return (ARCHIVE_OK); #else if (warning_done) /* Warning was already emitted; suppress further warnings. */ return (ARCHIVE_OK); archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Cannot restore access time on this system"); warning_done = 1; return (ARCHIVE_WARN); #endif } /* * Trivial implementations of gname/uname lookup functions. * These are normally overridden by the client, but these stub * versions ensure that we always have something that works. */ static const char * trivial_lookup_gname(void *private_data, int64_t gid) { (void)private_data; /* UNUSED */ (void)gid; /* UNUSED */ return (NULL); } static const char * trivial_lookup_uname(void *private_data, int64_t uid) { (void)private_data; /* UNUSED */ (void)uid; /* UNUSED */ return (NULL); } /* * Allocate memory for the reading buffer adjusted to the filesystem * alignment. */ static int setup_suitable_read_buffer(struct archive_read_disk *a) { struct tree *t = a->tree; struct filesystem *cf = t->current_filesystem; size_t asize; size_t s; if (cf->allocation_ptr == NULL) { /* If we couldn't get a filesystem alignment, * we use 4096 as default value but we won't use * O_DIRECT to open() and openat() operations. */ long xfer_align = (cf->xfer_align == -1)?4096:cf->xfer_align; if (cf->max_xfer_size != -1) asize = cf->max_xfer_size + xfer_align; else { long incr = cf->incr_xfer_size; /* Some platform does not set a proper value to * incr_xfer_size.*/ if (incr < 0) incr = cf->min_xfer_size; if (cf->min_xfer_size < 0) { incr = xfer_align; asize = xfer_align; } else asize = cf->min_xfer_size; /* Increase a buffer size up to 64K bytes in * a proper incremant size. */ while (asize < 1024*64) asize += incr; /* Take a margin to adjust to the filesystem * alignment. */ asize += xfer_align; } cf->allocation_ptr = malloc(asize); if (cf->allocation_ptr == NULL) { archive_set_error(&a->archive, ENOMEM, "Couldn't allocate memory"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } /* * Calculate proper address for the filesystem. */ s = (uintptr_t)cf->allocation_ptr; s %= xfer_align; if (s > 0) s = xfer_align - s; /* * Set a read buffer pointer in the proper alignment of * the current filesystem. */ cf->buff = cf->allocation_ptr + s; cf->buff_size = asize - xfer_align; } return (ARCHIVE_OK); } static int _archive_read_data_block(struct archive *_a, const void **buff, size_t *size, int64_t *offset) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t = a->tree; int r; ssize_t bytes; size_t buffbytes; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_data_block"); if (t->entry_eof || t->entry_remaining_bytes <= 0) { r = ARCHIVE_EOF; goto abort_read_data; } /* * Open the current file. */ if (t->entry_fd < 0) { int flags = O_RDONLY | O_BINARY; /* * Eliminate or reduce cache effects if we can. * * Carefully consider this to be enabled. */ #if defined(O_DIRECT) && 0/* Disabled for now */ if (t->current_filesystem->xfer_align != -1 && t->nlink == 1) flags |= O_DIRECT; #endif #if defined(O_NOATIME) /* * Linux has O_NOATIME flag; use it if we need. */ if ((t->flags & needsRestoreTimes) != 0 && t->restore_time.noatime == 0) flags |= O_NOATIME; do { #endif #ifdef HAVE_OPENAT t->entry_fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), flags); #else tree_enter_working_dir(t); t->entry_fd = open(tree_current_access_path(t), flags); #endif #if defined(O_NOATIME) /* * When we did open the file with O_NOATIME flag, * if successful, set 1 to t->restore_time.noatime * not to restore an atime of the file later. * if failed by EPERM, retry it without O_NOATIME flag. */ if (flags & O_NOATIME) { if (t->entry_fd >= 0) t->restore_time.noatime = 1; else if (errno == EPERM) { flags &= ~O_NOATIME; continue; } } } while (0); #endif if (t->entry_fd < 0) { archive_set_error(&a->archive, errno, "Couldn't open %s", tree_current_path(t)); r = ARCHIVE_FAILED; tree_enter_initial_dir(t); goto abort_read_data; } tree_enter_initial_dir(t); } /* * Allocate read buffer if not allocated. */ if (t->current_filesystem->allocation_ptr == NULL) { r = setup_suitable_read_buffer(a); if (r != ARCHIVE_OK) { a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } } t->entry_buff = t->current_filesystem->buff; t->entry_buff_size = t->current_filesystem->buff_size; buffbytes = t->entry_buff_size; if (buffbytes > t->current_sparse->length) buffbytes = t->current_sparse->length; /* * Skip hole. * TODO: Should we consider t->current_filesystem->xfer_align? */ if (t->current_sparse->offset > t->entry_total) { if (lseek(t->entry_fd, (off_t)t->current_sparse->offset, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek error"); r = ARCHIVE_FATAL; a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } bytes = t->current_sparse->offset - t->entry_total; t->entry_remaining_bytes -= bytes; t->entry_total += bytes; } /* * Read file contents. */ if (buffbytes > 0) { bytes = read(t->entry_fd, t->entry_buff, buffbytes); if (bytes < 0) { archive_set_error(&a->archive, errno, "Read error"); r = ARCHIVE_FATAL; a->archive.state = ARCHIVE_STATE_FATAL; goto abort_read_data; } } else bytes = 0; if (bytes == 0) { /* Get EOF */ t->entry_eof = 1; r = ARCHIVE_EOF; goto abort_read_data; } *buff = t->entry_buff; *size = bytes; *offset = t->entry_total; t->entry_total += bytes; t->entry_remaining_bytes -= bytes; if (t->entry_remaining_bytes == 0) { /* Close the current file descriptor */ close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; t->entry_eof = 1; } t->current_sparse->offset += bytes; t->current_sparse->length -= bytes; if (t->current_sparse->length == 0 && !t->entry_eof) t->current_sparse++; return (ARCHIVE_OK); abort_read_data: *buff = NULL; *size = 0; *offset = t->entry_total; if (t->entry_fd >= 0) { /* Close the current file descriptor */ close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } return (r); } static int _archive_read_next_header2(struct archive *_a, struct archive_entry *entry) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t; const struct stat *st; /* info to use for this entry */ const struct stat *lst;/* lstat() information */ int descend, fd = -1, r; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_read_next_header2"); t = a->tree; if (t->entry_fd >= 0) { close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } #if !(defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR)) /* Restore working directory. */ tree_enter_working_dir(t); #endif st = NULL; lst = NULL; do { switch (tree_next(t)) { case TREE_ERROR_FATAL: archive_set_error(&a->archive, t->tree_errno, "%s: Unable to continue traversing directory tree", tree_current_path(t)); a->archive.state = ARCHIVE_STATE_FATAL; tree_enter_initial_dir(t); return (ARCHIVE_FATAL); case TREE_ERROR_DIR: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "%s: Couldn't visit directory", tree_current_path(t)); tree_enter_initial_dir(t); return (ARCHIVE_FAILED); case 0: tree_enter_initial_dir(t); return (ARCHIVE_EOF); case TREE_POSTDESCENT: case TREE_POSTASCENT: break; case TREE_REGULAR: lst = tree_current_lstat(t); if (lst == NULL) { archive_set_error(&a->archive, errno, "%s: Cannot stat", tree_current_path(t)); tree_enter_initial_dir(t); return (ARCHIVE_FAILED); } break; } } while (lst == NULL); /* * Distinguish 'L'/'P'/'H' symlink following. */ switch(t->symlink_mode) { case 'H': /* 'H': After the first item, rest like 'P'. */ t->symlink_mode = 'P'; /* 'H': First item (from command line) like 'L'. */ /* FALLTHROUGH */ case 'L': /* 'L': Do descend through a symlink to dir. */ descend = tree_current_is_dir(t); /* 'L': Follow symlinks to files. */ a->symlink_mode = 'L'; a->follow_symlinks = 1; /* 'L': Archive symlinks as targets, if we can. */ st = tree_current_stat(t); if (st != NULL && !tree_target_is_same_as_parent(t, st)) break; /* If stat fails, we have a broken symlink; * in that case, don't follow the link. */ /* FALLTHROUGH */ default: /* 'P': Don't descend through a symlink to dir. */ descend = tree_current_is_physical_dir(t); /* 'P': Don't follow symlinks to files. */ a->symlink_mode = 'P'; a->follow_symlinks = 0; /* 'P': Archive symlinks as symlinks. */ st = lst; break; } if (update_current_filesystem(a, st->st_dev) != ARCHIVE_OK) { a->archive.state = ARCHIVE_STATE_FATAL; tree_enter_initial_dir(t); return (ARCHIVE_FATAL); } t->descend = descend; archive_entry_set_pathname(entry, tree_current_path(t)); archive_entry_copy_sourcepath(entry, tree_current_access_path(t)); archive_entry_copy_stat(entry, st); /* Save the times to be restored. */ t->restore_time.mtime = archive_entry_mtime(entry); t->restore_time.mtime_nsec = archive_entry_mtime_nsec(entry); t->restore_time.atime = archive_entry_atime(entry); t->restore_time.atime_nsec = archive_entry_atime_nsec(entry); t->restore_time.filetype = archive_entry_filetype(entry); t->restore_time.noatime = t->current_filesystem->noatime; #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) /* * Open the current file to freely gather its metadata anywhere in * working directory. * Note: A symbolic link file cannot be opened with O_NOFOLLOW. */ if (a->follow_symlinks || archive_entry_filetype(entry) != AE_IFLNK) fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY | O_NONBLOCK); /* Restore working directory if openat() operation failed or * the file is a symbolic link. */ if (fd < 0) tree_enter_working_dir(t); /* The current direcotry fd is needed at * archive_read_disk_entry_from_file() function to read link data * with readlinkat(). */ a->entry_wd_fd = tree_current_dir_fd(t); #endif /* * Populate the archive_entry with metadata from the disk. */ r = archive_read_disk_entry_from_file(&(a->archive), entry, fd, st); /* Close the file descriptor used for reding the current file * metadata at archive_read_disk_entry_from_file(). */ if (fd >= 0) close(fd); /* Return to the initial directory. */ tree_enter_initial_dir(t); archive_entry_copy_sourcepath(entry, tree_current_path(t)); /* * EOF and FATAL are persistent at this layer. By * modifying the state, we guarantee that future calls to * read a header or read data will fail. */ switch (r) { case ARCHIVE_EOF: a->archive.state = ARCHIVE_STATE_EOF; break; case ARCHIVE_OK: case ARCHIVE_WARN: t->entry_total = 0; if (archive_entry_filetype(entry) == AE_IFREG) { t->nlink = archive_entry_nlink(entry); t->entry_remaining_bytes = archive_entry_size(entry); t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0; if (!t->entry_eof && setup_sparse(a, entry) != ARCHIVE_OK) return (ARCHIVE_FATAL); } else { t->entry_remaining_bytes = 0; t->entry_eof = 1; } a->archive.state = ARCHIVE_STATE_DATA; break; case ARCHIVE_RETRY: break; case ARCHIVE_FATAL: a->archive.state = ARCHIVE_STATE_FATAL; break; } return (r); } static int setup_sparse(struct archive_read_disk *a, struct archive_entry *entry) { struct tree *t = a->tree; int64_t length, offset; int i; t->sparse_count = archive_entry_sparse_reset(entry); if (t->sparse_count+1 > t->sparse_list_size) { free(t->sparse_list); t->sparse_list_size = t->sparse_count + 1; t->sparse_list = malloc(sizeof(t->sparse_list[0]) * t->sparse_list_size); if (t->sparse_list == NULL) { t->sparse_list_size = 0; archive_set_error(&a->archive, ENOMEM, "Can't allocate data"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } } for (i = 0; i < t->sparse_count; i++) { archive_entry_sparse_next(entry, &offset, &length); t->sparse_list[i].offset = offset; t->sparse_list[i].length = length; } if (i == 0) { t->sparse_list[i].offset = 0; t->sparse_list[i].length = archive_entry_size(entry); } else { t->sparse_list[i].offset = archive_entry_size(entry); t->sparse_list[i].length = 0; } t->current_sparse = t->sparse_list; return (ARCHIVE_OK); } /* * Called by the client to mark the directory just returned from * tree_next() as needing to be visited. */ int archive_read_disk_descend(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct tree *t = a->tree; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_descend"); if (t->visit_type != TREE_REGULAR || !t->descend) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Ignored the request descending the current object"); return (ARCHIVE_WARN); } if (tree_current_is_physical_dir(t)) { tree_push(t, t->basename, t->current_filesystem_id, t->lst.st_dev, t->lst.st_ino, &t->restore_time); t->stack->flags |= isDir; } else if (tree_current_is_dir(t)) { tree_push(t, t->basename, t->current_filesystem_id, t->st.st_dev, t->st.st_ino, &t->restore_time); t->stack->flags |= isDirLink; } t->descend = 0; return (ARCHIVE_OK); } int archive_read_disk_open(struct archive *_a, const char *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, "archive_read_disk_open"); archive_clear_error(&a->archive); return (_archive_read_disk_open(_a, pathname)); } int archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; struct archive_string path; int ret; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, "archive_read_disk_open_w"); archive_clear_error(&a->archive); /* Make a char string from a wchar_t string. */ archive_string_init(&path); if (archive_string_append_from_wcs(&path, pathname, wcslen(pathname)) != 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't convert a path to a char string"); a->archive.state = ARCHIVE_STATE_FATAL; ret = ARCHIVE_FATAL; } else ret = _archive_read_disk_open(_a, path.s); archive_string_free(&path); return (ret); } static int _archive_read_disk_open(struct archive *_a, const char *pathname) { struct archive_read_disk *a = (struct archive_read_disk *)_a; if (a->tree != NULL) a->tree = tree_reopen(a->tree, pathname, a->restore_time); else a->tree = tree_open(pathname, a->symlink_mode, a->restore_time); if (a->tree == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate tar data"); a->archive.state = ARCHIVE_STATE_FATAL; return (ARCHIVE_FATAL); } a->archive.state = ARCHIVE_STATE_HEADER; return (ARCHIVE_OK); } /* * Return a current filesystem ID which is index of the filesystem entry * you've visited through archive_read_disk. */ int archive_read_disk_current_filesystem(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem_id); } static int update_current_filesystem(struct archive_read_disk *a, int64_t dev) { struct tree *t = a->tree; int i, fid; if (t->current_filesystem != NULL && t->current_filesystem->dev == dev) return (ARCHIVE_OK); for (i = 0; i < t->max_filesystem_id; i++) { if (t->filesystem_table[i].dev == dev) { /* There is the filesytem ID we've already generated. */ t->current_filesystem_id = i; t->current_filesystem = &(t->filesystem_table[i]); return (ARCHIVE_OK); } } /* * This is the new filesytem which we have to generate a new ID for. */ fid = t->max_filesystem_id++; if (t->max_filesystem_id > t->allocated_filesytem) { size_t s; s = t->max_filesystem_id * 2; t->filesystem_table = realloc(t->filesystem_table, s * sizeof(*t->filesystem_table)); if (t->filesystem_table == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate tar data"); return (ARCHIVE_FATAL); } t->allocated_filesytem = s; } t->current_filesystem_id = fid; t->current_filesystem = &(t->filesystem_table[fid]); t->current_filesystem->dev = dev; t->current_filesystem->allocation_ptr = NULL; t->current_filesystem->buff = NULL; /* Setup the current filesystem properties which depend on * platform specific. */ return (setup_current_filesystem(a)); } /* * Returns 1 if current filesystem is generated filesystem, 0 if it is not * or -1 if it is unknown. */ int archive_read_disk_current_filesystem_is_synthetic(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem->synthetic); } /* * Returns 1 if current filesystem is remote filesystem, 0 if it is not * or -1 if it is unknown. */ int archive_read_disk_current_filesystem_is_remote(struct archive *_a) { struct archive_read_disk *a = (struct archive_read_disk *)_a; archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_read_disk_current_filesystem"); return (a->tree->current_filesystem->remote); } #if defined(_PC_REC_INCR_XFER_SIZE) && defined(_PC_REC_MAX_XFER_SIZE) &&\ defined(_PC_REC_MIN_XFER_SIZE) && defined(_PC_REC_XFER_ALIGN) static int get_xfer_size(struct tree *t, int fd, const char *path) { t->current_filesystem->xfer_align = -1; errno = 0; if (fd >= 0) { t->current_filesystem->incr_xfer_size = fpathconf(fd, _PC_REC_INCR_XFER_SIZE); t->current_filesystem->max_xfer_size = fpathconf(fd, _PC_REC_MAX_XFER_SIZE); t->current_filesystem->min_xfer_size = fpathconf(fd, _PC_REC_MIN_XFER_SIZE); t->current_filesystem->xfer_align = fpathconf(fd, _PC_REC_XFER_ALIGN); } else if (path != NULL) { t->current_filesystem->incr_xfer_size = pathconf(path, _PC_REC_INCR_XFER_SIZE); t->current_filesystem->max_xfer_size = pathconf(path, _PC_REC_MAX_XFER_SIZE); t->current_filesystem->min_xfer_size = pathconf(path, _PC_REC_MIN_XFER_SIZE); t->current_filesystem->xfer_align = pathconf(path, _PC_REC_XFER_ALIGN); } /* At least we need an alignment size. */ if (t->current_filesystem->xfer_align == -1) return ((errno == EINVAL)?1:-1); else return (0); } #else static int get_xfer_size(struct tree *t, int fd, const char *path) { (void)t; /* UNUSED */ (void)fd; /* UNUSED */ (void)path; /* UNUSED */ return (1);/* Not supported */ } #endif #if defined(HAVE_STATFS) && defined(HAVE_FSTATFS) && defined(MNT_LOCAL) \ && !defined(ST_LOCAL) /* * Gather current filesystem properties on FreeBSD, OpenBSD and Mac OS X. */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statfs sfs; #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC) struct xvfsconf vfc; #endif int r, xr = 0; #if !defined(HAVE_STRUCT_STATFS_F_NAMEMAX) long nm; #endif t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } r = fstatfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else r = statfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { r = fstatfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); } if (r == -1 || xr == -1) { archive_set_error(&a->archive, errno, "statfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ t->current_filesystem->xfer_align = sfs.f_bsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_iosize; t->current_filesystem->incr_xfer_size = sfs.f_iosize; } if (sfs.f_flags & MNT_LOCAL) t->current_filesystem->remote = 0; else t->current_filesystem->remote = 1; #if defined(HAVE_GETVFSBYNAME) && defined(VFCF_SYNTHETIC) r = getvfsbyname(sfs.f_fstypename, &vfc); if (r == -1) { archive_set_error(&a->archive, errno, "getvfsbyname failed"); return (ARCHIVE_FAILED); } if (vfc.vfc_flags & VFCF_SYNTHETIC) t->current_filesystem->synthetic = 1; else t->current_filesystem->synthetic = 0; #endif #if defined(MNT_NOATIME) if (sfs.f_flags & MNT_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ #if defined(HAVE_STRUCT_STATFS_F_NAMEMAX) t->current_filesystem->name_max = sfs.f_namemax; #else /* Mac OS X does not have f_namemax in struct statfs. */ if (tree_current_is_symblic_link_target(t)) nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX); else nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX); if (nm == -1) t->current_filesystem->name_max = NAME_MAX; else t->current_filesystem->name_max = nm; #endif #endif /* HAVE_READDIR_R */ return (ARCHIVE_OK); } #elif (defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS)) && defined(ST_LOCAL) /* * Gather current filesystem properties on NetBSD */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statvfs sfs; int r, xr = 0; t->current_filesystem->synthetic = -1; if (tree_current_is_symblic_link_target(t)) { r = statvfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); } else { #ifdef HAVE_FSTATVFS r = fstatvfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #else r = statvfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1) { t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statvfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* Usuall come here unless NetBSD supports _PC_REC_XFER_ALIGN * for pathconf() function. */ t->current_filesystem->xfer_align = sfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_iosize; t->current_filesystem->incr_xfer_size = sfs.f_iosize; } if (sfs.f_flag & ST_LOCAL) t->current_filesystem->remote = 0; else t->current_filesystem->remote = 1; if (sfs.f_flag & ST_NOATIME) t->current_filesystem->noatime = 1; else t->current_filesystem->noatime = 0; /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namemax; return (ARCHIVE_OK); } #elif defined(HAVE_SYS_STATFS_H) && defined(HAVE_LINUX_MAGIC_H) &&\ defined(HAVE_STATFS) && defined(HAVE_FSTATFS) /* * Note: statfs is deprecated since LSB 3.2 */ #ifndef CIFS_SUPER_MAGIC #define CIFS_SUPER_MAGIC 0xFF534D42 #endif #ifndef DEVFS_SUPER_MAGIC #define DEVFS_SUPER_MAGIC 0x1373 #endif /* * Gather current filesystem properties on Linux */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statfs sfs; struct statvfs svfs; int r, vr = 0, xr = 0; if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } vr = fstatvfs(fd, &svfs);/* for f_flag, mount flags */ r = fstatfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else vr = statvfs(tree_current_access_path(t), &svfs); r = statfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { #ifdef HAVE_FSTATFS vr = fstatvfs(tree_current_dir_fd(t), &svfs); r = fstatfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #elif defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) #error "Unexpected case. Please tell us about this error." #else vr = statvfs(".", &svfs); r = statfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1 || vr == -1) { t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ t->current_filesystem->xfer_align = svfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = svfs.f_bsize; t->current_filesystem->incr_xfer_size = svfs.f_bsize; } switch (sfs.f_type) { case AFS_SUPER_MAGIC: case CIFS_SUPER_MAGIC: case CODA_SUPER_MAGIC: case NCP_SUPER_MAGIC:/* NetWare */ case NFS_SUPER_MAGIC: case SMB_SUPER_MAGIC: t->current_filesystem->remote = 1; t->current_filesystem->synthetic = 0; break; case DEVFS_SUPER_MAGIC: case PROC_SUPER_MAGIC: case USBDEVICE_SUPER_MAGIC: t->current_filesystem->remote = 0; t->current_filesystem->synthetic = 1; break; default: t->current_filesystem->remote = 0; t->current_filesystem->synthetic = 0; break; } #if defined(ST_NOATIME) if (svfs.f_flag & ST_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namelen; #endif return (ARCHIVE_OK); } #elif defined(HAVE_SYS_STATVFS_H) &&\ (defined(HAVE_STATVFS) || defined(HAVE_FSTATVFS)) /* * Gather current filesystem properties on other posix platform. */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; struct statvfs sfs; int r, xr = 0; t->current_filesystem->synthetic = -1;/* Not supported */ t->current_filesystem->remote = -1;/* Not supported */ if (tree_current_is_symblic_link_target(t)) { #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) /* * Get file system statistics on any directory * where current is. */ int fd = openat(tree_current_dir_fd(t), tree_current_access_path(t), O_RDONLY); if (fd < 0) { archive_set_error(&a->archive, errno, "openat failed"); return (ARCHIVE_FAILED); } r = fstatvfs(fd, &sfs); if (r == 0) xr = get_xfer_size(t, fd, NULL); close(fd); #else r = statvfs(tree_current_access_path(t), &sfs); if (r == 0) xr = get_xfer_size(t, -1, tree_current_access_path(t)); #endif } else { #ifdef HAVE_FSTATVFS r = fstatvfs(tree_current_dir_fd(t), &sfs); if (r == 0) xr = get_xfer_size(t, tree_current_dir_fd(t), NULL); #elif defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) #error "Unexpected case. Please tell us about this error." #else r = statvfs(".", &sfs); if (r == 0) xr = get_xfer_size(t, -1, "."); #endif } if (r == -1 || xr == -1) { t->current_filesystem->synthetic = -1; t->current_filesystem->remote = -1; archive_set_error(&a->archive, errno, "statvfs failed"); return (ARCHIVE_FAILED); } else if (xr == 1) { /* pathconf(_PC_REX_*) operations are not supported. */ t->current_filesystem->xfer_align = sfs.f_frsize; t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = sfs.f_bsize; t->current_filesystem->incr_xfer_size = sfs.f_bsize; } #if defined(ST_NOATIME) if (sfs.f_flag & ST_NOATIME) t->current_filesystem->noatime = 1; else #endif t->current_filesystem->noatime = 0; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ t->current_filesystem->name_max = sfs.f_namemax; #endif return (ARCHIVE_OK); } #else /* * Generic: Gather current filesystem properties. * TODO: Is this generic function really needed? */ static int setup_current_filesystem(struct archive_read_disk *a) { struct tree *t = a->tree; #if defined(_PC_NAME_MAX) && defined(HAVE_READDIR_R) long nm; #endif t->current_filesystem->synthetic = -1;/* Not supported */ t->current_filesystem->remote = -1;/* Not supported */ t->current_filesystem->noatime = 0; (void)get_xfer_size(t, -1, ".");/* Dummy call to avoid build error. */ t->current_filesystem->xfer_align = -1;/* Unknown */ t->current_filesystem->max_xfer_size = -1; t->current_filesystem->min_xfer_size = -1; t->current_filesystem->incr_xfer_size = -1; #if defined(HAVE_READDIR_R) /* Set maximum filename length. */ # if defined(_PC_NAME_MAX) if (tree_current_is_symblic_link_target(t)) nm = pathconf(tree_current_access_path(t), _PC_NAME_MAX); else nm = fpathconf(tree_current_dir_fd(t), _PC_NAME_MAX); if (nm == -1) # endif /* _PC_NAME_MAX */ /* * Some sysmtes (HP-UX or others?) incorrectly defined * NAME_MAX macro to be a smaller value. */ # if defined(NAME_MAX) && NAME_MAX >= 255 t->current_filesystem->name_max = NAME_MAX; # else /* No way to get a trusted value of maximum filename * length. */ t->current_filesystem->name_max = PATH_MAX; # endif /* NAME_MAX */ # if defined(_PC_NAME_MAX) else t->current_filesystem->name_max = nm; # endif /* _PC_NAME_MAX */ #endif /* HAVE_READDIR_R */ return (ARCHIVE_OK); } #endif static int close_and_restore_time(int fd, struct tree *t, struct restore_time *rt) { #ifndef HAVE_UTIMES (void)a; /* UNUSED */ return (close(fd)); #else #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__) struct timespec timespecs[2]; #endif struct timeval times[2]; if ((t->flags & needsRestoreTimes) == 0 || rt->noatime) { if (fd >= 0) return (close(fd)); else return (0); } #if defined(HAVE_FUTIMENS) && !defined(__CYGWIN__) timespecs[1].tv_sec = rt->mtime; timespecs[1].tv_nsec = rt->mtime_nsec; timespecs[0].tv_sec = rt->atime; timespecs[0].tv_nsec = rt->atime_nsec; /* futimens() is defined in POSIX.1-2008. */ if (futimens(fd, timespecs) == 0) return (close(fd)); #endif times[1].tv_sec = rt->mtime; times[1].tv_usec = rt->mtime_nsec / 1000; times[0].tv_sec = rt->atime; times[0].tv_usec = rt->atime_nsec / 1000; #if !defined(HAVE_FUTIMENS) && defined(HAVE_FUTIMES) && !defined(__CYGWIN__) if (futimes(fd, times) == 0) return (close(fd)); #endif close(fd); #if defined(HAVE_FUTIMESAT) if (futimesat(tree_current_dir_fd(t), rt->name, times) == 0) return (0); #endif #ifdef HAVE_LUTIMES if (lutimes(rt->name, times) != 0) #else if (AE_IFLNK != rt->filetype && utimes(rt->name, times) != 0) #endif return (-1); #endif return (0); } /* * Add a directory path to the current stack. */ static void tree_push(struct tree *t, const char *path, int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt) { struct tree_entry *te; te = malloc(sizeof(*te)); memset(te, 0, sizeof(*te)); te->next = t->stack; te->parent = t->current; if (te->parent) te->depth = te->parent->depth + 1; t->stack = te; archive_string_init(&te->name); te->symlink_parent_fd = -1; archive_strcpy(&te->name, path); te->flags = needsDescent | needsOpen | needsAscent; te->filesystem_id = filesystem_id; te->dev = dev; te->ino = ino; te->dirname_length = t->dirname_length; te->restore_time.name = te->name.s; if (rt != NULL) { te->restore_time.mtime = rt->mtime; te->restore_time.mtime_nsec = rt->mtime_nsec; te->restore_time.atime = rt->atime; te->restore_time.atime_nsec = rt->atime_nsec; te->restore_time.filetype = rt->filetype; te->restore_time.noatime = rt->noatime; } } /* * Append a name to the current dir path. */ static void tree_append(struct tree *t, const char *name, size_t name_length) { size_t size_needed; t->path.s[t->dirname_length] = '\0'; t->path.length = t->dirname_length; /* Strip trailing '/' from name, unless entire name is "/". */ while (name_length > 1 && name[name_length - 1] == '/') name_length--; /* Resize pathname buffer as needed. */ size_needed = name_length + t->dirname_length + 2; archive_string_ensure(&t->path, size_needed); /* Add a separating '/' if it's needed. */ if (t->dirname_length > 0 && t->path.s[archive_strlen(&t->path)-1] != '/') archive_strappend_char(&t->path, '/'); t->basename = t->path.s + archive_strlen(&t->path); archive_strncat(&t->path, name, name_length); t->restore_time.name = t->basename; } /* * Open a directory tree for traversal. */ static struct tree * tree_open(const char *path, int symlink_mode, int restore_time) { struct tree *t; if ((t = malloc(sizeof(*t))) == NULL) return (NULL); memset(t, 0, sizeof(*t)); archive_string_init(&t->path); archive_string_ensure(&t->path, 31); t->initial_symlink_mode = symlink_mode; return (tree_reopen(t, path, restore_time)); } static struct tree * tree_reopen(struct tree *t, const char *path, int restore_time) { t->flags = (restore_time)?needsRestoreTimes:0; t->visit_type = 0; t->tree_errno = 0; t->dirname_length = 0; t->depth = 0; t->descend = 0; t->current = NULL; t->d = INVALID_DIR_HANDLE; t->symlink_mode = t->initial_symlink_mode; archive_string_empty(&t->path); t->entry_fd = -1; t->entry_eof = 0; t->entry_remaining_bytes = 0; /* First item is set up a lot like a symlink traversal. */ tree_push(t, path, 0, 0, 0, NULL); t->stack->flags = needsFirstVisit; t->maxOpenCount = t->openCount = 1; t->initial_dir_fd = open(".", O_RDONLY); t->working_dir_fd = dup(t->initial_dir_fd); return (t); } static int tree_descent(struct tree *t) { int r = 0; #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) int new_fd; t->dirname_length = archive_strlen(&t->path); new_fd = openat(t->working_dir_fd, t->stack->name.s, O_RDONLY); if (new_fd < 0) { t->tree_errno = errno; r = TREE_ERROR_DIR; } else { t->depth++; /* If it is a link, set up fd for the ascent. */ if (t->stack->flags & isDirLink) { t->stack->symlink_parent_fd = t->working_dir_fd; t->openCount++; if (t->openCount > t->maxOpenCount) t->maxOpenCount = t->openCount; } else close(t->working_dir_fd); t->working_dir_fd = new_fd; } #else /* If it is a link, set up fd for the ascent. */ if (t->stack->flags & isDirLink) t->stack->symlink_parent_fd = t->working_dir_fd; else { close(t->working_dir_fd); t->openCount--; } t->working_dir_fd = -1; t->dirname_length = archive_strlen(&t->path); if (chdir(t->stack->name.s) != 0) { t->tree_errno = errno; r = TREE_ERROR_DIR; } else { t->depth++; t->working_dir_fd = open(".", O_RDONLY); t->openCount++; if (t->openCount > t->maxOpenCount) t->maxOpenCount = t->openCount; } #endif return (r); } /* * We've finished a directory; ascend back to the parent. */ static int tree_ascend(struct tree *t) { struct tree_entry *te; int r = 0, prev_dir_fd; te = t->stack; prev_dir_fd = t->working_dir_fd; #if defined(HAVE_OPENAT) && defined(HAVE_FSTATAT) && defined(HAVE_FDOPENDIR) if (te->flags & isDirLink) t->working_dir_fd = te->symlink_parent_fd; else { int new_fd = openat(t->working_dir_fd, "..", O_RDONLY); if (new_fd < 0) { t->tree_errno = errno; r = TREE_ERROR_FATAL; } else t->working_dir_fd = new_fd; } #else if (te->flags & isDirLink) { if (fchdir(te->symlink_parent_fd) != 0) { t->tree_errno = errno; r = TREE_ERROR_FATAL; } else t->working_dir_fd = te->symlink_parent_fd; } else { if (chdir("..") != 0) { t->tree_errno = errno; r = TREE_ERROR_FATAL; } else t->working_dir_fd = open(".", O_RDONLY); } #endif if (r == 0) { /* Current directory has been changed, we should * close an fd of previous working directory. */ close_and_restore_time(prev_dir_fd, t, &te->restore_time); if (te->flags & isDirLink) { t->openCount--; te->symlink_parent_fd = -1; } t->depth--; } return (r); } /* * Return to the initial directory where tree_open() was performed. */ static int tree_enter_initial_dir(struct tree *t) { int r = 0; if (t->flags & onWorkingDir) { r = fchdir(t->initial_dir_fd); if (r == 0) t->flags &= ~onWorkingDir; } return (r); } /* * Restore working directory of directory traversals. */ static int tree_enter_working_dir(struct tree *t) { int r = 0; /* * Change the current directory if really needed. * Sometimes this is unneeded when we did not do * descent. */ if (t->depth > 0 && (t->flags & onWorkingDir) == 0) { r = fchdir(t->working_dir_fd); if (r == 0) t->flags |= onWorkingDir; } return (r); } static int tree_current_dir_fd(struct tree *t) { return (t->working_dir_fd); } /* * Pop the working stack. */ static void tree_pop(struct tree *t) { struct tree_entry *te; t->path.s[t->dirname_length] = '\0'; t->path.length = t->dirname_length; if (t->stack == t->current && t->current != NULL) t->current = t->current->parent; te = t->stack; t->stack = te->next; t->dirname_length = te->dirname_length; t->basename = t->path.s + t->dirname_length; while (t->basename[0] == '/') t->basename++; archive_string_free(&te->name); free(te); } /* * Get the next item in the tree traversal. */ static int tree_next(struct tree *t) { int r; while (t->stack != NULL) { /* If there's an open dir, get the next entry from there. */ if (t->d != INVALID_DIR_HANDLE) { r = tree_dir_next_posix(t); if (r == 0) continue; return (r); } if (t->stack->flags & needsFirstVisit) { /* Top stack item needs a regular visit. */ t->current = t->stack; tree_append(t, t->stack->name.s, archive_strlen(&(t->stack->name))); /* t->dirname_length = t->path_length; */ /* tree_pop(t); */ t->stack->flags &= ~needsFirstVisit; return (t->visit_type = TREE_REGULAR); } else if (t->stack->flags & needsDescent) { /* Top stack item is dir to descend into. */ t->current = t->stack; tree_append(t, t->stack->name.s, archive_strlen(&(t->stack->name))); t->stack->flags &= ~needsDescent; r = tree_descent(t); if (r != 0) { tree_pop(t); t->visit_type = r; } else t->visit_type = TREE_POSTDESCENT; return (t->visit_type); } else if (t->stack->flags & needsOpen) { t->stack->flags &= ~needsOpen; r = tree_dir_next_posix(t); if (r == 0) continue; return (r); } else if (t->stack->flags & needsAscent) { /* Top stack item is dir and we're done with it. */ r = tree_ascend(t); tree_pop(t); t->visit_type = r != 0 ? r : TREE_POSTASCENT; return (t->visit_type); } else { /* Top item on stack is dead. */ tree_pop(t); t->flags &= ~hasLstat; t->flags &= ~hasStat; } } return (t->visit_type = 0); } static int tree_dir_next_posix(struct tree *t) { int r; const char *name; size_t namelen; if (t->d == NULL) { #if defined(HAVE_READDIR_R) size_t dirent_size; #endif #if defined(HAVE_FDOPENDIR) if ((t->d = fdopendir(dup(t->working_dir_fd))) == NULL) { #else if ((t->d = opendir(".")) == NULL) { #endif r = tree_ascend(t); /* Undo "chdir" */ tree_pop(t); t->tree_errno = errno; t->visit_type = r != 0 ? r : TREE_ERROR_DIR; return (t->visit_type); } #if defined(HAVE_READDIR_R) dirent_size = offsetof(struct dirent, d_name) + t->filesystem_table[t->current->filesystem_id].name_max + 1; if (t->dirent == NULL || t->dirent_allocated < dirent_size) { free(t->dirent); t->dirent = malloc(dirent_size); if (t->dirent == NULL) { closedir(t->d); t->d = INVALID_DIR_HANDLE; (void)tree_ascend(t); tree_pop(t); t->tree_errno = ENOMEM; t->visit_type = TREE_ERROR_DIR; return (t->visit_type); } t->dirent_allocated = dirent_size; } #endif /* HAVE_READDIR_R */ } for (;;) { #if defined(HAVE_READDIR_R) r = readdir_r(t->d, t->dirent, &t->de); if (r != 0 || t->de == NULL) { #else errno = 0; t->de = readdir(t->d); if (t->de == NULL) { r = errno; #endif closedir(t->d); t->d = INVALID_DIR_HANDLE; if (r != 0) { t->tree_errno = r; t->visit_type = TREE_ERROR_DIR; return (t->visit_type); } else return (0); } name = t->de->d_name; namelen = D_NAMELEN(t->de); t->flags &= ~hasLstat; t->flags &= ~hasStat; if (name[0] == '.' && name[1] == '\0') continue; if (name[0] == '.' && name[1] == '.' && name[2] == '\0') continue; tree_append(t, name, namelen); return (t->visit_type = TREE_REGULAR); } } /* * Get the stat() data for the entry just returned from tree_next(). */ static const struct stat * tree_current_stat(struct tree *t) { if (!(t->flags & hasStat)) { #ifdef HAVE_FSTATAT if (fstatat(tree_current_dir_fd(t), tree_current_access_path(t), &t->st, 0) != 0) #else if (stat(tree_current_access_path(t), &t->st) != 0) #endif return NULL; t->flags |= hasStat; } return (&t->st); } /* * Get the lstat() data for the entry just returned from tree_next(). */ static const struct stat * tree_current_lstat(struct tree *t) { if (!(t->flags & hasLstat)) { #ifdef HAVE_FSTATAT if (fstatat(tree_current_dir_fd(t), tree_current_access_path(t), &t->lst, AT_SYMLINK_NOFOLLOW) != 0) #else if (lstat(tree_current_access_path(t), &t->lst) != 0) #endif return NULL; t->flags |= hasLstat; } return (&t->lst); } /* * Test whether current entry is a dir or link to a dir. */ static int tree_current_is_dir(struct tree *t) { const struct stat *st; /* * If we already have lstat() info, then try some * cheap tests to determine if this is a dir. */ if (t->flags & hasLstat) { /* If lstat() says it's a dir, it must be a dir. */ if (S_ISDIR(tree_current_lstat(t)->st_mode)) return 1; /* Not a dir; might be a link to a dir. */ /* If it's not a link, then it's not a link to a dir. */ if (!S_ISLNK(tree_current_lstat(t)->st_mode)) return 0; /* * It's a link, but we don't know what it's a link to, * so we'll have to use stat(). */ } st = tree_current_stat(t); /* If we can't stat it, it's not a dir. */ if (st == NULL) return 0; /* Use the definitive test. Hopefully this is cached. */ return (S_ISDIR(st->st_mode)); } /* * Test whether current entry is a physical directory. Usually, we * already have at least one of stat() or lstat() in memory, so we * use tricks to try to avoid an extra trip to the disk. */ static int tree_current_is_physical_dir(struct tree *t) { const struct stat *st; /* * If stat() says it isn't a dir, then it's not a dir. * If stat() data is cached, this check is free, so do it first. */ if ((t->flags & hasStat) && (!S_ISDIR(tree_current_stat(t)->st_mode))) return 0; /* * Either stat() said it was a dir (in which case, we have * to determine whether it's really a link to a dir) or * stat() info wasn't available. So we use lstat(), which * hopefully is already cached. */ st = tree_current_lstat(t); /* If we can't stat it, it's not a dir. */ if (st == NULL) return 0; /* Use the definitive test. Hopefully this is cached. */ return (S_ISDIR(st->st_mode)); } /* * Test whether the same file has been in the tree as its parent. */ static int tree_target_is_same_as_parent(struct tree *t, const struct stat *st) { struct tree_entry *te; for (te = t->current->parent; te != NULL; te = te->parent) { - if (te->dev == st->st_dev && te->ino == st->st_ino) + if (te->dev == (int64_t)st->st_dev && + te->ino == (int64_t)st->st_ino) return (1); } return (0); } /* * Test whether the current file is symbolic link target and * on the other filesystem. */ static int tree_current_is_symblic_link_target(struct tree *t) { static const struct stat *lst, *st; lst = tree_current_lstat(t); st = tree_current_stat(t); - return (st != NULL && st->st_dev == t->current_filesystem->dev && + return (st != NULL && + (int64_t)st->st_dev == t->current_filesystem->dev && st->st_dev != lst->st_dev); } /* * Return the access path for the entry just returned from tree_next(). */ static const char * tree_current_access_path(struct tree *t) { return (t->basename); } /* * Return the full path for the entry just returned from tree_next(). */ static const char * tree_current_path(struct tree *t) { return (t->path.s); } /* * Terminate the traversal. */ static void tree_close(struct tree *t) { if (t == NULL) return; if (t->entry_fd >= 0) { close_and_restore_time(t->entry_fd, t, &t->restore_time); t->entry_fd = -1; } /* Close the handle of readdir(). */ if (t->d != INVALID_DIR_HANDLE) { closedir(t->d); t->d = INVALID_DIR_HANDLE; } /* Release anything remaining in the stack. */ while (t->stack != NULL) { if (t->stack->flags & isDirLink) close(t->stack->symlink_parent_fd); tree_pop(t); } if (t->working_dir_fd >= 0) { close(t->working_dir_fd); t->working_dir_fd = -1; } if (t->initial_dir_fd >= 0) { close(t->initial_dir_fd); t->initial_dir_fd = -1; } } /* * Release any resources. */ static void tree_free(struct tree *t) { int i; if (t == NULL) return; archive_string_free(&t->path); #if defined(HAVE_READDIR_R) free(t->dirent); #endif free(t->sparse_list); for (i = 0; i < t->max_filesystem_id; i++) free(t->filesystem_table[i].allocation_ptr); free(t->filesystem_table); free(t); } #endif Index: vendor/libarchive/dist/libarchive/archive_read_support_format_rar.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_read_support_format_rar.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_read_support_format_rar.c (revision 231297) @@ -1,2574 +1,2574 @@ /*- * Copyright (c) 2003-2007 Tim Kientzle * Copyright (c) 2011 Andres Mejia * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" #ifdef HAVE_ERRNO_H #include #endif #include #include #ifdef HAVE_ZLIB_H #include /* crc32 */ #endif #include "archive.h" #ifndef HAVE_ZLIB_H #include "archive_crc32.h" #endif #include "archive_endian.h" #include "archive_entry.h" #include "archive_entry_locale.h" #include "archive_ppmd7_private.h" #include "archive_private.h" #include "archive_read_private.h" /* RAR signature, also known as the mark header */ #define RAR_SIGNATURE "\x52\x61\x72\x21\x1A\x07\x00" /* Header types */ #define MARK_HEAD 0x72 #define MAIN_HEAD 0x73 #define FILE_HEAD 0x74 #define COMM_HEAD 0x75 #define AV_HEAD 0x76 #define SUB_HEAD 0x77 #define PROTECT_HEAD 0x78 #define SIGN_HEAD 0x79 #define NEWSUB_HEAD 0x7a #define ENDARC_HEAD 0x7b /* Main Header Flags */ #define MHD_VOLUME 0x0001 #define MHD_COMMENT 0x0002 #define MHD_LOCK 0x0004 #define MHD_SOLID 0x0008 #define MHD_NEWNUMBERING 0x0010 #define MHD_AV 0x0020 #define MHD_PROTECT 0x0040 #define MHD_PASSWORD 0x0080 #define MHD_FIRSTVOLUME 0x0100 #define MHD_ENCRYPTVER 0x0200 /* Flags common to all headers */ #define HD_MARKDELETION 0x4000 #define HD_ADD_SIZE_PRESENT 0x8000 /* File Header Flags */ #define FHD_SPLIT_BEFORE 0x0001 #define FHD_SPLIT_AFTER 0x0002 #define FHD_PASSWORD 0x0004 #define FHD_COMMENT 0x0008 #define FHD_SOLID 0x0010 #define FHD_LARGE 0x0100 #define FHD_UNICODE 0x0200 #define FHD_SALT 0x0400 #define FHD_VERSION 0x0800 #define FHD_EXTTIME 0x1000 #define FHD_EXTFLAGS 0x2000 /* File dictionary sizes */ #define DICTIONARY_SIZE_64 0x00 #define DICTIONARY_SIZE_128 0x20 #define DICTIONARY_SIZE_256 0x40 #define DICTIONARY_SIZE_512 0x60 #define DICTIONARY_SIZE_1024 0x80 #define DICTIONARY_SIZE_2048 0xA0 #define DICTIONARY_SIZE_4096 0xC0 #define FILE_IS_DIRECTORY 0xE0 #define DICTIONARY_MASK FILE_IS_DIRECTORY /* OS Flags */ #define OS_MSDOS 0 #define OS_OS2 1 #define OS_WIN32 2 #define OS_UNIX 3 #define OS_MAC_OS 4 #define OS_BEOS 5 /* Compression Methods */ #define COMPRESS_METHOD_STORE 0x30 /* LZSS */ #define COMPRESS_METHOD_FASTEST 0x31 #define COMPRESS_METHOD_FAST 0x32 #define COMPRESS_METHOD_NORMAL 0x33 /* PPMd Variant H */ #define COMPRESS_METHOD_GOOD 0x34 #define COMPRESS_METHOD_BEST 0x35 #define CRC_POLYNOMIAL 0xEDB88320 #define NS_UNIT 10000000 #define DICTIONARY_MAX_SIZE 0x400000 #define MAINCODE_SIZE 299 #define OFFSETCODE_SIZE 60 #define LOWOFFSETCODE_SIZE 17 #define LENGTHCODE_SIZE 28 #define HUFFMAN_TABLE_SIZE \ MAINCODE_SIZE + OFFSETCODE_SIZE + LOWOFFSETCODE_SIZE + LENGTHCODE_SIZE #define MAX_SYMBOL_LENGTH 0xF #define MAX_SYMBOLS 20 /* * Considering L1,L2 cache miss and a calling of write system-call, * the best size of the output buffer(uncompressed buffer) is 128K. * If the structure of extracting process is changed, this value * might be researched again. */ #define UNP_BUFFER_SIZE (128 * 1024) /* Define this here for non-Windows platforms */ #if !((defined(__WIN32__) || defined(_WIN32) || defined(__WIN32)) && !defined(__CYGWIN__)) #define FILE_ATTRIBUTE_DIRECTORY 0x10 #endif /* Fields common to all headers */ struct rar_header { char crc[2]; char type; char flags[2]; char size[2]; }; /* Fields common to all file headers */ struct rar_file_header { char pack_size[4]; char unp_size[4]; char host_os; char file_crc[4]; char file_time[4]; char unp_ver; char method; char name_size[2]; char file_attr[4]; }; struct huffman_tree_node { int branches[2]; }; struct huffman_table_entry { unsigned int length; int value; }; struct huffman_code { struct huffman_tree_node *tree; int numentries; int minlength; int maxlength; int tablesize; struct huffman_table_entry *table; }; struct lzss { unsigned char *window; int mask; int64_t position; }; struct rar { /* Entries from main RAR header */ unsigned main_flags; unsigned long file_crc; char reserved1[2]; char reserved2[4]; char encryptver; /* File header entries */ char compression_method; unsigned file_flags; int64_t packed_size; int64_t unp_size; time_t mtime; long mnsec; mode_t mode; char *filename; size_t filename_allocated; /* File header optional entries */ char salt[8]; time_t atime; long ansec; time_t ctime; long cnsec; time_t arctime; long arcnsec; /* Fields to help with tracking decompression of files. */ int64_t bytes_unconsumed; int64_t bytes_remaining; int64_t bytes_uncopied; int64_t offset; int64_t offset_outgoing; char valid; unsigned int unp_offset; unsigned int unp_buffer_size; unsigned char *unp_buffer; unsigned int dictionary_size; char start_new_block; char entry_eof; unsigned long crc_calculated; int found_first_header; /* LZSS members */ struct huffman_code maincode; struct huffman_code offsetcode; struct huffman_code lowoffsetcode; struct huffman_code lengthcode; unsigned char lengthtable[HUFFMAN_TABLE_SIZE]; struct lzss lzss; char output_last_match; unsigned int lastlength; unsigned int lastoffset; unsigned int oldoffset[4]; unsigned int lastlowoffset; unsigned int numlowoffsetrepeats; int64_t filterstart; char start_new_table; /* PPMd Variant H members */ char ppmd_valid; char ppmd_eod; char is_ppmd_block; int ppmd_escape; CPpmd7 ppmd7_context; CPpmd7z_RangeDec range_dec; IByteIn bytein; /* * String conversion object. */ int init_default_conversion; struct archive_string_conv *sconv_default; struct archive_string_conv *opt_sconv; struct archive_string_conv *sconv_utf8; struct archive_string_conv *sconv_utf16be; /* * Bit stream reader. */ struct rar_br { #define CACHE_TYPE uint64_t #define CACHE_BITS (8 * sizeof(CACHE_TYPE)) /* Cache buffer. */ CACHE_TYPE cache_buffer; /* Indicates how many bits avail in cache_buffer. */ int cache_avail; ssize_t avail_in; const unsigned char *next_in; } br; }; static int archive_read_format_rar_bid(struct archive_read *, int); static int archive_read_format_rar_options(struct archive_read *, const char *, const char *); static int archive_read_format_rar_read_header(struct archive_read *, struct archive_entry *); static int archive_read_format_rar_read_data(struct archive_read *, const void **, size_t *, int64_t *); static int archive_read_format_rar_read_data_skip(struct archive_read *a); static int archive_read_format_rar_cleanup(struct archive_read *); /* Support functions */ static int read_header(struct archive_read *, struct archive_entry *, char); -static time_t get_time(int time); +static time_t get_time(int); static int read_exttime(const char *, struct rar *, const char *); static int read_symlink_stored(struct archive_read *, struct archive_entry *, struct archive_string_conv *); static int read_data_stored(struct archive_read *, const void **, size_t *, int64_t *); static int read_data_compressed(struct archive_read *, const void **, size_t *, int64_t *); static int rar_br_preparation(struct archive_read *, struct rar_br *); static int parse_codes(struct archive_read *); static void free_codes(struct archive_read *); static int read_next_symbol(struct archive_read *, struct huffman_code *); static int create_code(struct archive_read *, struct huffman_code *, unsigned char *, int, char); static int add_value(struct archive_read *, struct huffman_code *, int, int, int); static int new_node(struct huffman_code *); static int make_table(struct archive_read *, struct huffman_code *); static int make_table_recurse(struct archive_read *, struct huffman_code *, int, struct huffman_table_entry *, int, int); static int64_t expand(struct archive_read *, int64_t); static int copy_from_lzss_window(struct archive_read *, const void **, int64_t, int); /* * Bit stream reader. */ /* Check that the cache buffer has enough bits. */ #define rar_br_has(br, n) ((br)->cache_avail >= n) /* Get compressed data by bit. */ #define rar_br_bits(br, n) \ (((uint32_t)((br)->cache_buffer >> \ ((br)->cache_avail - (n)))) & cache_masks[n]) #define rar_br_bits_forced(br, n) \ (((uint32_t)((br)->cache_buffer << \ ((n) - (br)->cache_avail))) & cache_masks[n]) /* Read ahead to make sure the cache buffer has enough compressed data we * will use. * True : completed, there is enough data in the cache buffer. * False : there is no data in the stream. */ #define rar_br_read_ahead(a, br, n) \ ((rar_br_has(br, (n)) || rar_br_fillup(a, br)) || rar_br_has(br, (n))) /* Notify how many bits we consumed. */ #define rar_br_consume(br, n) ((br)->cache_avail -= (n)) #define rar_br_consume_unalined_bits(br) ((br)->cache_avail &= ~7) static const uint32_t cache_masks[] = { 0x00000000, 0x00000001, 0x00000003, 0x00000007, 0x0000000F, 0x0000001F, 0x0000003F, 0x0000007F, 0x000000FF, 0x000001FF, 0x000003FF, 0x000007FF, 0x00000FFF, 0x00001FFF, 0x00003FFF, 0x00007FFF, 0x0000FFFF, 0x0001FFFF, 0x0003FFFF, 0x0007FFFF, 0x000FFFFF, 0x001FFFFF, 0x003FFFFF, 0x007FFFFF, 0x00FFFFFF, 0x01FFFFFF, 0x03FFFFFF, 0x07FFFFFF, 0x0FFFFFFF, 0x1FFFFFFF, 0x3FFFFFFF, 0x7FFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF }; /* * Shift away used bits in the cache data and fill it up with following bits. * Call this when cache buffer does not have enough bits you need. * * Returns 1 if the cache buffer is full. * Returns 0 if the cache buffer is not full; input buffer is empty. */ static int rar_br_fillup(struct archive_read *a, struct rar_br *br) { struct rar *rar = (struct rar *)(a->format->data); int n = CACHE_BITS - br->cache_avail; for (;;) { switch (n >> 3) { case 8: if (br->avail_in >= 8) { br->cache_buffer = ((uint64_t)br->next_in[0]) << 56 | ((uint64_t)br->next_in[1]) << 48 | ((uint64_t)br->next_in[2]) << 40 | ((uint64_t)br->next_in[3]) << 32 | ((uint32_t)br->next_in[4]) << 24 | ((uint32_t)br->next_in[5]) << 16 | ((uint32_t)br->next_in[6]) << 8 | (uint32_t)br->next_in[7]; br->next_in += 8; br->avail_in -= 8; br->cache_avail += 8 * 8; rar->bytes_unconsumed += 8; rar->bytes_remaining -= 8; return (1); } break; case 7: if (br->avail_in >= 7) { br->cache_buffer = (br->cache_buffer << 56) | ((uint64_t)br->next_in[0]) << 48 | ((uint64_t)br->next_in[1]) << 40 | ((uint64_t)br->next_in[2]) << 32 | ((uint32_t)br->next_in[3]) << 24 | ((uint32_t)br->next_in[4]) << 16 | ((uint32_t)br->next_in[5]) << 8 | (uint32_t)br->next_in[6]; br->next_in += 7; br->avail_in -= 7; br->cache_avail += 7 * 8; rar->bytes_unconsumed += 7; rar->bytes_remaining -= 7; return (1); } break; case 6: if (br->avail_in >= 6) { br->cache_buffer = (br->cache_buffer << 48) | ((uint64_t)br->next_in[0]) << 40 | ((uint64_t)br->next_in[1]) << 32 | ((uint32_t)br->next_in[2]) << 24 | ((uint32_t)br->next_in[3]) << 16 | ((uint32_t)br->next_in[4]) << 8 | (uint32_t)br->next_in[5]; br->next_in += 6; br->avail_in -= 6; br->cache_avail += 6 * 8; rar->bytes_unconsumed += 6; rar->bytes_remaining -= 6; return (1); } break; case 0: /* We have enough compressed data in * the cache buffer.*/ return (1); default: break; } if (br->avail_in <= 0) { if (rar->bytes_unconsumed > 0) { /* Consume as much as the decompressor * actually used. */ __archive_read_consume(a, rar->bytes_unconsumed); rar->bytes_unconsumed = 0; } br->next_in = __archive_read_ahead(a, 1, &(br->avail_in)); if (br->next_in == NULL) return (0); if (br->avail_in > rar->bytes_remaining) br->avail_in = rar->bytes_remaining; if (br->avail_in == 0) return (0); } br->cache_buffer = (br->cache_buffer << 8) | *br->next_in++; br->avail_in--; br->cache_avail += 8; n -= 8; rar->bytes_unconsumed++; rar->bytes_remaining--; } } static int rar_br_preparation(struct archive_read *a, struct rar_br *br) { struct rar *rar = (struct rar *)(a->format->data); if (rar->bytes_remaining > 0) { br->next_in = __archive_read_ahead(a, 1, &(br->avail_in)); if (br->next_in == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); return (ARCHIVE_FATAL); } if (br->avail_in > rar->bytes_remaining) br->avail_in = rar->bytes_remaining; if (br->cache_avail == 0) (void)rar_br_fillup(a, br); } return (ARCHIVE_OK); } /* Find last bit set */ static inline int rar_fls(unsigned int word) { word |= (word >> 1); word |= (word >> 2); word |= (word >> 4); word |= (word >> 8); word |= (word >> 16); return word - (word >> 1); } /* LZSS functions */ static inline int64_t lzss_position(struct lzss *lzss) { return lzss->position; } static inline int lzss_mask(struct lzss *lzss) { return lzss->mask; } static inline int lzss_size(struct lzss *lzss) { return lzss->mask + 1; } static inline int lzss_offset_for_position(struct lzss *lzss, int64_t pos) { return pos & lzss->mask; } static inline unsigned char * lzss_pointer_for_position(struct lzss *lzss, int64_t pos) { return &lzss->window[lzss_offset_for_position(lzss, pos)]; } static inline int lzss_current_offset(struct lzss *lzss) { return lzss_offset_for_position(lzss, lzss->position); } static inline uint8_t * lzss_current_pointer(struct lzss *lzss) { return lzss_pointer_for_position(lzss, lzss->position); } static inline void lzss_emit_literal(struct rar *rar, uint8_t literal) { *lzss_current_pointer(&rar->lzss) = literal; rar->lzss.position++; } static inline void lzss_emit_match(struct rar *rar, int offset, int length) { int dstoffs = lzss_current_offset(&rar->lzss); int srcoffs = (dstoffs - offset) & lzss_mask(&rar->lzss); int l, li, remaining; unsigned char *d, *s; remaining = length; while (remaining > 0) { l = remaining; if (dstoffs > srcoffs) { if (l > lzss_size(&rar->lzss) - dstoffs) l = lzss_size(&rar->lzss) - dstoffs; } else { if (l > lzss_size(&rar->lzss) - srcoffs) l = lzss_size(&rar->lzss) - srcoffs; } d = &(rar->lzss.window[dstoffs]); s = &(rar->lzss.window[srcoffs]); if ((dstoffs + l < srcoffs) || (srcoffs + l < dstoffs)) memcpy(d, s, l); else { for (li = 0; li < l; li++) d[li] = s[li]; } remaining -= l; dstoffs = (dstoffs + l) & lzss_mask(&(rar->lzss)); srcoffs = (srcoffs + l) & lzss_mask(&(rar->lzss)); } rar->lzss.position += length; } static void * ppmd_alloc(void *p, size_t size) { (void)p; return malloc(size); } static void ppmd_free(void *p, void *address) { (void)p; free(address); } static ISzAlloc g_szalloc = { ppmd_alloc, ppmd_free }; static Byte ppmd_read(void *p) { struct archive_read *a = ((IByteIn*)p)->a; struct rar *rar = (struct rar *)(a->format->data); struct rar_br *br = &(rar->br); Byte b; if (!rar_br_read_ahead(a, br, 8)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); rar->valid = 0; return 0; } b = rar_br_bits(br, 8); rar_br_consume(br, 8); return b; } int archive_read_support_format_rar(struct archive *_a) { struct archive_read *a = (struct archive_read *)_a; struct rar *rar; int r; archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, "archive_read_support_format_rar"); rar = (struct rar *)malloc(sizeof(*rar)); if (rar == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate rar data"); return (ARCHIVE_FATAL); } memset(rar, 0, sizeof(*rar)); r = __archive_read_register_format(a, rar, "rar", archive_read_format_rar_bid, archive_read_format_rar_options, archive_read_format_rar_read_header, archive_read_format_rar_read_data, archive_read_format_rar_read_data_skip, archive_read_format_rar_cleanup); if (r != ARCHIVE_OK) free(rar); return (r); } static int archive_read_format_rar_bid(struct archive_read *a, int best_bid) { const char *p; /* If there's already a bid > 30, we'll never win. */ if (best_bid > 30) return (-1); if ((p = __archive_read_ahead(a, 7, NULL)) == NULL) return (-1); if (memcmp(p, RAR_SIGNATURE, 7) == 0) return (30); if ((p[0] == 'M' && p[1] == 'Z') || memcmp(p, "\x7F\x45LF", 4) == 0) { /* This is a PE file */ ssize_t offset = 0x10000; ssize_t window = 4096; ssize_t bytes_avail; while (offset + window <= (1024 * 128)) { const char *buff = __archive_read_ahead(a, offset + window, &bytes_avail); if (buff == NULL) { /* Remaining bytes are less than window. */ window >>= 1; if (window < 0x40) return (0); continue; } p = buff + offset; while (p + 7 < buff + bytes_avail) { if (memcmp(p, RAR_SIGNATURE, 7) == 0) return (30); p += 0x10; } offset = p - buff; } } return (0); } static int skip_sfx(struct archive_read *a) { const void *h; const char *p, *q; size_t skip, total; ssize_t bytes, window; total = 0; window = 4096; while (total + window <= (1024 * 128)) { h = __archive_read_ahead(a, window, &bytes); if (h == NULL) { /* Remaining bytes are less than window. */ window >>= 1; if (window < 0x40) goto fatal; continue; } if (bytes < 0x40) goto fatal; p = h; q = p + bytes; /* * Scan ahead until we find something that looks * like the RAR header. */ while (p + 7 < q) { if (memcmp(p, RAR_SIGNATURE, 7) == 0) { skip = p - (const char *)h; __archive_read_consume(a, skip); return (ARCHIVE_OK); } p += 0x10; } skip = p - (const char *)h; __archive_read_consume(a, skip); total += skip; } fatal: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Couldn't find out RAR header"); return (ARCHIVE_FATAL); } static int archive_read_format_rar_options(struct archive_read *a, const char *key, const char *val) { struct rar *rar; int ret = ARCHIVE_FAILED; rar = (struct rar *)(a->format->data); if (strcmp(key, "hdrcharset") == 0) { if (val == NULL || val[0] == 0) archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "rar: hdrcharset option needs a character-set name"); else { rar->opt_sconv = archive_string_conversion_from_charset( &a->archive, val, 0); if (rar->opt_sconv != NULL) ret = ARCHIVE_OK; else ret = ARCHIVE_FATAL; } } else archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "rar: unknown keyword ``%s''", key); return (ret); } static int archive_read_format_rar_read_header(struct archive_read *a, struct archive_entry *entry) { const void *h; const char *p; struct rar *rar; size_t skip; char head_type; int ret; unsigned flags; a->archive.archive_format = ARCHIVE_FORMAT_RAR; if (a->archive.archive_format_name == NULL) a->archive.archive_format_name = "RAR"; rar = (struct rar *)(a->format->data); /* RAR files can be generated without EOF headers, so return ARCHIVE_EOF if * this fails. */ if ((h = __archive_read_ahead(a, 7, NULL)) == NULL) return (ARCHIVE_EOF); p = h; if (rar->found_first_header == 0 && ((p[0] == 'M' && p[1] == 'Z') || memcmp(p, "\x7F\x45LF", 4) == 0)) { /* This is an executable ? Must be self-extracting... */ ret = skip_sfx(a); if (ret < ARCHIVE_WARN) return (ret); } rar->found_first_header = 1; while (1) { unsigned long crc32_val; if ((h = __archive_read_ahead(a, 7, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; head_type = p[2]; switch(head_type) { case MARK_HEAD: if (memcmp(p, RAR_SIGNATURE, 7) != 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid marker header"); return (ARCHIVE_FATAL); } __archive_read_consume(a, 7); break; case MAIN_HEAD: rar->main_flags = archive_le16dec(p + 3); skip = archive_le16dec(p + 5); if (skip < 7 + sizeof(rar->reserved1) + sizeof(rar->reserved2)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } if ((h = __archive_read_ahead(a, skip, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; memcpy(rar->reserved1, p + 7, sizeof(rar->reserved1)); memcpy(rar->reserved2, p + 7 + sizeof(rar->reserved1), sizeof(rar->reserved2)); if (rar->main_flags & MHD_ENCRYPTVER) { if (skip < 7 + sizeof(rar->reserved1) + sizeof(rar->reserved2)+1) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } rar->encryptver = *(p + 7 + sizeof(rar->reserved1) + sizeof(rar->reserved2)); } if (rar->main_flags & MHD_VOLUME || rar->main_flags & MHD_FIRSTVOLUME) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "RAR volume support unavailable."); return (ARCHIVE_FATAL); } if (rar->main_flags & MHD_PASSWORD) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "RAR encryption support unavailable."); return (ARCHIVE_FATAL); } crc32_val = crc32(0, (const unsigned char *)p + 2, skip - 2); if ((crc32_val & 0xffff) != archive_le16dec(p)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Header CRC error"); return (ARCHIVE_FATAL); } __archive_read_consume(a, skip); break; case FILE_HEAD: return read_header(a, entry, head_type); case COMM_HEAD: case AV_HEAD: case SUB_HEAD: case PROTECT_HEAD: case SIGN_HEAD: flags = archive_le16dec(p + 3); skip = archive_le16dec(p + 5); if (skip < 7) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } if (skip > 7) { if ((h = __archive_read_ahead(a, skip, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; } if (flags & HD_ADD_SIZE_PRESENT) { if (skip < 7 + 4) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } skip += archive_le32dec(p + 7); if ((h = __archive_read_ahead(a, skip, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; } crc32_val = crc32(0, (const unsigned char *)p + 2, skip - 2); if ((crc32_val & 0xffff) != archive_le16dec(p)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Header CRC error"); return (ARCHIVE_FATAL); } __archive_read_consume(a, skip); break; case NEWSUB_HEAD: if ((ret = read_header(a, entry, head_type)) < ARCHIVE_WARN) return ret; break; case ENDARC_HEAD: return (ARCHIVE_EOF); default: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Bad RAR file"); return (ARCHIVE_FATAL); } } } static int archive_read_format_rar_read_data(struct archive_read *a, const void **buff, size_t *size, int64_t *offset) { struct rar *rar = (struct rar *)(a->format->data); int ret; if (rar->bytes_unconsumed > 0) { /* Consume as much as the decompressor actually used. */ __archive_read_consume(a, rar->bytes_unconsumed); rar->bytes_unconsumed = 0; } if (rar->entry_eof) { *buff = NULL; *size = 0; *offset = rar->offset; return (ARCHIVE_EOF); } switch (rar->compression_method) { case COMPRESS_METHOD_STORE: ret = read_data_stored(a, buff, size, offset); break; case COMPRESS_METHOD_FASTEST: case COMPRESS_METHOD_FAST: case COMPRESS_METHOD_NORMAL: case COMPRESS_METHOD_GOOD: case COMPRESS_METHOD_BEST: ret = read_data_compressed(a, buff, size, offset); if (ret != ARCHIVE_OK && ret != ARCHIVE_WARN) __archive_ppmd7_functions.Ppmd7_Free(&rar->ppmd7_context, &g_szalloc); break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Unsupported compression method for RAR file."); ret = ARCHIVE_FATAL; break; } return (ret); } static int archive_read_format_rar_read_data_skip(struct archive_read *a) { struct rar *rar; int64_t bytes_skipped; rar = (struct rar *)(a->format->data); if (rar->bytes_unconsumed > 0) { /* Consume as much as the decompressor actually used. */ __archive_read_consume(a, rar->bytes_unconsumed); rar->bytes_unconsumed = 0; } if (rar->bytes_remaining > 0) { bytes_skipped = __archive_read_consume(a, rar->bytes_remaining); if (bytes_skipped < 0) return (ARCHIVE_FATAL); } return (ARCHIVE_OK); } static int archive_read_format_rar_cleanup(struct archive_read *a) { struct rar *rar; rar = (struct rar *)(a->format->data); free_codes(a); free(rar->filename); free(rar->unp_buffer); free(rar->lzss.window); __archive_ppmd7_functions.Ppmd7_Free(&rar->ppmd7_context, &g_szalloc); free(rar); (a->format->data) = NULL; return (ARCHIVE_OK); } static int read_header(struct archive_read *a, struct archive_entry *entry, char head_type) { const void *h; const char *p, *endp; struct rar *rar; struct rar_header rar_header; struct rar_file_header file_header; int64_t header_size; unsigned filename_size, end; char *filename; char *strp; char packed_size[8]; char unp_size[8]; int ttime; struct archive_string_conv *sconv, *fn_sconv; unsigned long crc32_val; int ret = (ARCHIVE_OK), ret2; rar = (struct rar *)(a->format->data); /* Setup a string conversion object for non-rar-unicode filenames. */ sconv = rar->opt_sconv; if (sconv == NULL) { if (!rar->init_default_conversion) { rar->sconv_default = archive_string_default_conversion_for_read( &(a->archive)); rar->init_default_conversion = 1; } sconv = rar->sconv_default; } if ((h = __archive_read_ahead(a, 7, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; memcpy(&rar_header, p, sizeof(rar_header)); rar->file_flags = archive_le16dec(rar_header.flags); header_size = archive_le16dec(rar_header.size); - if (header_size < sizeof(file_header) + 7) { + if (header_size < (int64_t)sizeof(file_header) + 7) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } crc32_val = crc32(0, (const unsigned char *)p + 2, 7 - 2); __archive_read_consume(a, 7); if (!(rar->file_flags & FHD_SOLID)) { rar->compression_method = 0; rar->packed_size = 0; rar->unp_size = 0; rar->mtime = 0; rar->ctime = 0; rar->atime = 0; rar->arctime = 0; rar->mode = 0; memset(&rar->salt, 0, sizeof(rar->salt)); rar->atime = 0; rar->ansec = 0; rar->ctime = 0; rar->cnsec = 0; rar->mtime = 0; rar->mnsec = 0; rar->arctime = 0; rar->arcnsec = 0; } else { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "RAR solid archive support unavailable."); return (ARCHIVE_FATAL); } if ((h = __archive_read_ahead(a, header_size - 7, NULL)) == NULL) return (ARCHIVE_FATAL); /* File Header CRC check. */ crc32_val = crc32(crc32_val, h, header_size - 7); if ((crc32_val & 0xffff) != archive_le16dec(rar_header.crc)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Header CRC error"); return (ARCHIVE_FATAL); } /* If no CRC error, Go on parsing File Header. */ p = h; endp = p + header_size - 7; memcpy(&file_header, p, sizeof(file_header)); p += sizeof(file_header); rar->compression_method = file_header.method; ttime = archive_le32dec(file_header.file_time); rar->mtime = get_time(ttime); rar->file_crc = archive_le32dec(file_header.file_crc); if (rar->file_flags & FHD_PASSWORD) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "RAR encryption support unavailable."); return (ARCHIVE_FATAL); } if (rar->file_flags & FHD_LARGE) { memcpy(packed_size, file_header.pack_size, 4); memcpy(packed_size + 4, p, 4); /* High pack size */ p += 4; memcpy(unp_size, file_header.unp_size, 4); memcpy(unp_size + 4, p, 4); /* High unpack size */ p += 4; rar->packed_size = archive_le64dec(&packed_size); rar->unp_size = archive_le64dec(&unp_size); } else { rar->packed_size = archive_le32dec(file_header.pack_size); rar->unp_size = archive_le32dec(file_header.unp_size); } /* TODO: Need to use CRC check for these kind of cases. * For now, check if sizes are not < 0. */ if (rar->packed_size < 0 || rar->unp_size < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid sizes specified."); return (ARCHIVE_FATAL); } /* TODO: RARv3 subblocks contain comments. For now the complete block is * consumed at the end. */ if (head_type == NEWSUB_HEAD) { size_t distance = p - (const char *)h; header_size += rar->packed_size; /* Make sure we have the extended data. */ if ((h = __archive_read_ahead(a, header_size - 7, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; endp = p + header_size - 7; p += distance; } filename_size = archive_le16dec(file_header.name_size); if (p + filename_size > endp) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid filename size"); return (ARCHIVE_FATAL); } if (rar->filename_allocated < filename_size+2) { rar->filename = realloc(rar->filename, filename_size+2); if (rar->filename == NULL) { archive_set_error(&a->archive, ENOMEM, "Couldn't allocate memory."); return (ARCHIVE_FATAL); } } filename = rar->filename; memcpy(filename, p, filename_size); filename[filename_size] = '\0'; if (rar->file_flags & FHD_UNICODE) { if (filename_size != strlen(filename)) { unsigned char highbyte, flagbits, flagbyte, length, offset; end = filename_size; filename_size = 0; offset = strlen(filename) + 1; highbyte = *(p + offset++); flagbits = 0; flagbyte = 0; while (offset < end && filename_size < end) { if (!flagbits) { flagbyte = *(p + offset++); flagbits = 8; } flagbits -= 2; switch((flagbyte >> flagbits) & 3) { case 0: filename[filename_size++] = '\0'; filename[filename_size++] = *(p + offset++); break; case 1: filename[filename_size++] = highbyte; filename[filename_size++] = *(p + offset++); break; case 2: filename[filename_size++] = *(p + offset + 1); filename[filename_size++] = *(p + offset); offset += 2; break; case 3: { length = *(p + offset++); while (length) { if (filename_size >= end) break; filename[filename_size++] = *(p + offset); length--; } } break; } } if (filename_size >= end) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid filename"); return (ARCHIVE_FATAL); } filename[filename_size++] = '\0'; filename[filename_size++] = '\0'; /* Decoded unicode form is UTF-16BE, so we have to update a string * conversion object for it. */ if (rar->sconv_utf16be == NULL) { rar->sconv_utf16be = archive_string_conversion_from_charset( &a->archive, "UTF-16BE", 1); if (rar->sconv_utf16be == NULL) return (ARCHIVE_FATAL); } fn_sconv = rar->sconv_utf16be; strp = filename; while (memcmp(strp, "\x00\x00", 2)) { if (!memcmp(strp, "\x00\\", 2)) *(strp + 1) = '/'; strp += 2; } p += offset; } else { /* * If FHD_UNICODE is set but no unicode data, this file name form * is UTF-8, so we have to update a string conversion object for * it accordingly. */ if (rar->sconv_utf8 == NULL) { rar->sconv_utf8 = archive_string_conversion_from_charset( &a->archive, "UTF-8", 1); if (rar->sconv_utf8 == NULL) return (ARCHIVE_FATAL); } fn_sconv = rar->sconv_utf8; while ((strp = strchr(filename, '\\')) != NULL) *strp = '/'; p += filename_size; } } else { fn_sconv = sconv; while ((strp = strchr(filename, '\\')) != NULL) *strp = '/'; p += filename_size; } if (rar->file_flags & FHD_SALT) { if (p + 8 > endp) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } memcpy(rar->salt, p, 8); p += 8; } if (rar->file_flags & FHD_EXTTIME) { if (read_exttime(p, rar, endp) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid header size"); return (ARCHIVE_FATAL); } } __archive_read_consume(a, header_size - 7); switch(file_header.host_os) { case OS_MSDOS: case OS_OS2: case OS_WIN32: rar->mode = archive_le32dec(file_header.file_attr); if (rar->mode & FILE_ATTRIBUTE_DIRECTORY) rar->mode = AE_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH; else rar->mode = AE_IFREG; rar->mode |= S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH; break; case OS_UNIX: case OS_MAC_OS: case OS_BEOS: rar->mode = archive_le32dec(file_header.file_attr); break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Unknown file attributes from RAR file's host OS"); return (ARCHIVE_FATAL); } rar->bytes_remaining = rar->packed_size; rar->bytes_uncopied = rar->bytes_unconsumed = 0; rar->lzss.position = rar->dictionary_size = rar->offset = 0; rar->offset_outgoing = 0; rar->br.cache_avail = 0; rar->br.avail_in = 0; rar->crc_calculated = 0; rar->entry_eof = 0; rar->valid = 1; rar->is_ppmd_block = 0; rar->start_new_table = 1; free(rar->unp_buffer); rar->unp_buffer = NULL; rar->unp_offset = 0; rar->unp_buffer_size = UNP_BUFFER_SIZE; memset(rar->lengthtable, 0, sizeof(rar->lengthtable)); __archive_ppmd7_functions.Ppmd7_Free(&rar->ppmd7_context, &g_szalloc); rar->ppmd_valid = rar->ppmd_eod = 0; /* Don't set any archive entries for non-file header types */ if (head_type == NEWSUB_HEAD) return ret; archive_entry_set_mtime(entry, rar->mtime, rar->mnsec); archive_entry_set_ctime(entry, rar->ctime, rar->cnsec); archive_entry_set_atime(entry, rar->atime, rar->ansec); archive_entry_set_size(entry, rar->unp_size); archive_entry_set_mode(entry, rar->mode); if (archive_entry_copy_pathname_l(entry, filename, filename_size, fn_sconv)) { if (errno == ENOMEM) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for Pathname"); return (ARCHIVE_FATAL); } archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Pathname cannot be converted from %s to current locale.", archive_string_conversion_charset_name(fn_sconv)); ret = (ARCHIVE_WARN); } if (((rar->mode) & AE_IFMT) == AE_IFLNK) { /* Make sure a symbolic-link file does not have its body. */ rar->bytes_remaining = 0; archive_entry_set_size(entry, 0); /* Read a symbolic-link name. */ if ((ret2 = read_symlink_stored(a, entry, sconv)) < (ARCHIVE_WARN)) return ret2; if (ret > ret2) ret = ret2; } if (rar->bytes_remaining == 0) rar->entry_eof = 1; return ret; } static time_t get_time(int ttime) { struct tm tm; tm.tm_sec = 2 * (ttime & 0x1f); tm.tm_min = (ttime >> 5) & 0x3f; tm.tm_hour = (ttime >> 11) & 0x1f; tm.tm_mday = (ttime >> 16) & 0x1f; tm.tm_mon = ((ttime >> 21) & 0x0f) - 1; tm.tm_year = ((ttime >> 25) & 0x7f) + 80; tm.tm_isdst = -1; return mktime(&tm); } static int read_exttime(const char *p, struct rar *rar, const char *endp) { unsigned rmode, flags, rem, j, count; int ttime, i; struct tm *tm; time_t t; long nsec; if (p + 2 > endp) return (-1); flags = archive_le16dec(p); p += 2; for (i = 3; i >= 0; i--) { t = 0; if (i == 3) t = rar->mtime; rmode = flags >> i * 4; if (rmode & 8) { if (!t) { if (p + 4 > endp) return (-1); ttime = archive_le32dec(p); t = get_time(ttime); p += 4; } rem = 0; count = rmode & 3; if (p + count > endp) return (-1); for (j = 0; j < count; j++) { rem = ((*p) << 16) | (rem >> 8); p++; } tm = localtime(&t); nsec = tm->tm_sec + rem / NS_UNIT; if (rmode & 4) { tm->tm_sec++; t = mktime(tm); } if (i == 3) { rar->mtime = t; rar->mnsec = nsec; } else if (i == 2) { rar->ctime = t; rar->cnsec = nsec; } else if (i == 1) { rar->atime = t; rar->ansec = nsec; } else { rar->arctime = t; rar->arcnsec = nsec; } } } return (0); } static int read_symlink_stored(struct archive_read *a, struct archive_entry *entry, struct archive_string_conv *sconv) { const void *h; const char *p; struct rar *rar; int ret = (ARCHIVE_OK); rar = (struct rar *)(a->format->data); if ((h = __archive_read_ahead(a, rar->packed_size, NULL)) == NULL) return (ARCHIVE_FATAL); p = h; if (archive_entry_copy_symlink_l(entry, p, rar->packed_size, sconv)) { if (errno == ENOMEM) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for link"); return (ARCHIVE_FATAL); } archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "link cannot be converted from %s to current locale.", archive_string_conversion_charset_name(sconv)); ret = (ARCHIVE_WARN); } __archive_read_consume(a, rar->packed_size); return ret; } static int read_data_stored(struct archive_read *a, const void **buff, size_t *size, int64_t *offset) { struct rar *rar; ssize_t bytes_avail; rar = (struct rar *)(a->format->data); if (rar->bytes_remaining == 0) { *buff = NULL; *size = 0; *offset = rar->offset; if (rar->file_crc != rar->crc_calculated) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "File CRC error"); return (ARCHIVE_FATAL); } rar->entry_eof = 1; return (ARCHIVE_EOF); } *buff = __archive_read_ahead(a, 1, &bytes_avail); if (bytes_avail <= 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); return (ARCHIVE_FATAL); } if (bytes_avail > rar->bytes_remaining) bytes_avail = rar->bytes_remaining; *size = bytes_avail; *offset = rar->offset; rar->offset += bytes_avail; rar->bytes_remaining -= bytes_avail; rar->bytes_unconsumed = bytes_avail; /* Calculate File CRC. */ rar->crc_calculated = crc32(rar->crc_calculated, *buff, bytes_avail); return (ARCHIVE_OK); } static int read_data_compressed(struct archive_read *a, const void **buff, size_t *size, int64_t *offset) { struct rar *rar; int64_t start, end, actualend; size_t bs; int ret = (ARCHIVE_OK), sym, code, lzss_offset, length, i; rar = (struct rar *)(a->format->data); do { if (!rar->valid) return (ARCHIVE_FATAL); if (rar->ppmd_eod || (rar->dictionary_size && rar->offset >= rar->unp_size)) { if (rar->unp_offset > 0) { /* * We have unprocessed extracted data. write it out. */ *buff = rar->unp_buffer; *size = rar->unp_offset; *offset = rar->offset_outgoing; rar->offset_outgoing += *size; /* Calculate File CRC. */ rar->crc_calculated = crc32(rar->crc_calculated, *buff, *size); rar->unp_offset = 0; return (ARCHIVE_OK); } *buff = NULL; *size = 0; *offset = rar->offset; if (rar->file_crc != rar->crc_calculated) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "File CRC error"); return (ARCHIVE_FATAL); } rar->entry_eof = 1; return (ARCHIVE_EOF); } if (!rar->is_ppmd_block && rar->dictionary_size && rar->bytes_uncopied > 0) { if (rar->bytes_uncopied > (rar->unp_buffer_size - rar->unp_offset)) bs = rar->unp_buffer_size - rar->unp_offset; else bs = rar->bytes_uncopied; ret = copy_from_lzss_window(a, buff, rar->offset, bs); if (ret != ARCHIVE_OK) return (ret); rar->offset += bs; rar->bytes_uncopied -= bs; if (*buff != NULL) { rar->unp_offset = 0; *size = rar->unp_buffer_size; *offset = rar->offset_outgoing; rar->offset_outgoing += *size; /* Calculate File CRC. */ rar->crc_calculated = crc32(rar->crc_calculated, *buff, *size); return (ret); } continue; } if (!rar->br.next_in && (ret = rar_br_preparation(a, &(rar->br))) < ARCHIVE_WARN) return (ret); if (rar->start_new_table && ((ret = parse_codes(a)) < (ARCHIVE_WARN))) return (ret); if (rar->is_ppmd_block) { if ((sym = __archive_ppmd7_functions.Ppmd7_DecodeSymbol( &rar->ppmd7_context, &rar->range_dec.p)) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid symbol"); return (ARCHIVE_FATAL); } if(sym != rar->ppmd_escape) { lzss_emit_literal(rar, sym); rar->bytes_uncopied++; } else { if ((code = __archive_ppmd7_functions.Ppmd7_DecodeSymbol( &rar->ppmd7_context, &rar->range_dec.p)) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid symbol"); return (ARCHIVE_FATAL); } switch(code) { case 0: rar->start_new_table = 1; return read_data_compressed(a, buff, size, offset); case 2: rar->ppmd_eod = 1;/* End Of ppmd Data. */ continue; case 3: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Parsing filters is unsupported."); return (ARCHIVE_FAILED); case 4: lzss_offset = 0; for (i = 2; i >= 0; i--) { if ((code = __archive_ppmd7_functions.Ppmd7_DecodeSymbol( &rar->ppmd7_context, &rar->range_dec.p)) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid symbol"); return (ARCHIVE_FATAL); } lzss_offset |= code << (i * 8); } if ((length = __archive_ppmd7_functions.Ppmd7_DecodeSymbol( &rar->ppmd7_context, &rar->range_dec.p)) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid symbol"); return (ARCHIVE_FATAL); } lzss_emit_match(rar, lzss_offset + 2, length + 32); rar->bytes_uncopied += length + 32; break; case 5: if ((length = __archive_ppmd7_functions.Ppmd7_DecodeSymbol( &rar->ppmd7_context, &rar->range_dec.p)) < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid symbol"); return (ARCHIVE_FATAL); } lzss_emit_match(rar, 1, length + 4); rar->bytes_uncopied += length + 4; break; default: lzss_emit_literal(rar, sym); rar->bytes_uncopied++; } } } else { start = rar->offset; end = start + rar->dictionary_size; rar->filterstart = INT64_MAX; if ((actualend = expand(a, end)) < 0) return ((int)actualend); rar->bytes_uncopied = actualend - start; if (rar->bytes_uncopied == 0) { /* Broken RAR files cause this case. * NOTE: If this case were possible on a normal RAR file * we would find out where it was actually bad and * what we would do to solve it. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Internal error extracting RAR file"); return (ARCHIVE_FATAL); } } if (rar->bytes_uncopied > (rar->unp_buffer_size - rar->unp_offset)) bs = rar->unp_buffer_size - rar->unp_offset; else bs = rar->bytes_uncopied; ret = copy_from_lzss_window(a, buff, rar->offset, bs); if (ret != ARCHIVE_OK) return (ret); rar->offset += bs; rar->bytes_uncopied -= bs; /* * If *buff is NULL, it means unp_buffer is not full. * So we have to continue extracting a RAR file. */ } while (*buff == NULL); rar->unp_offset = 0; *size = rar->unp_buffer_size; *offset = rar->offset_outgoing; rar->offset_outgoing += *size; /* Calculate File CRC. */ rar->crc_calculated = crc32(rar->crc_calculated, *buff, *size); return ret; } static int parse_codes(struct archive_read *a) { int i, j, val, n, r; unsigned char bitlengths[MAX_SYMBOLS], zerocount, ppmd_flags; unsigned int maxorder; struct huffman_code precode; struct rar *rar = (struct rar *)(a->format->data); struct rar_br *br = &(rar->br); free_codes(a); /* Skip to the next byte */ rar_br_consume_unalined_bits(br); /* PPMd block flag */ if (!rar_br_read_ahead(a, br, 1)) goto truncated_data; if ((rar->is_ppmd_block = rar_br_bits(br, 1)) != 0) { rar_br_consume(br, 1); if (!rar_br_read_ahead(a, br, 7)) goto truncated_data; ppmd_flags = rar_br_bits(br, 7); rar_br_consume(br, 7); /* Memory is allocated in MB */ if (ppmd_flags & 0x20) { if (!rar_br_read_ahead(a, br, 8)) goto truncated_data; rar->dictionary_size = (rar_br_bits(br, 8) + 1) << 20; rar_br_consume(br, 8); } if (ppmd_flags & 0x40) { if (!rar_br_read_ahead(a, br, 8)) goto truncated_data; rar->ppmd_escape = rar->ppmd7_context.InitEsc = rar_br_bits(br, 8); rar_br_consume(br, 8); } else rar->ppmd_escape = 2; if (ppmd_flags & 0x20) { maxorder = (ppmd_flags & 0x1F) + 1; if(maxorder > 16) maxorder = 16 + (maxorder - 16) * 3; if (maxorder == 1) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); return (ARCHIVE_FATAL); } /* Make sure ppmd7_contest is freed before Ppmd7_Construct * because reading a broken file cause this abnormal sequence. */ __archive_ppmd7_functions.Ppmd7_Free(&rar->ppmd7_context, &g_szalloc); rar->bytein.a = a; rar->bytein.Read = &ppmd_read; __archive_ppmd7_functions.PpmdRAR_RangeDec_CreateVTable(&rar->range_dec); rar->range_dec.Stream = &rar->bytein; __archive_ppmd7_functions.Ppmd7_Construct(&rar->ppmd7_context); if (!__archive_ppmd7_functions.Ppmd7_Alloc(&rar->ppmd7_context, rar->dictionary_size, &g_szalloc)) { archive_set_error(&a->archive, ENOMEM, "Out of memory"); return (ARCHIVE_FATAL); } if (!__archive_ppmd7_functions.PpmdRAR_RangeDec_Init(&rar->range_dec)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Unable to initialize PPMd range decoder"); return (ARCHIVE_FATAL); } __archive_ppmd7_functions.Ppmd7_Init(&rar->ppmd7_context, maxorder); rar->ppmd_valid = 1; } else { if (!rar->ppmd_valid) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid PPMd sequence"); return (ARCHIVE_FATAL); } if (!__archive_ppmd7_functions.PpmdRAR_RangeDec_Init(&rar->range_dec)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Unable to initialize PPMd range decoder"); return (ARCHIVE_FATAL); } } } else { rar_br_consume(br, 1); /* Keep existing table flag */ if (!rar_br_read_ahead(a, br, 1)) goto truncated_data; if (!rar_br_bits(br, 1)) memset(rar->lengthtable, 0, sizeof(rar->lengthtable)); rar_br_consume(br, 1); memset(&bitlengths, 0, sizeof(bitlengths)); for (i = 0; i < MAX_SYMBOLS;) { if (!rar_br_read_ahead(a, br, 4)) goto truncated_data; bitlengths[i++] = rar_br_bits(br, 4); rar_br_consume(br, 4); if (bitlengths[i-1] == 0xF) { if (!rar_br_read_ahead(a, br, 4)) goto truncated_data; zerocount = rar_br_bits(br, 4); rar_br_consume(br, 4); if (zerocount) { i--; for (j = 0; j < zerocount + 2 && i < MAX_SYMBOLS; j++) bitlengths[i++] = 0; } } } memset(&precode, 0, sizeof(precode)); r = create_code(a, &precode, bitlengths, MAX_SYMBOLS, MAX_SYMBOL_LENGTH); if (r != ARCHIVE_OK) { free(precode.tree); free(precode.table); return (r); } for (i = 0; i < HUFFMAN_TABLE_SIZE;) { if ((val = read_next_symbol(a, &precode)) < 0) { free(precode.tree); free(precode.table); return (ARCHIVE_FATAL); } if (val < 16) { rar->lengthtable[i] = (rar->lengthtable[i] + val) & 0xF; i++; } else if (val < 18) { if (i == 0) { free(precode.tree); free(precode.table); archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Internal error extracting RAR file."); return (ARCHIVE_FATAL); } if(val == 16) { if (!rar_br_read_ahead(a, br, 3)) { free(precode.tree); free(precode.table); goto truncated_data; } n = rar_br_bits(br, 3) + 3; rar_br_consume(br, 3); } else { if (!rar_br_read_ahead(a, br, 7)) { free(precode.tree); free(precode.table); goto truncated_data; } n = rar_br_bits(br, 7) + 11; rar_br_consume(br, 7); } for (j = 0; j < n && i < HUFFMAN_TABLE_SIZE; j++) { rar->lengthtable[i] = rar->lengthtable[i-1]; i++; } } else { if(val == 18) { if (!rar_br_read_ahead(a, br, 3)) { free(precode.tree); free(precode.table); goto truncated_data; } n = rar_br_bits(br, 3) + 3; rar_br_consume(br, 3); } else { if (!rar_br_read_ahead(a, br, 7)) { free(precode.tree); free(precode.table); goto truncated_data; } n = rar_br_bits(br, 7) + 11; rar_br_consume(br, 7); } for(j = 0; j < n && i < HUFFMAN_TABLE_SIZE; j++) rar->lengthtable[i++] = 0; } } free(precode.tree); free(precode.table); r = create_code(a, &rar->maincode, &rar->lengthtable[0], MAINCODE_SIZE, MAX_SYMBOL_LENGTH); if (r != ARCHIVE_OK) return (r); r = create_code(a, &rar->offsetcode, &rar->lengthtable[MAINCODE_SIZE], OFFSETCODE_SIZE, MAX_SYMBOL_LENGTH); if (r != ARCHIVE_OK) return (r); r = create_code(a, &rar->lowoffsetcode, &rar->lengthtable[MAINCODE_SIZE + OFFSETCODE_SIZE], LOWOFFSETCODE_SIZE, MAX_SYMBOL_LENGTH); if (r != ARCHIVE_OK) return (r); r = create_code(a, &rar->lengthcode, &rar->lengthtable[MAINCODE_SIZE + OFFSETCODE_SIZE + LOWOFFSETCODE_SIZE], LENGTHCODE_SIZE, MAX_SYMBOL_LENGTH); if (r != ARCHIVE_OK) return (r); } if (!rar->dictionary_size || !rar->lzss.window) { /* Seems as though dictionary sizes are not used. Even so, minimize * memory usage as much as possible. */ if (rar->unp_size >= DICTIONARY_MAX_SIZE) rar->dictionary_size = DICTIONARY_MAX_SIZE; else rar->dictionary_size = rar_fls(rar->unp_size) << 1; rar->lzss.window = (unsigned char *)realloc(rar->lzss.window, rar->dictionary_size); if (rar->lzss.window == NULL) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for uncompressed data."); return (ARCHIVE_FATAL); } memset(rar->lzss.window, 0, rar->dictionary_size); rar->lzss.mask = rar->dictionary_size - 1; } rar->start_new_table = 0; return (ARCHIVE_OK); truncated_data: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); rar->valid = 0; return (ARCHIVE_FATAL); } static void free_codes(struct archive_read *a) { struct rar *rar = (struct rar *)(a->format->data); free(rar->maincode.tree); free(rar->offsetcode.tree); free(rar->lowoffsetcode.tree); free(rar->lengthcode.tree); free(rar->maincode.table); free(rar->offsetcode.table); free(rar->lowoffsetcode.table); free(rar->lengthcode.table); memset(&rar->maincode, 0, sizeof(rar->maincode)); memset(&rar->offsetcode, 0, sizeof(rar->offsetcode)); memset(&rar->lowoffsetcode, 0, sizeof(rar->lowoffsetcode)); memset(&rar->lengthcode, 0, sizeof(rar->lengthcode)); } static int read_next_symbol(struct archive_read *a, struct huffman_code *code) { unsigned char bit; unsigned int bits; int length, value, node; struct rar *rar; struct rar_br *br; if (!code->table) { if (make_table(a, code) != (ARCHIVE_OK)) return -1; } rar = (struct rar *)(a->format->data); br = &(rar->br); /* Look ahead (peek) at bits */ if (!rar_br_read_ahead(a, br, code->tablesize)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); rar->valid = 0; return -1; } bits = rar_br_bits(br, code->tablesize); length = code->table[bits].length; value = code->table[bits].value; if (length < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid prefix code in bitstream"); return -1; } if (length <= code->tablesize) { /* Skip length bits */ rar_br_consume(br, length); return value; } /* Skip tablesize bits */ rar_br_consume(br, code->tablesize); node = value; while (!(code->tree[node].branches[0] == code->tree[node].branches[1])) { if (!rar_br_read_ahead(a, br, 1)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); rar->valid = 0; return -1; } bit = rar_br_bits(br, 1); rar_br_consume(br, 1); if (code->tree[node].branches[bit] < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid prefix code in bitstream"); return -1; } node = code->tree[node].branches[bit]; } return code->tree[node].branches[0]; } static int create_code(struct archive_read *a, struct huffman_code *code, unsigned char *lengths, int numsymbols, char maxlength) { int i, j, codebits = 0, symbolsleft = numsymbols; if (new_node(code) < 0) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for node data."); return (ARCHIVE_FATAL); } code->numentries = 1; code->minlength = INT_MAX; code->maxlength = INT_MIN; codebits = 0; for(i = 1; i <= maxlength; i++) { for(j = 0; j < numsymbols; j++) { if (lengths[j] != i) continue; if (add_value(a, code, j, codebits, i) != ARCHIVE_OK) return (ARCHIVE_FATAL); codebits++; if (--symbolsleft <= 0) { break; break; } } codebits <<= 1; } return (ARCHIVE_OK); } static int add_value(struct archive_read *a, struct huffman_code *code, int value, int codebits, int length) { int repeatpos, lastnode, bitpos, bit, repeatnode, nextnode; free(code->table); code->table = NULL; if(length > code->maxlength) code->maxlength = length; if(length < code->minlength) code->minlength = length; repeatpos = -1; if (repeatpos == 0 || (repeatpos >= 0 && (((codebits >> (repeatpos - 1)) & 3) == 0 || ((codebits >> (repeatpos - 1)) & 3) == 3))) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid repeat position"); return (ARCHIVE_FATAL); } lastnode = 0; for (bitpos = length - 1; bitpos >= 0; bitpos--) { bit = (codebits >> bitpos) & 1; /* Leaf node check */ if (code->tree[lastnode].branches[0] == code->tree[lastnode].branches[1]) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Prefix found"); return (ARCHIVE_FATAL); } if (bitpos == repeatpos) { /* Open branch check */ if (!(code->tree[lastnode].branches[bit] < 0)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid repeating code"); return (ARCHIVE_FATAL); } if ((repeatnode = new_node(code)) < 0) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for node data."); return (ARCHIVE_FATAL); } if ((nextnode = new_node(code)) < 0) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for node data."); return (ARCHIVE_FATAL); } /* Set branches */ code->tree[lastnode].branches[bit] = repeatnode; code->tree[repeatnode].branches[bit] = repeatnode; code->tree[repeatnode].branches[bit^1] = nextnode; lastnode = nextnode; bitpos++; /* terminating bit already handled, skip it */ } else { /* Open branch check */ if (code->tree[lastnode].branches[bit] < 0) { if (new_node(code) < 0) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for node data."); return (ARCHIVE_FATAL); } code->tree[lastnode].branches[bit] = code->numentries++; } /* set to branch */ lastnode = code->tree[lastnode].branches[bit]; } } if (!(code->tree[lastnode].branches[0] == -1 && code->tree[lastnode].branches[1] == -2)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Prefix found"); return (ARCHIVE_FATAL); } /* Set leaf value */ code->tree[lastnode].branches[0] = value; code->tree[lastnode].branches[1] = value; return (ARCHIVE_OK); } static int new_node(struct huffman_code *code) { code->tree = (struct huffman_tree_node *)realloc(code->tree, (code->numentries + 1) * sizeof(*code->tree)); if (code->tree == NULL) return (-1); code->tree[code->numentries].branches[0] = -1; code->tree[code->numentries].branches[1] = -2; return 1; } static int make_table(struct archive_read *a, struct huffman_code *code) { if (code->maxlength < code->minlength || code->maxlength > 10) code->tablesize = 10; else code->tablesize = code->maxlength; code->table = (struct huffman_table_entry *)malloc(sizeof(*code->table) * (1 << code->tablesize)); return make_table_recurse(a, code, 0, code->table, 0, code->tablesize); } static int make_table_recurse(struct archive_read *a, struct huffman_code *code, int node, struct huffman_table_entry *table, int depth, int maxdepth) { int currtablesize, i, ret = (ARCHIVE_OK); if (!code->tree) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Huffman tree was not created."); return (ARCHIVE_FATAL); } if (node < 0 || node >= code->numentries) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid location to Huffman tree specified."); return (ARCHIVE_FATAL); } currtablesize = 1 << (maxdepth - depth); if (code->tree[node].branches[0] == code->tree[node].branches[1]) { for(i = 0; i < currtablesize; i++) { table[i].length = depth; table[i].value = code->tree[node].branches[0]; } } else if (node < 0) { for(i = 0; i < currtablesize; i++) table[i].length = -1; } else { if(depth == maxdepth) { table[0].length = maxdepth + 1; table[0].value = node; } else { ret |= make_table_recurse(a, code, code->tree[node].branches[0], table, depth + 1, maxdepth); ret |= make_table_recurse(a, code, code->tree[node].branches[1], table + currtablesize / 2, depth + 1, maxdepth); } } return ret; } static int64_t expand(struct archive_read *a, int64_t end) { static const unsigned char lengthbases[] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224 }; static const unsigned char lengthbits[] = { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5 }; static const unsigned int offsetbases[] = { 0, 1, 2, 3, 4, 6, 8, 12, 16, 24, 32, 48, 64, 96, 128, 192, 256, 384, 512, 768, 1024, 1536, 2048, 3072, 4096, 6144, 8192, 12288, 16384, 24576, 32768, 49152, 65536, 98304, 131072, 196608, 262144, 327680, 393216, 458752, 524288, 589824, 655360, 720896, 786432, 851968, 917504, 983040, 1048576, 1310720, 1572864, 1835008, 2097152, 2359296, 2621440, 2883584, 3145728, 3407872, 3670016, 3932160 }; static const unsigned char offsetbits[] = { 0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18 }; static const unsigned char shortbases[] = { 0, 4, 8, 16, 32, 64, 128, 192 }; static const unsigned char shortbits[] = { 2, 2, 3, 4, 5, 6, 6, 6 }; int symbol, offs, len, offsindex, lensymbol, i, offssymbol, lowoffsetsymbol; unsigned char newfile; struct rar *rar = (struct rar *)(a->format->data); struct rar_br *br = &(rar->br); if (rar->filterstart < end) end = rar->filterstart; while (1) { if (rar->output_last_match && lzss_position(&rar->lzss) + rar->lastlength <= end) { lzss_emit_match(rar, rar->lastoffset, rar->lastlength); rar->output_last_match = 0; } if(rar->is_ppmd_block || rar->output_last_match || lzss_position(&rar->lzss) >= end) return lzss_position(&rar->lzss); if ((symbol = read_next_symbol(a, &rar->maincode)) < 0) return (ARCHIVE_FATAL); rar->output_last_match = 0; if (symbol < 256) { lzss_emit_literal(rar, symbol); continue; } else if (symbol == 256) { if (!rar_br_read_ahead(a, br, 1)) goto truncated_data; newfile = !rar_br_bits(br, 1); rar_br_consume(br, 1); if(newfile) { rar->start_new_block = 1; if (!rar_br_read_ahead(a, br, 1)) goto truncated_data; rar->start_new_table = rar_br_bits(br, 1); rar_br_consume(br, 1); return lzss_position(&rar->lzss); } else { if (parse_codes(a) != ARCHIVE_OK) return (ARCHIVE_FATAL); continue; } } else if(symbol==257) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Parsing filters is unsupported."); return (ARCHIVE_FAILED); } else if(symbol==258) { if(rar->lastlength == 0) continue; offs = rar->lastoffset; len = rar->lastlength; } else if (symbol <= 262) { offsindex = symbol - 259; offs = rar->oldoffset[offsindex]; if ((lensymbol = read_next_symbol(a, &rar->lengthcode)) < 0) goto bad_data; if (lensymbol > (int)(sizeof(lengthbases)/sizeof(lengthbases[0]))) goto bad_data; if (lensymbol > (int)(sizeof(lengthbits)/sizeof(lengthbits[0]))) goto bad_data; len = lengthbases[lensymbol] + 2; if (lengthbits[lensymbol] > 0) { if (!rar_br_read_ahead(a, br, lengthbits[lensymbol])) goto truncated_data; len += rar_br_bits(br, lengthbits[lensymbol]); rar_br_consume(br, lengthbits[lensymbol]); } for (i = offsindex; i > 0; i--) rar->oldoffset[i] = rar->oldoffset[i-1]; rar->oldoffset[0] = offs; } else if(symbol<=270) { offs = shortbases[symbol-263] + 1; if(shortbits[symbol-263] > 0) { if (!rar_br_read_ahead(a, br, shortbits[symbol-263])) goto truncated_data; offs += rar_br_bits(br, shortbits[symbol-263]); rar_br_consume(br, shortbits[symbol-263]); } len = 2; for(i = 3; i > 0; i--) rar->oldoffset[i] = rar->oldoffset[i-1]; rar->oldoffset[0] = offs; } else { if (symbol-271 > (int)(sizeof(lengthbases)/sizeof(lengthbases[0]))) goto bad_data; if (symbol-271 > (int)(sizeof(lengthbits)/sizeof(lengthbits[0]))) goto bad_data; len = lengthbases[symbol-271]+3; if(lengthbits[symbol-271] > 0) { if (!rar_br_read_ahead(a, br, lengthbits[symbol-271])) goto truncated_data; len += rar_br_bits(br, lengthbits[symbol-271]); rar_br_consume(br, lengthbits[symbol-271]); } if ((offssymbol = read_next_symbol(a, &rar->offsetcode)) < 0) goto bad_data; if (offssymbol > (int)(sizeof(offsetbases)/sizeof(offsetbases[0]))) goto bad_data; if (offssymbol > (int)(sizeof(offsetbits)/sizeof(offsetbits[0]))) goto bad_data; offs = offsetbases[offssymbol]+1; if(offsetbits[offssymbol] > 0) { if(offssymbol > 9) { if(offsetbits[offssymbol] > 4) { if (!rar_br_read_ahead(a, br, offsetbits[offssymbol] - 4)) goto truncated_data; offs += rar_br_bits(br, offsetbits[offssymbol] - 4) << 4; rar_br_consume(br, offsetbits[offssymbol] - 4); } if(rar->numlowoffsetrepeats > 0) { rar->numlowoffsetrepeats--; offs += rar->lastlowoffset; } else { if ((lowoffsetsymbol = read_next_symbol(a, &rar->lowoffsetcode)) < 0) return (ARCHIVE_FATAL); if(lowoffsetsymbol == 16) { rar->numlowoffsetrepeats = 15; offs += rar->lastlowoffset; } else { offs += lowoffsetsymbol; rar->lastlowoffset = lowoffsetsymbol; } } } else { if (!rar_br_read_ahead(a, br, offsetbits[offssymbol])) goto truncated_data; offs += rar_br_bits(br, offsetbits[offssymbol]); rar_br_consume(br, offsetbits[offssymbol]); } } if (offs >= 0x40000) len++; if (offs >= 0x2000) len++; for(i = 3; i > 0; i--) rar->oldoffset[i] = rar->oldoffset[i-1]; rar->oldoffset[0] = offs; } rar->lastoffset = offs; rar->lastlength = len; rar->output_last_match = 1; } truncated_data: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated RAR file data"); rar->valid = 0; return (ARCHIVE_FATAL); bad_data: archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Bad RAR file data"); return (ARCHIVE_FATAL); } static int copy_from_lzss_window(struct archive_read *a, const void **buffer, int64_t startpos, int length) { int windowoffs, firstpart; struct rar *rar = (struct rar *)(a->format->data); if (!rar->unp_buffer) { if ((rar->unp_buffer = malloc(rar->unp_buffer_size)) == NULL) { archive_set_error(&a->archive, ENOMEM, "Unable to allocate memory for uncompressed data."); return (ARCHIVE_FATAL); } } windowoffs = lzss_offset_for_position(&rar->lzss, startpos); if(windowoffs + length <= lzss_size(&rar->lzss)) memcpy(&rar->unp_buffer[rar->unp_offset], &rar->lzss.window[windowoffs], length); else { firstpart = lzss_size(&rar->lzss) - windowoffs; if (firstpart < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Bad RAR file data"); return (ARCHIVE_FATAL); } if (firstpart < length) { memcpy(&rar->unp_buffer[rar->unp_offset], &rar->lzss.window[windowoffs], firstpart); memcpy(&rar->unp_buffer[rar->unp_offset + firstpart], &rar->lzss.window[0], length - firstpart); } else memcpy(&rar->unp_buffer[rar->unp_offset], &rar->lzss.window[windowoffs], length); } rar->unp_offset += length; if (rar->unp_offset >= rar->unp_buffer_size) *buffer = rar->unp_buffer; else *buffer = NULL; return (ARCHIVE_OK); } Index: vendor/libarchive/dist/libarchive/archive_read_support_format_zip.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_read_support_format_zip.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_read_support_format_zip.c (revision 231297) @@ -1,1287 +1,1287 @@ /*- * Copyright (c) 2004 Tim Kientzle * Copyright (c) 2011 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD: head/lib/libarchive/archive_read_support_format_zip.c 201102 2009-12-28 03:11:36Z kientzle $"); #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_ZLIB_H #include #endif #include "archive.h" #include "archive_entry.h" #include "archive_entry_locale.h" #include "archive_private.h" #include "archive_read_private.h" #include "archive_endian.h" #ifndef HAVE_ZLIB_H #include "archive_crc32.h" #endif struct zip_entry { int64_t local_header_offset; int64_t compressed_size; int64_t uncompressed_size; int64_t gid; int64_t uid; struct archive_entry *entry; time_t mtime; time_t atime; time_t ctime; uint32_t crc32; uint16_t mode; uint16_t flags; char compression; char system; }; struct zip { /* Structural information about the archive. */ int64_t central_directory_offset; size_t central_directory_size; size_t central_directory_entries; char have_central_directory; /* List of entries (seekable Zip only) */ size_t entries_remaining; struct zip_entry *zip_entries; struct zip_entry *entry; size_t unconsumed; /* entry_bytes_remaining is the number of bytes we expect. */ int64_t entry_bytes_remaining; /* These count the number of bytes actually read for the entry. */ int64_t entry_compressed_bytes_read; int64_t entry_uncompressed_bytes_read; /* Running CRC32 of the decompressed data */ unsigned long entry_crc32; /* Flags to mark progress of decompression. */ char decompress_init; char end_of_entry; ssize_t filename_length; ssize_t extra_length; unsigned char *uncompressed_buffer; size_t uncompressed_buffer_size; #ifdef HAVE_ZLIB_H z_stream stream; char stream_valid; #endif struct archive_string extra; struct archive_string_conv *sconv; struct archive_string_conv *sconv_default; struct archive_string_conv *sconv_utf8; int init_default_conversion; char format_name[64]; }; #define ZIP_LENGTH_AT_END 8 #define ZIP_ENCRYPTED (1<<0) #define ZIP_STRONG_ENCRYPTED (1<<6) #define ZIP_UTF8_NAME (1<<11) static int archive_read_format_zip_streamable_bid(struct archive_read *, int); static int archive_read_format_zip_seekable_bid(struct archive_read *, int); static int archive_read_format_zip_options(struct archive_read *, const char *, const char *); static int archive_read_format_zip_cleanup(struct archive_read *); static int archive_read_format_zip_read_data(struct archive_read *, const void **, size_t *, int64_t *); static int archive_read_format_zip_read_data_skip(struct archive_read *a); static int archive_read_format_zip_seekable_read_header(struct archive_read *, struct archive_entry *); static int archive_read_format_zip_streamable_read_header(struct archive_read *, struct archive_entry *); #ifdef HAVE_ZLIB_H static int zip_read_data_deflate(struct archive_read *a, const void **buff, size_t *size, int64_t *offset); #endif static int zip_read_data_none(struct archive_read *a, const void **buff, size_t *size, int64_t *offset); static int zip_read_local_file_header(struct archive_read *a, struct archive_entry *entry, struct zip *); static time_t zip_time(const char *); static const char *compression_name(int compression); static void process_extra(const char *, size_t, struct zip_entry *); int archive_read_support_format_zip_streamable(struct archive *); int archive_read_support_format_zip_seekable(struct archive *); int archive_read_support_format_zip_streamable(struct archive *_a) { struct archive_read *a = (struct archive_read *)_a; struct zip *zip; int r; archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, "archive_read_support_format_zip"); zip = (struct zip *)malloc(sizeof(*zip)); if (zip == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate zip data"); return (ARCHIVE_FATAL); } memset(zip, 0, sizeof(*zip)); r = __archive_read_register_format(a, zip, "zip", archive_read_format_zip_streamable_bid, archive_read_format_zip_options, archive_read_format_zip_streamable_read_header, archive_read_format_zip_read_data, archive_read_format_zip_read_data_skip, archive_read_format_zip_cleanup); if (r != ARCHIVE_OK) free(zip); return (ARCHIVE_OK); } int archive_read_support_format_zip_seekable(struct archive *_a) { struct archive_read *a = (struct archive_read *)_a; struct zip *zip; int r; archive_check_magic(_a, ARCHIVE_READ_MAGIC, ARCHIVE_STATE_NEW, "archive_read_support_format_zip_seekable"); zip = (struct zip *)malloc(sizeof(*zip)); if (zip == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate zip data"); return (ARCHIVE_FATAL); } memset(zip, 0, sizeof(*zip)); r = __archive_read_register_format(a, zip, "zip", archive_read_format_zip_seekable_bid, archive_read_format_zip_options, archive_read_format_zip_seekable_read_header, archive_read_format_zip_read_data, archive_read_format_zip_read_data_skip, archive_read_format_zip_cleanup); if (r != ARCHIVE_OK) free(zip); return (ARCHIVE_OK); } int archive_read_support_format_zip(struct archive *a) { int r; r = archive_read_support_format_zip_streamable(a); if (r != ARCHIVE_OK) return r; return (archive_read_support_format_zip_seekable(a)); } /* * TODO: This is a performance sink because it forces the read core to * drop buffered data from the start of file, which will then have to * be re-read again if this bidder loses. * * We workaround this a little by passing in the best bid so far so * that later bidders can do nothing if they know they'll never * outbid. But we can certainly do better... */ static int archive_read_format_zip_seekable_bid(struct archive_read *a, int best_bid) { struct zip *zip = (struct zip *)a->format->data; int64_t filesize; const char *p; /* If someone has already bid more than 32, then avoid trashing the look-ahead buffers with a seek. */ if (best_bid > 32) return (-1); filesize = __archive_read_seek(a, -22, SEEK_END); /* If we can't seek, then we can't bid. */ if (filesize <= 0) return 0; /* TODO: More robust search for end of central directory record. */ if ((p = __archive_read_ahead(a, 22, NULL)) == NULL) return 0; /* First four bytes are signature for end of central directory record. Four zero bytes ensure this isn't a multi-volume Zip file (which we don't yet support). */ if (memcmp(p, "PK\005\006\000\000\000\000", 8) != 0) return 0; /* Since we've already done the hard work of finding the end of central directory record, let's save the important information. */ zip->central_directory_entries = archive_le16dec(p + 10); zip->central_directory_size = archive_le32dec(p + 12); zip->central_directory_offset = archive_le32dec(p + 16); /* Just one volume, so central dir must all be on this volume. */ if (zip->central_directory_entries != archive_le16dec(p + 8)) return 0; /* Central directory can't extend beyond end of this file. */ - if (zip->central_directory_offset + zip->central_directory_size > filesize) + if (zip->central_directory_offset + (int64_t)zip->central_directory_size > filesize) return 0; /* This is just a tiny bit higher than the maximum returned by the streaming Zip bidder. This ensures that the more accurate seeking Zip parser wins whenever seek is available. */ return 32; } static int slurp_central_directory(struct archive_read *a, struct zip *zip) { unsigned i; __archive_read_seek(a, zip->central_directory_offset, SEEK_SET); zip->zip_entries = calloc(zip->central_directory_entries, sizeof(struct zip_entry)); for (i = 0; i < zip->central_directory_entries; ++i) { struct zip_entry *zip_entry = &zip->zip_entries[i]; size_t filename_length, extra_length, comment_length; uint32_t external_attributes; const char *p; if ((p = __archive_read_ahead(a, 46, NULL)) == NULL) return ARCHIVE_FATAL; if (memcmp(p, "PK\001\002", 4) != 0) { archive_set_error(&a->archive, -1, "Invalid central directory signature"); return ARCHIVE_FATAL; } zip->have_central_directory = 1; /* version = p[4]; */ zip_entry->system = p[5]; /* version_required = archive_le16dec(p + 6); */ zip_entry->flags = archive_le16dec(p + 8); zip_entry->compression = archive_le16dec(p + 10); zip_entry->mtime = zip_time(p + 12); zip_entry->crc32 = archive_le32dec(p + 16); zip_entry->compressed_size = archive_le32dec(p + 20); zip_entry->uncompressed_size = archive_le32dec(p + 24); filename_length = archive_le16dec(p + 28); extra_length = archive_le16dec(p + 30); comment_length = archive_le16dec(p + 32); /* disk_start = archive_le16dec(p + 34); */ /* Better be zero. */ /* internal_attributes = archive_le16dec(p + 36); */ /* text bit */ external_attributes = archive_le32dec(p + 38); zip_entry->local_header_offset = archive_le32dec(p + 42); /* If we can't guess the mode, leave it zero here; when we read the local file header we might get more information. */ zip_entry->mode = 0; if (zip_entry->system == 3) { zip_entry->mode = external_attributes >> 16; } /* We don't read the filename until we get to the local file header. Reading it here would speed up table-of-contents operations (removing the need to find and read local file header to get the filename) at the cost of requiring a lot of extra space. */ /* We don't read the extra block here. We assume it will be duplicated at the local file header. */ __archive_read_consume(a, 46 + filename_length + extra_length + comment_length); } /* TODO: Sort zip entries by file offset so that we can optimize get_next_header() to use skip instead of seek. */ return ARCHIVE_OK; } static int archive_read_format_zip_seekable_read_header(struct archive_read *a, struct archive_entry *entry) { struct zip *zip = (struct zip *)a->format->data; int r; a->archive.archive_format = ARCHIVE_FORMAT_ZIP; if (a->archive.archive_format_name == NULL) a->archive.archive_format_name = "ZIP"; if (zip->zip_entries == NULL) { r = slurp_central_directory(a, zip); zip->entries_remaining = zip->central_directory_entries; if (r != ARCHIVE_OK) return r; zip->entry = zip->zip_entries; } else { ++zip->entry; } if (zip->entries_remaining <= 0) return ARCHIVE_EOF; --zip->entries_remaining; /* TODO: If entries are sorted by offset within the file, we should be able to skip here instead of seeking. Skipping is typically faster (easier for I/O layer to optimize). */ __archive_read_seek(a, zip->entry->local_header_offset, SEEK_SET); zip->unconsumed = 0; r = zip_read_local_file_header(a, entry, zip); if (r != ARCHIVE_OK) return r; if ((zip->entry->mode & AE_IFMT) == AE_IFLNK) { const void *p; size_t linkname_length = archive_entry_size(entry); archive_entry_set_size(entry, 0); p = __archive_read_ahead(a, linkname_length, NULL); if (p == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Truncated Zip file"); return ARCHIVE_FATAL; } if (archive_entry_copy_symlink_l(entry, p, linkname_length, NULL) != 0) { /* NOTE: If the last argument is NULL, this will * fail only by memeory allocation failure. */ archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for Symlink"); return (ARCHIVE_FATAL); } /* TODO: handle character-set issues? */ } return ARCHIVE_OK; } static int archive_read_format_zip_streamable_bid(struct archive_read *a, int best_bid) { const char *p; (void)best_bid; /* UNUSED */ if ((p = __archive_read_ahead(a, 4, NULL)) == NULL) return (-1); /* * Bid of 30 here is: 16 bits for "PK", * next 16-bit field has four options (-2 bits). * 16 + 16-2 = 30. */ if (p[0] == 'P' && p[1] == 'K') { if ((p[2] == '\001' && p[3] == '\002') || (p[2] == '\003' && p[3] == '\004') || (p[2] == '\005' && p[3] == '\006') || (p[2] == '\007' && p[3] == '\010') || (p[2] == '0' && p[3] == '0')) return (30); } /* TODO: It's worth looking ahead a little bit for a valid * PK signature. In particular, that would make it possible * to read some UUEncoded SFX files or SFX files coming from * a network socket. */ return (0); } static int archive_read_format_zip_options(struct archive_read *a, const char *key, const char *val) { struct zip *zip; int ret = ARCHIVE_FAILED; zip = (struct zip *)(a->format->data); if (strcmp(key, "compat-2x") == 0) { /* Handle filnames as libarchive 2.x */ zip->init_default_conversion = (val != NULL) ? 1 : 0; ret = ARCHIVE_OK; } else if (strcmp(key, "hdrcharset") == 0) { if (val == NULL || val[0] == 0) archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "zip: hdrcharset option needs a character-set name"); else { zip->sconv = archive_string_conversion_from_charset( &a->archive, val, 0); if (zip->sconv != NULL) { if (strcmp(val, "UTF-8") == 0) zip->sconv_utf8 = zip->sconv; ret = ARCHIVE_OK; } else ret = ARCHIVE_FATAL; } } else archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "zip: unknown keyword ``%s''", key); return (ret); } static int archive_read_format_zip_streamable_read_header(struct archive_read *a, struct archive_entry *entry) { struct zip *zip; a->archive.archive_format = ARCHIVE_FORMAT_ZIP; if (a->archive.archive_format_name == NULL) a->archive.archive_format_name = "ZIP"; zip = (struct zip *)(a->format->data); /* Make sure we have a zip_entry structure to use. */ if (zip->zip_entries == NULL) { zip->zip_entries = malloc(sizeof(struct zip_entry)); if (zip->zip_entries == NULL) { archive_set_error(&a->archive, ENOMEM, "Out of memory"); return ARCHIVE_FATAL; } } zip->entry = zip->zip_entries; memset(zip->entry, 0, sizeof(struct zip_entry)); /* Search ahead for the next local file header. */ __archive_read_consume(a, zip->unconsumed); zip->unconsumed = 0; for (;;) { int64_t skipped = 0; const char *p, *end; ssize_t bytes; p = __archive_read_ahead(a, 4, &bytes); if (p == NULL) return (ARCHIVE_FATAL); end = p + bytes; while (p + 4 <= end) { if (p[0] == 'P' && p[1] == 'K') { if (p[2] == '\001' && p[3] == '\002') /* Beginning of central directory. */ return (ARCHIVE_EOF); if (p[2] == '\003' && p[3] == '\004') { /* Regular file entry. */ __archive_read_consume(a, skipped); return zip_read_local_file_header(a, entry, zip); } if (p[2] == '\005' && p[3] == '\006') /* End of central directory. */ return (ARCHIVE_EOF); } ++p; ++skipped; } __archive_read_consume(a, skipped); } } /* * Assumes file pointer is at beginning of local file header. */ static int zip_read_local_file_header(struct archive_read *a, struct archive_entry *entry, struct zip *zip) { const char *p; const void *h; const wchar_t *wp; const char *cp; size_t len, filename_length, extra_length; struct archive_string_conv *sconv; struct zip_entry *zip_entry = zip->entry; uint32_t local_crc32; int64_t compressed_size, uncompressed_size; int ret = ARCHIVE_OK; char version; zip->decompress_init = 0; zip->end_of_entry = 0; zip->entry_uncompressed_bytes_read = 0; zip->entry_compressed_bytes_read = 0; zip->entry_crc32 = crc32(0, NULL, 0); /* Setup default conversion. */ if (zip->sconv == NULL && !zip->init_default_conversion) { zip->sconv_default = archive_string_default_conversion_for_read(&(a->archive)); zip->init_default_conversion = 1; } if ((p = __archive_read_ahead(a, 30, NULL)) == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file header"); return (ARCHIVE_FATAL); } if (memcmp(p, "PK\003\004", 4) != 0) { archive_set_error(&a->archive, -1, "Damaged Zip archive"); return ARCHIVE_FATAL; } version = p[4]; zip_entry->system = p[5]; zip_entry->flags = archive_le16dec(p + 6); zip_entry->compression = archive_le16dec(p + 8); zip_entry->mtime = zip_time(p + 10); local_crc32 = archive_le32dec(p + 14); compressed_size = archive_le32dec(p + 18); uncompressed_size = archive_le32dec(p + 22); filename_length = archive_le16dec(p + 26); extra_length = archive_le16dec(p + 28); __archive_read_consume(a, 30); if (zip->have_central_directory) { /* If we read the central dir entry, we must have size information as well, so ignore the length-at-end flag. */ zip_entry->flags &= ~ZIP_LENGTH_AT_END; /* If we have values from both the local file header and the central directory, warn about mismatches which might indicate a damaged file. But some writers always put zero in the local header; don't bother warning about that. */ if (local_crc32 != 0 && local_crc32 != zip_entry->crc32) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Inconsistent CRC32 values"); ret = ARCHIVE_WARN; } if (compressed_size != 0 && compressed_size != zip_entry->compressed_size) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Inconsistent compressed size"); ret = ARCHIVE_WARN; } if (uncompressed_size != 0 && uncompressed_size != zip_entry->uncompressed_size) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Inconsistent uncompressed size"); ret = ARCHIVE_WARN; } } else { /* If we don't have the CD info, use whatever we do have. */ zip_entry->crc32 = local_crc32; zip_entry->compressed_size = compressed_size; zip_entry->uncompressed_size = uncompressed_size; } /* Read the filename. */ if ((h = __archive_read_ahead(a, filename_length, NULL)) == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file header"); return (ARCHIVE_FATAL); } if (zip_entry->flags & ZIP_UTF8_NAME) { /* The filename is stored to be UTF-8. */ if (zip->sconv_utf8 == NULL) { zip->sconv_utf8 = archive_string_conversion_from_charset( &a->archive, "UTF-8", 1); if (zip->sconv_utf8 == NULL) return (ARCHIVE_FATAL); } sconv = zip->sconv_utf8; } else if (zip->sconv != NULL) sconv = zip->sconv; else sconv = zip->sconv_default; if (archive_entry_copy_pathname_l(entry, h, filename_length, sconv) != 0) { if (errno == ENOMEM) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for Pathname"); return (ARCHIVE_FATAL); } archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Pathname cannot be converted " "from %s to current locale.", archive_string_conversion_charset_name(sconv)); ret = ARCHIVE_WARN; } __archive_read_consume(a, filename_length); if (zip_entry->mode == 0) { /* Especially in streaming mode, we can end up here without having seen any mode information. Guess from the filename. */ wp = archive_entry_pathname_w(entry); if (wp != NULL) { len = wcslen(wp); if (len > 0 && wp[len - 1] == L'/') zip_entry->mode = AE_IFDIR | 0777; else zip_entry->mode = AE_IFREG | 0777; } else { cp = archive_entry_pathname(entry); len = (cp != NULL)?strlen(cp):0; if (len > 0 && cp[len - 1] == '/') zip_entry->mode = AE_IFDIR | 0777; else zip_entry->mode = AE_IFREG | 0777; } } /* Read the extra data. */ if ((h = __archive_read_ahead(a, extra_length, NULL)) == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file header"); return (ARCHIVE_FATAL); } process_extra(h, extra_length, zip_entry); __archive_read_consume(a, extra_length); /* Populate some additional entry fields: */ archive_entry_set_mode(entry, zip_entry->mode); archive_entry_set_uid(entry, zip_entry->uid); archive_entry_set_gid(entry, zip_entry->gid); archive_entry_set_mtime(entry, zip_entry->mtime, 0); archive_entry_set_ctime(entry, zip_entry->ctime, 0); archive_entry_set_atime(entry, zip_entry->atime, 0); /* Set the size only if it's meaningful. */ if (0 == (zip_entry->flags & ZIP_LENGTH_AT_END)) archive_entry_set_size(entry, zip_entry->uncompressed_size); zip->entry_bytes_remaining = zip_entry->compressed_size; /* If there's no body, force read_data() to return EOF immediately. */ if (0 == (zip_entry->flags & ZIP_LENGTH_AT_END) && zip->entry_bytes_remaining < 1) zip->end_of_entry = 1; /* Set up a more descriptive format name. */ sprintf(zip->format_name, "ZIP %d.%d (%s)", version / 10, version % 10, compression_name(zip->entry->compression)); a->archive.archive_format_name = zip->format_name; return (ret); } static const char * compression_name(int compression) { static const char *compression_names[] = { "uncompressed", "shrinking", "reduced-1", "reduced-2", "reduced-3", "reduced-4", "imploded", "reserved", "deflation" }; if (0 <= compression && compression < (int)(sizeof(compression_names)/sizeof(compression_names[0]))) return compression_names[compression]; else return "??"; } /* Convert an MSDOS-style date/time into Unix-style time. */ static time_t zip_time(const char *p) { int msTime, msDate; struct tm ts; msTime = (0xff & (unsigned)p[0]) + 256 * (0xff & (unsigned)p[1]); msDate = (0xff & (unsigned)p[2]) + 256 * (0xff & (unsigned)p[3]); memset(&ts, 0, sizeof(ts)); ts.tm_year = ((msDate >> 9) & 0x7f) + 80; /* Years since 1900. */ ts.tm_mon = ((msDate >> 5) & 0x0f) - 1; /* Month number. */ ts.tm_mday = msDate & 0x1f; /* Day of month. */ ts.tm_hour = (msTime >> 11) & 0x1f; ts.tm_min = (msTime >> 5) & 0x3f; ts.tm_sec = (msTime << 1) & 0x3e; ts.tm_isdst = -1; return mktime(&ts); } static int archive_read_format_zip_read_data(struct archive_read *a, const void **buff, size_t *size, int64_t *offset) { int r; struct zip *zip = (struct zip *)(a->format->data); *offset = zip->entry_uncompressed_bytes_read; *size = 0; *buff = NULL; /* If we hit end-of-entry last time, return ARCHIVE_EOF. */ if (zip->end_of_entry) return (ARCHIVE_EOF); /* Return EOF immediately if this is a non-regular file. */ if (AE_IFREG != (zip->entry->mode & AE_IFMT)) return (ARCHIVE_EOF); if (zip->entry->flags & (ZIP_ENCRYPTED | ZIP_STRONG_ENCRYPTED)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Encrypted file is unsupported"); return (ARCHIVE_FAILED); } __archive_read_consume(a, zip->unconsumed); zip->unconsumed = 0; switch(zip->entry->compression) { case 0: /* No compression. */ r = zip_read_data_none(a, buff, size, offset); break; #ifdef HAVE_ZLIB_H case 8: /* Deflate compression. */ r = zip_read_data_deflate(a, buff, size, offset); break; #endif default: /* Unsupported compression. */ /* Return a warning. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Unsupported ZIP compression method (%s)", compression_name(zip->entry->compression)); /* We can't decompress this entry, but we will * be able to skip() it and try the next entry. */ return (ARCHIVE_FAILED); break; } if (r != ARCHIVE_OK) return (r); /* Update checksum */ if (*size) zip->entry_crc32 = crc32(zip->entry_crc32, *buff, *size); /* If we hit the end, swallow any end-of-data marker. */ if (zip->end_of_entry) { /* Check file size, CRC against these values. */ if (zip->entry->compressed_size != zip->entry_compressed_bytes_read) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "ZIP compressed data is wrong size (read %jd, expected %jd)", (intmax_t)zip->entry_compressed_bytes_read, (intmax_t)zip->entry->compressed_size); return (ARCHIVE_WARN); } /* Size field only stores the lower 32 bits of the actual * size. */ if ((zip->entry->uncompressed_size & UINT32_MAX) != (zip->entry_uncompressed_bytes_read & UINT32_MAX)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "ZIP uncompressed data is wrong size (read %jd, expected %jd)", (intmax_t)zip->entry_uncompressed_bytes_read, (intmax_t)zip->entry->uncompressed_size); return (ARCHIVE_WARN); } /* Check computed CRC against header */ if (zip->entry->crc32 != zip->entry_crc32) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "ZIP bad CRC: 0x%lx should be 0x%lx", (unsigned long)zip->entry_crc32, (unsigned long)zip->entry->crc32); return (ARCHIVE_WARN); } } return (ARCHIVE_OK); } /* * Read "uncompressed" data. There are three cases: * 1) We know the size of the data. This is always true for the * seeking reader (we've examined the Central Directory already). * 2) ZIP_LENGTH_AT_END was set, but only the CRC was deferred. * Info-ZIP seems to do this; we know the size but have to grab * the CRC from the data descriptor afterwards. * 3) We're streaming and ZIP_LENGTH_AT_END was specified and * we have no size information. In this case, we can do pretty * well by watching for the data descriptor record. The data * descriptor is 16 bytes and includes a computed CRC that should * provide a strong check. * * TODO: Technically, the PK\007\010 signature is optional. * In the original spec, the data descriptor contained CRC * and size fields but had no leading signature. In practice, * newer writers seem to provide the signature pretty consistently, * but we might need to do something more complex here if * we want to handle older archives that lack that signature. * * Returns ARCHIVE_OK if successful, ARCHIVE_FATAL otherwise, sets * zip->end_of_entry if it consumes all of the data. */ static int zip_read_data_none(struct archive_read *a, const void **_buff, size_t *size, int64_t *offset) { struct zip *zip; const char *buff; ssize_t bytes_avail; (void)offset; /* UNUSED */ zip = (struct zip *)(a->format->data); if (zip->entry->flags & ZIP_LENGTH_AT_END) { const char *p; /* Grab at least 16 bytes. */ buff = __archive_read_ahead(a, 16, &bytes_avail); if (bytes_avail < 16) { /* Zip archives have end-of-archive markers that are longer than this, so a failure to get at least 16 bytes really does indicate a truncated file. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file data"); return (ARCHIVE_FATAL); } /* Check for a complete PK\007\010 signature. */ p = buff; if (p[0] == 'P' && p[1] == 'K' && p[2] == '\007' && p[3] == '\010' && archive_le32dec(p + 4) == zip->entry_crc32 && archive_le32dec(p + 8) == zip->entry_compressed_bytes_read && archive_le32dec(p + 12) == zip->entry_uncompressed_bytes_read) { zip->entry->crc32 = archive_le32dec(p + 4); zip->entry->compressed_size = archive_le32dec(p + 8); zip->entry->uncompressed_size = archive_le32dec(p + 12); zip->end_of_entry = 1; zip->unconsumed = 16; return (ARCHIVE_OK); } /* If not at EOF, ensure we consume at least one byte. */ ++p; /* Scan forward until we see where a PK\007\010 signature might be. */ /* Return bytes up until that point. On the next call, the code above will verify the data descriptor. */ while (p < buff + bytes_avail - 4) { if (p[3] == 'P') { p += 3; } else if (p[3] == 'K') { p += 2; } else if (p[3] == '\007') { p += 1; } else if (p[3] == '\010' && p[2] == '\007' && p[1] == 'K' && p[0] == 'P') { break; } else { p += 4; } } bytes_avail = p - buff; } else { if (zip->entry_bytes_remaining == 0) { zip->end_of_entry = 1; return (ARCHIVE_OK); } /* Grab a bunch of bytes. */ buff = __archive_read_ahead(a, 1, &bytes_avail); if (bytes_avail <= 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file data"); return (ARCHIVE_FATAL); } if (bytes_avail > zip->entry_bytes_remaining) bytes_avail = zip->entry_bytes_remaining; } *size = bytes_avail; zip->entry_bytes_remaining -= bytes_avail; zip->entry_uncompressed_bytes_read += bytes_avail; zip->entry_compressed_bytes_read += bytes_avail; zip->unconsumed += bytes_avail; *_buff = buff; return (ARCHIVE_OK); } #ifdef HAVE_ZLIB_H static int zip_read_data_deflate(struct archive_read *a, const void **buff, size_t *size, int64_t *offset) { struct zip *zip; ssize_t bytes_avail; const void *compressed_buff; int r; (void)offset; /* UNUSED */ zip = (struct zip *)(a->format->data); /* If the buffer hasn't been allocated, allocate it now. */ if (zip->uncompressed_buffer == NULL) { zip->uncompressed_buffer_size = 256 * 1024; zip->uncompressed_buffer = (unsigned char *)malloc(zip->uncompressed_buffer_size); if (zip->uncompressed_buffer == NULL) { archive_set_error(&a->archive, ENOMEM, "No memory for ZIP decompression"); return (ARCHIVE_FATAL); } } /* If we haven't yet read any data, initialize the decompressor. */ if (!zip->decompress_init) { if (zip->stream_valid) r = inflateReset(&zip->stream); else r = inflateInit2(&zip->stream, -15 /* Don't check for zlib header */); if (r != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't initialize ZIP decompression."); return (ARCHIVE_FATAL); } /* Stream structure has been set up. */ zip->stream_valid = 1; /* We've initialized decompression for this stream. */ zip->decompress_init = 1; } /* * Note: '1' here is a performance optimization. * Recall that the decompression layer returns a count of * available bytes; asking for more than that forces the * decompressor to combine reads by copying data. */ compressed_buff = __archive_read_ahead(a, 1, &bytes_avail); if (0 == (zip->entry->flags & ZIP_LENGTH_AT_END) && bytes_avail > zip->entry_bytes_remaining) { bytes_avail = zip->entry_bytes_remaining; } if (bytes_avail <= 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file body"); return (ARCHIVE_FATAL); } /* * A bug in zlib.h: stream.next_in should be marked 'const' * but isn't (the library never alters data through the * next_in pointer, only reads it). The result: this ugly * cast to remove 'const'. */ zip->stream.next_in = (Bytef *)(uintptr_t)(const void *)compressed_buff; zip->stream.avail_in = bytes_avail; zip->stream.total_in = 0; zip->stream.next_out = zip->uncompressed_buffer; zip->stream.avail_out = zip->uncompressed_buffer_size; zip->stream.total_out = 0; r = inflate(&zip->stream, 0); switch (r) { case Z_OK: break; case Z_STREAM_END: zip->end_of_entry = 1; break; case Z_MEM_ERROR: archive_set_error(&a->archive, ENOMEM, "Out of memory for ZIP decompression"); return (ARCHIVE_FATAL); default: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "ZIP decompression failed (%d)", r); return (ARCHIVE_FATAL); } /* Consume as much as the compressor actually used. */ bytes_avail = zip->stream.total_in; __archive_read_consume(a, bytes_avail); zip->entry_bytes_remaining -= bytes_avail; zip->entry_compressed_bytes_read += bytes_avail; *size = zip->stream.total_out; zip->entry_uncompressed_bytes_read += zip->stream.total_out; *buff = zip->uncompressed_buffer; if (zip->end_of_entry && (zip->entry->flags & ZIP_LENGTH_AT_END)) { const char *p; if (NULL == (p = __archive_read_ahead(a, 16, NULL))) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP end-of-file record"); return (ARCHIVE_FATAL); } /* Consume the optional PK\007\010 marker. */ if (p[0] == 'P' && p[1] == 'K' && p[2] == '\007' && p[3] == '\010') { zip->entry->crc32 = archive_le32dec(p + 4); zip->entry->compressed_size = archive_le32dec(p + 8); zip->entry->uncompressed_size = archive_le32dec(p + 12); zip->unconsumed = 16; } } return (ARCHIVE_OK); } #endif static int archive_read_format_zip_read_data_skip(struct archive_read *a) { struct zip *zip; zip = (struct zip *)(a->format->data); /* If we've already read to end of data, we're done. */ if (zip->end_of_entry) return (ARCHIVE_OK); /* If we're seeking, we're done. */ if (zip->have_central_directory) return (ARCHIVE_OK); /* So we know we're streaming... */ if (0 == (zip->entry->flags & ZIP_LENGTH_AT_END)) { /* We know the compressed length, so we can just skip. */ int64_t bytes_skipped = __archive_read_consume(a, zip->entry_bytes_remaining + zip->unconsumed); if (bytes_skipped < 0) return (ARCHIVE_FATAL); zip->unconsumed = 0; return (ARCHIVE_OK); } /* We're streaming and we don't know the length. */ /* If the body is compressed and we know the format, we can * find an exact end-of-entry by decompressing it. */ switch (zip->entry->compression) { #ifdef HAVE_ZLIB_H case 8: /* Deflate compression. */ while (!zip->end_of_entry) { int64_t offset = 0; const void *buff = NULL; size_t size = 0; int r; r = zip_read_data_deflate(a, &buff, &size, &offset); if (r != ARCHIVE_OK) return (r); } break; #endif default: /* Uncompressed or unknown. */ /* Scan for a PK\007\010 signature. */ __archive_read_consume(a, zip->unconsumed); zip->unconsumed = 0; for (;;) { const char *p, *buff; ssize_t bytes_avail; buff = __archive_read_ahead(a, 16, &bytes_avail); if (bytes_avail < 16) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Truncated ZIP file data"); return (ARCHIVE_FATAL); } p = buff; while (p <= buff + bytes_avail - 16) { if (p[3] == 'P') { p += 3; } else if (p[3] == 'K') { p += 2; } else if (p[3] == '\007') { p += 1; } else if (p[3] == '\010' && p[2] == '\007' && p[1] == 'K' && p[0] == 'P') { __archive_read_consume(a, p - buff + 16); return ARCHIVE_OK; } else { p += 4; } } __archive_read_consume(a, p - buff); } } return ARCHIVE_OK; } static int archive_read_format_zip_cleanup(struct archive_read *a) { struct zip *zip; zip = (struct zip *)(a->format->data); #ifdef HAVE_ZLIB_H if (zip->stream_valid) inflateEnd(&zip->stream); #endif free(zip->zip_entries); free(zip->uncompressed_buffer); archive_string_free(&(zip->extra)); free(zip); (a->format->data) = NULL; return (ARCHIVE_OK); } /* * The extra data is stored as a list of * id1+size1+data1 + id2+size2+data2 ... * triplets. id and size are 2 bytes each. */ static void process_extra(const char *p, size_t extra_length, struct zip_entry* zip_entry) { unsigned offset = 0; while (offset < extra_length - 4) { unsigned short headerid = archive_le16dec(p + offset); unsigned short datasize = archive_le16dec(p + offset + 2); offset += 4; if (offset + datasize > extra_length) break; #ifdef DEBUG fprintf(stderr, "Header id 0x%x, length %d\n", headerid, datasize); #endif switch (headerid) { case 0x0001: /* Zip64 extended information extra field. */ if (datasize >= 8) zip_entry->uncompressed_size = archive_le64dec(p + offset); if (datasize >= 16) zip_entry->compressed_size = archive_le64dec(p + offset + 8); break; case 0x5455: { /* Extended time field "UT". */ int flags = p[offset]; offset++; datasize--; /* Flag bits indicate which dates are present. */ if (flags & 0x01) { #ifdef DEBUG fprintf(stderr, "mtime: %lld -> %d\n", (long long)zip_entry->mtime, archive_le32dec(p + offset)); #endif if (datasize < 4) break; zip_entry->mtime = archive_le32dec(p + offset); offset += 4; datasize -= 4; } if (flags & 0x02) { if (datasize < 4) break; zip_entry->atime = archive_le32dec(p + offset); offset += 4; datasize -= 4; } if (flags & 0x04) { if (datasize < 4) break; zip_entry->ctime = archive_le32dec(p + offset); offset += 4; datasize -= 4; } break; } case 0x5855: { /* Info-ZIP Unix Extra Field (old version) "UX". */ if (datasize >= 8) { zip_entry->atime = archive_le32dec(p + offset); zip_entry->mtime = archive_le32dec(p + offset + 4); } if (datasize >= 12) { zip_entry->uid = archive_le16dec(p + offset + 8); zip_entry->gid = archive_le16dec(p + offset + 10); } break; } case 0x7855: /* Info-ZIP Unix Extra Field (type 2) "Ux". */ #ifdef DEBUG fprintf(stderr, "uid %d gid %d\n", archive_le16dec(p + offset), archive_le16dec(p + offset + 2)); #endif if (datasize >= 2) zip_entry->uid = archive_le16dec(p + offset); if (datasize >= 4) zip_entry->gid = archive_le16dec(p + offset + 2); break; case 0x7875: { /* Info-Zip Unix Extra Field (type 3) "ux". */ int uidsize = 0, gidsize = 0; if (datasize >= 1 && p[offset] == 1) {/* version=1 */ if (datasize >= 4) { /* get a uid size. */ uidsize = p[offset+1]; if (uidsize == 2) zip_entry->uid = archive_le16dec( p + offset + 2); else if (uidsize == 4 && datasize >= 6) zip_entry->uid = archive_le32dec( p + offset + 2); } if (datasize >= (2 + uidsize + 3)) { /* get a gid size. */ gidsize = p[offset+2+uidsize]; if (gidsize == 2) zip_entry->gid = archive_le16dec( p+offset+2+uidsize+1); else if (gidsize == 4 && datasize >= (2 + uidsize + 5)) zip_entry->gid = archive_le32dec( p+offset+2+uidsize+1); } } break; } default: break; } offset += datasize; } #ifdef DEBUG if (offset != extra_length) { fprintf(stderr, "Extra data field contents do not match reported size!\n"); } #endif } Index: vendor/libarchive/dist/libarchive/archive_string.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_string.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_string.c (revision 231297) @@ -1,4237 +1,4237 @@ /*- * Copyright (c) 2003-2011 Tim Kientzle * Copyright (c) 2011 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD: head/lib/libarchive/archive_string.c 201095 2009-12-28 02:33:22Z kientzle $"); /* * Basic resizable string support, to simplify manipulating arbitrary-sized * strings while minimizing heap activity. * * In particular, the buffer used by a string object is only grown, it * never shrinks, so you can clear and reuse the same string object * without incurring additional memory allocations. */ #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_ICONV_H #include #endif #ifdef HAVE_LANGINFO_H #include #endif #ifdef HAVE_LOCALCHARSET_H #include #endif #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_WCHAR_H #include #endif #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #endif #if defined(__APPLE__) #include #endif #include "archive_endian.h" #include "archive_private.h" #include "archive_string.h" #include "archive_string_composition.h" #if !defined(HAVE_WMEMCPY) && !defined(wmemcpy) #define wmemcpy(a,b,i) (wchar_t *)memcpy((a), (b), (i) * sizeof(wchar_t)) #endif struct archive_string_conv { struct archive_string_conv *next; char *from_charset; char *to_charset; unsigned from_cp; unsigned to_cp; /* Set 1 if from_charset and to_charset are the same. */ int same; int flag; #define SCONV_TO_CHARSET 1 /* MBS is being converted to specified * charset. */ #define SCONV_FROM_CHARSET (1<<1) /* MBS is being converted from * specified charset. */ #define SCONV_BEST_EFFORT (1<<2) /* Copy at least ASCII code. */ #define SCONV_WIN_CP (1<<3) /* Use Windows API for converting * MBS. */ #define SCONV_UTF8_LIBARCHIVE_2 (1<<4) /* Incorrect UTF-8 made by libarchive * 2.x in the wrong assumption. */ #define SCONV_NORMALIZATION_C (1<<6) /* Need normalization to be Form C. * Before UTF-8 characters are actually * processed. */ #define SCONV_NORMALIZATION_D (1<<7) /* Need normalization to be Form D. * Before UTF-8 characters are actually * processed. * Currently this only for MAC OS X. */ #define SCONV_TO_UTF8 (1<<8) /* "to charset" side is UTF-8. */ #define SCONV_FROM_UTF8 (1<<9) /* "from charset" side is UTF-8. */ #define SCONV_TO_UTF16BE (1<<10) /* "to charset" side is UTF-16BE. */ #define SCONV_FROM_UTF16BE (1<<11) /* "from charset" side is UTF-16BE. */ #define SCONV_TO_UTF16LE (1<<12) /* "to charset" side is UTF-16LE. */ #define SCONV_FROM_UTF16LE (1<<13) /* "from charset" side is UTF-16LE. */ #define SCONV_TO_UTF16 (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE) #define SCONV_FROM_UTF16 (SCONV_FROM_UTF16BE | SCONV_FROM_UTF16LE) #if HAVE_ICONV iconv_t cd; iconv_t cd_w;/* Use at archive_mstring on * Windows. */ #endif /* A temporary buffer for normalization. */ struct archive_string utftmp; #if defined(__APPLE__) UnicodeToTextInfo uniInfo; struct archive_string utf16nfc; struct archive_string utf16nfd; #endif int (*converter[2])(struct archive_string *, const void *, size_t, struct archive_string_conv *); int nconverter; }; #define CP_C_LOCALE 0 /* "C" locale only for this file. */ #define CP_UTF16LE 1200 #define CP_UTF16BE 1201 #define IS_HIGH_SURROGATE_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDBFF) #define IS_LOW_SURROGATE_LA(uc) ((uc) >= 0xDC00 && (uc) <= 0xDFFF) #define IS_SURROGATE_PAIR_LA(uc) ((uc) >= 0xD800 && (uc) <= 0xDFFF) #define UNICODE_MAX 0x10FFFF #define UNICODE_R_CHAR 0xFFFD /* Replacement character. */ /* Set U+FFFD(Replacement character) in UTF-8. */ #define UTF8_SET_R_CHAR(outp) do { \ (outp)[0] = 0xef; \ (outp)[1] = 0xbf; \ (outp)[2] = 0xbd; \ } while (0) #define UTF8_R_CHAR_SIZE 3 static struct archive_string_conv *find_sconv_object(struct archive *, const char *, const char *); static void add_sconv_object(struct archive *, struct archive_string_conv *); static struct archive_string_conv *create_sconv_object(const char *, const char *, unsigned, int); static void free_sconv_object(struct archive_string_conv *); static struct archive_string_conv *get_sconv_object(struct archive *, const char *, const char *, int); static unsigned make_codepage_from_charset(const char *); static unsigned get_current_codepage(void); static unsigned get_current_oemcp(void); static size_t mbsnbytes(const void *, size_t); static size_t utf16nbytes(const void *, size_t); #if defined(_WIN32) && !defined(__CYGWIN__) static int archive_wstring_append_from_mbs_in_codepage( struct archive_wstring *, const char *, size_t, struct archive_string_conv *); static int archive_string_append_from_wcs_in_codepage(struct archive_string *, const wchar_t *, size_t, struct archive_string_conv *); static int is_big_endian(void); static int strncat_in_codepage(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_from_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_from_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_to_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int win_strncat_to_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); #endif static int best_effort_strncat_from_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_from_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_to_utf16be(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int best_effort_strncat_to_utf16le(struct archive_string *, const void *, size_t, struct archive_string_conv *); #if defined(HAVE_ICONV) static int iconv_strncat_in_locale(struct archive_string *, const void *, size_t, struct archive_string_conv *); #endif static int best_effort_strncat_in_locale(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int _utf8_to_unicode(uint32_t *, const char *, size_t); static int utf8_to_unicode(uint32_t *, const char *, size_t); static inline uint32_t combine_surrogate_pair(uint32_t, uint32_t); static int cesu8_to_unicode(uint32_t *, const char *, size_t); static size_t unicode_to_utf8(char *, size_t, uint32_t); static int utf16_to_unicode(uint32_t *, const char *, size_t, int); static size_t unicode_to_utf16be(char *, size_t, uint32_t); static size_t unicode_to_utf16le(char *, size_t, uint32_t); static int strncat_from_utf8_libarchive2(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int strncat_from_utf8_to_utf8(struct archive_string *, const void *, size_t, struct archive_string_conv *); static int archive_string_normalize_C(struct archive_string *, const void *, size_t, struct archive_string_conv *); #if defined(__APPLE__) static int archive_string_normalize_D(struct archive_string *, const void *, size_t, struct archive_string_conv *); #endif static int archive_string_append_unicode(struct archive_string *, const void *, size_t, struct archive_string_conv *); static struct archive_string * archive_string_append(struct archive_string *as, const char *p, size_t s) { if (archive_string_ensure(as, as->length + s + 1) == NULL) return (NULL); memcpy(as->s + as->length, p, s); as->length += s; as->s[as->length] = 0; return (as); } static struct archive_wstring * archive_wstring_append(struct archive_wstring *as, const wchar_t *p, size_t s) { if (archive_wstring_ensure(as, as->length + s + 1) == NULL) return (NULL); wmemcpy(as->s + as->length, p, s); as->length += s; as->s[as->length] = 0; return (as); } void archive_string_concat(struct archive_string *dest, struct archive_string *src) { if (archive_string_append(dest, src->s, src->length) == NULL) __archive_errx(1, "Out of memory"); } void archive_wstring_concat(struct archive_wstring *dest, struct archive_wstring *src) { if (archive_wstring_append(dest, src->s, src->length) == NULL) __archive_errx(1, "Out of memory"); } void archive_string_free(struct archive_string *as) { as->length = 0; as->buffer_length = 0; free(as->s); as->s = NULL; } void archive_wstring_free(struct archive_wstring *as) { as->length = 0; as->buffer_length = 0; free(as->s); as->s = NULL; } struct archive_wstring * archive_wstring_ensure(struct archive_wstring *as, size_t s) { return (struct archive_wstring *) archive_string_ensure((struct archive_string *)as, s * sizeof(wchar_t)); } /* Returns NULL on any allocation failure. */ struct archive_string * archive_string_ensure(struct archive_string *as, size_t s) { char *p; size_t new_length; /* If buffer is already big enough, don't reallocate. */ if (as->s && (s <= as->buffer_length)) return (as); /* * Growing the buffer at least exponentially ensures that * append operations are always linear in the number of * characters appended. Using a smaller growth rate for * larger buffers reduces memory waste somewhat at the cost of * a larger constant factor. */ if (as->buffer_length < 32) /* Start with a minimum 32-character buffer. */ new_length = 32; else if (as->buffer_length < 8192) /* Buffers under 8k are doubled for speed. */ new_length = as->buffer_length + as->buffer_length; else { /* Buffers 8k and over grow by at least 25% each time. */ new_length = as->buffer_length + as->buffer_length / 4; /* Be safe: If size wraps, fail. */ if (new_length < as->buffer_length) { /* On failure, wipe the string and return NULL. */ archive_string_free(as); errno = ENOMEM;/* Make sure errno has ENOMEM. */ return (NULL); } } /* * The computation above is a lower limit to how much we'll * grow the buffer. In any case, we have to grow it enough to * hold the request. */ if (new_length < s) new_length = s; /* Now we can reallocate the buffer. */ p = (char *)realloc(as->s, new_length); if (p == NULL) { /* On failure, wipe the string and return NULL. */ archive_string_free(as); errno = ENOMEM;/* Make sure errno has ENOMEM. */ return (NULL); } as->s = p; as->buffer_length = new_length; return (as); } /* * TODO: See if there's a way to avoid scanning * the source string twice. Then test to see * if it actually helps (remember that we're almost * always called with pretty short arguments, so * such an optimization might not help). */ struct archive_string * archive_strncat(struct archive_string *as, const void *_p, size_t n) { size_t s; const char *p, *pp; p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } if ((as = archive_string_append(as, p, s)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_wstring * archive_wstrncat(struct archive_wstring *as, const wchar_t *p, size_t n) { size_t s; const wchar_t *pp; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } if ((as = archive_wstring_append(as, p, s)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_string * archive_strcat(struct archive_string *as, const void *p) { /* strcat is just strncat without an effective limit. * Assert that we'll never get called with a source * string over 16MB. * TODO: Review all uses of strcat in the source * and try to replace them with strncat(). */ return archive_strncat(as, p, 0x1000000); } struct archive_wstring * archive_wstrcat(struct archive_wstring *as, const wchar_t *p) { /* Ditto. */ return archive_wstrncat(as, p, 0x1000000); } struct archive_string * archive_strappend_char(struct archive_string *as, char c) { if ((as = archive_string_append(as, &c, 1)) == NULL) __archive_errx(1, "Out of memory"); return (as); } struct archive_wstring * archive_wstrappend_wchar(struct archive_wstring *as, wchar_t c) { if ((as = archive_wstring_append(as, &c, 1)) == NULL) __archive_errx(1, "Out of memory"); return (as); } /* * Get the "current character set" name to use with iconv. * On FreeBSD, the empty character set name "" chooses * the correct character encoding for the current locale, * so this isn't necessary. * But iconv on Mac OS 10.6 doesn't seem to handle this correctly; * on that system, we have to explicitly call nl_langinfo() * to get the right name. Not sure about other platforms. * * NOTE: GNU libiconv does not recognize the character-set name * which some platform nl_langinfo(CODESET) returns, so we should * use locale_charset() instead of nl_langinfo(CODESET) for GNU libiconv. */ static const char * default_iconv_charset(const char *charset) { if (charset != NULL && charset[0] != '\0') return charset; #if HAVE_LOCALE_CHARSET && !defined(__APPLE__) /* locale_charset() is broken on Mac OS */ return locale_charset(); #elif HAVE_NL_LANGINFO return nl_langinfo(CODESET); #else return ""; #endif } #if defined(_WIN32) && !defined(__CYGWIN__) /* * Convert MBS to WCS. * Note: returns -1 if conversion fails. */ int archive_wstring_append_from_mbs(struct archive_wstring *dest, const char *p, size_t len) { int r = archive_wstring_append_from_mbs_in_codepage(dest, p, len, NULL); if (r != 0 && errno == ENOMEM) __archive_errx(1, "No memory"); return (r); } static int archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest, const char *s, size_t length, struct archive_string_conv *sc) { int count, ret = 0; UINT from_cp; if (sc != NULL) from_cp = sc->from_cp; else from_cp = get_current_codepage(); if (from_cp == CP_C_LOCALE) { /* * "C" locale special process. */ wchar_t *ws; const unsigned char *mp; if (NULL == archive_wstring_ensure(dest, dest->length + length + 1)) return (-1); ws = dest->s + dest->length; mp = (const unsigned char *)s; count = 0; while (count < (int)length && *mp) { *ws++ = (wchar_t)*mp++; count++; } } else if (sc != NULL && (sc->flag & SCONV_NORMALIZATION_C)) { /* * Normalize UTF-8 and UTF-16BE and convert it directly * to UTF-16 as wchar_t. */ struct archive_string u16; int saved_flag = sc->flag;/* save current flag. */ if (is_big_endian()) sc->flag |= SCONV_TO_UTF16BE; else sc->flag |= SCONV_TO_UTF16LE; if (sc->flag & SCONV_FROM_UTF16) { /* * UTF-16BE/LE NFD ===> UTF-16 NFC */ count = utf16nbytes(s, length); } else { /* * UTF-8 NFD ===> UTF-16 NFC */ count = mbsnbytes(s, length); } u16.s = (char *)dest->s; u16.length = dest->length << 1;; u16.buffer_length = dest->buffer_length; ret = archive_string_normalize_C(&u16, s, count, sc); dest->s = (wchar_t *)u16.s; dest->length = u16.length >> 1; dest->buffer_length = u16.buffer_length; sc->flag = saved_flag;/* restore the saved flag. */ return (ret); } else if (sc != NULL && (sc->flag & SCONV_FROM_UTF16)) { count = utf16nbytes(s, length); count >>= 1; /* to be WCS length */ /* Allocate memory for WCS. */ if (NULL == archive_wstring_ensure(dest, dest->length + count + 1)) return (-1); wmemcpy(dest->s + dest->length, (wchar_t *)s, count); if ((sc->flag & SCONV_FROM_UTF16BE) && !is_big_endian()) { uint16_t *u16 = (uint16_t *)(dest->s + dest->length); int b; for (b = 0; b < count; b++) { uint16_t val = archive_le16dec(u16+b); archive_be16enc(u16+b, val); } } else if ((sc->flag & SCONV_FROM_UTF16LE) && is_big_endian()) { uint16_t *u16 = (uint16_t *)(dest->s + dest->length); int b; for (b = 0; b < count; b++) { uint16_t val = archive_be16dec(u16+b); archive_le16enc(u16+b, val); } } } else { DWORD mbflag; if (sc == NULL) mbflag = 0; else if (sc->flag & SCONV_FROM_CHARSET) { /* Do not trust the length which comes from * an archive file. */ length = mbsnbytes(s, length); mbflag = 0; } else mbflag = MB_PRECOMPOSED; if (length == 0) { /* * We do not need to convert any characters but make * sure `dest' has a valid buffer(no NULL pointer). */ if (NULL == archive_wstring_ensure(dest, dest->length + 1)) return (-1); dest->s[dest->length] = L'\0'; return (0); } /* * Count how many bytes are needed for WCS. */ count = MultiByteToWideChar(from_cp, mbflag, s, length, NULL, 0); if (count == 0) { if (dest->s == NULL) { if (NULL == archive_wstring_ensure(dest, dest->length + 1)) return (-1); } dest->s[dest->length] = L'\0'; return (-1); } /* Allocate memory for WCS. */ if (NULL == archive_wstring_ensure(dest, dest->length + count + 1)) return (-1); /* Convert MBS to WCS. */ count = MultiByteToWideChar(from_cp, mbflag, s, length, dest->s + dest->length, count); if (count == 0) ret = -1; } dest->length += count; dest->s[dest->length] = L'\0'; return (ret); } #elif defined(HAVE_MBSNRTOWCS) /* * Convert MBS to WCS. * Note: returns -1 if conversion fails. */ int archive_wstring_append_from_mbs(struct archive_wstring *dest, const char *p, size_t len) { size_t r; /* * No single byte will be more than one wide character, * so this length estimate will always be big enough. */ size_t wcs_length = len; size_t mbs_length = len; const char *mbs = p; wchar_t *wcs; mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) __archive_errx(1, "No memory for archive_wstring_append_from_mbs()"); wcs = dest->s + dest->length; r = mbsnrtowcs(wcs, &mbs, mbs_length, wcs_length, &shift_state); if (r != (size_t)-1) { dest->length += r; dest->s[dest->length] = L'\0'; return (0); } dest->s[dest->length] = L'\0'; return (-1); } #else /* * Convert MBS to WCS. * Note: returns -1 if conversion fails. */ int archive_wstring_append_from_mbs(struct archive_wstring *dest, const char *p, size_t len) { size_t r; /* * No single byte will be more than one wide character, * so this length estimate will always be big enough. */ size_t wcs_length = len; size_t mbs_length = len; const char *mbs = p; wchar_t *wcs; #if HAVE_MBRTOWC mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #endif if (NULL == archive_wstring_ensure(dest, dest->length + wcs_length + 1)) __archive_errx(1, "No memory for archive_wstring_append_from_mbs()"); wcs = dest->s + dest->length; /* * We cannot use mbsrtowcs/mbstowcs here because those may convert * extra MBS when strlen(p) > len and one wide character consis of * multi bytes. */ while (wcs_length > 0 && *mbs && mbs_length > 0) { #if HAVE_MBRTOWC r = mbrtowc(wcs, mbs, wcs_length, &shift_state); #else r = mbtowc(wcs, mbs, wcs_length); #endif if (r == (size_t)-1 || r == (size_t)-2) { dest->s[dest->length] = L'\0'; return (-1); } if (r == 0 || r > mbs_length) break; wcs++; wcs_length--; mbs += r; mbs_length -= r; } dest->length = wcs - dest->s; dest->s[dest->length] = L'\0'; return (0); } #endif #if defined(_WIN32) && !defined(__CYGWIN__) /* * WCS ==> MBS. * Note: returns -1 if conversion fails. * * Win32 builds use WideCharToMultiByte from the Windows API. * (Maybe Cygwin should too? WideCharToMultiByte will know a * lot more about local character encodings than the wcrtomb() * wrapper is going to know.) */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { int r = archive_string_append_from_wcs_in_codepage(as, w, len, NULL); if (r != 0 && errno == ENOMEM) __archive_errx(1, "No memory"); return (r); } static int archive_string_append_from_wcs_in_codepage(struct archive_string *as, const wchar_t *ws, size_t len, struct archive_string_conv *sc) { BOOL defchar_used, *dp; int count, ret = 0; UINT to_cp; int wslen = (int)len; if (sc != NULL) to_cp = sc->to_cp; else to_cp = get_current_codepage(); if (to_cp == CP_C_LOCALE) { /* * "C" locale special process. */ const wchar_t *wp = ws; char *p; if (NULL == archive_string_ensure(as, as->length + wslen +1)) return (-1); p = as->s + as->length; count = 0; defchar_used = 0; while (count < wslen && *wp) { if (*wp > 255) { *p++ = '?'; wp++; defchar_used = 1; } else *p++ = (char)*wp++; count++; } } else if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) { uint16_t *u16; if (NULL == archive_string_ensure(as, as->length + len * 2 + 2)) return (-1); u16 = (uint16_t *)(as->s + as->length); count = 0; defchar_used = 0; if (sc->flag & SCONV_TO_UTF16BE) { while (count < (int)len && *ws) { archive_be16enc(u16+count, *ws); ws++; count++; } } else { while (count < (int)len && *ws) { archive_le16enc(u16+count, *ws); ws++; count++; } } count <<= 1; /* to be byte size */ } else { /* Make sure the MBS buffer has plenty to set. */ if (NULL == archive_string_ensure(as, as->length + len * 2 + 1)) return (-1); do { defchar_used = 0; if (to_cp == CP_UTF8 || sc == NULL) dp = NULL; else dp = &defchar_used; count = WideCharToMultiByte(to_cp, 0, ws, wslen, as->s + as->length, as->buffer_length-1, NULL, dp); if (count == 0 && GetLastError() == ERROR_INSUFFICIENT_BUFFER) { /* Expand the MBS buffer and retry. */ if (NULL == archive_string_ensure(as, as->buffer_length + len)) return (-1); continue; } if (count == 0) ret = -1; } while (0); } as->length += count; as->s[as->length] = '\0'; return (defchar_used?-1:ret); } #elif defined(HAVE_WCSNRTOMBS) /* * Translates a wide character string into current locale character set * and appends to the archive_string. Note: returns -1 if conversion * fails. */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { mbstate_t shift_state; size_t r, ndest, nwc; char *dest; const wchar_t *wp, *wpp; int ret_val = 0; wp = w; nwc = len; ndest = len * 2; /* Initialize the shift state. */ memset(&shift_state, 0, sizeof(shift_state)); while (nwc > 0) { /* Allocate buffer for MBS. */ if (archive_string_ensure(as, as->length + ndest + 1) == NULL) __archive_errx(1, "Out of memory"); dest = as->s + as->length; wpp = wp; r = wcsnrtombs(dest, &wp, nwc, as->buffer_length - as->length -1, &shift_state); if (r == (size_t)-1) { if (errno == EILSEQ) { /* Retry conversion just for safe WCS. */ size_t xwc = wp - wpp; wp = wpp; r = wcsnrtombs(dest, &wp, xwc, as->buffer_length - as->length -1, &shift_state); if (r == (size_t)-1) /* This would not happen. */ return (-1); as->length += r; nwc -= wp - wpp; /* Skip an illegal wide char. */ as->s[as->length++] = '?'; wp++; nwc--; ret_val = -1; continue; } else { ret_val = -1; break; } } as->length += r; if (wp == NULL || (wp - wpp) >= nwc) break; /* Get a remaining WCS lenth. */ nwc -= wp - wpp; } /* All wide characters are translated to MBS. */ as->s[as->length] = '\0'; return (ret_val); } #elif defined(HAVE_WCTOMB) || defined(HAVE_WCRTOMB) /* * Translates a wide character string into current locale character set * and appends to the archive_string. Note: returns -1 if conversion * fails. */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { /* We cannot use the standard wcstombs() here because it * cannot tell us how big the output buffer should be. So * I've built a loop around wcrtomb() or wctomb() that * converts a character at a time and resizes the string as * needed. We prefer wcrtomb() when it's available because * it's thread-safe. */ int n, ret_val = 0; char *p; char *end; #if HAVE_WCRTOMB mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ wctomb(NULL, L'\0'); #endif /* * Allocate buffer for MBS. * We need this allocation here since it is possible that * as->s is still NULL. */ if (archive_string_ensure(as, as->length + len + 1) == NULL) __archive_errx(1, "Out of memory"); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; while (*w != L'\0' && len > 0) { if (p >= end) { as->length = p - as->s; as->s[as->length] = '\0'; /* Re-allocate buffer for MBS. */ if (archive_string_ensure(as, as->length + len * 2 + 1) == NULL) __archive_errx(1, "Out of memory"); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; } #if HAVE_WCRTOMB n = wcrtomb(p, *w++, &shift_state); #else n = wctomb(p, *w++); #endif if (n == -1) { if (errno == EILSEQ) { /* Skip an illegal wide char. */ *p++ = '?'; ret_val = -1; } else { ret_val = -1; break; } } else p += n; len--; } as->length = p - as->s; as->s[as->length] = '\0'; return (ret_val); } #else /* HAVE_WCTOMB || HAVE_WCRTOMB */ /* * TODO: Test if __STDC_ISO_10646__ is defined. * Non-Windows uses ISO C wcrtomb() or wctomb() to perform the conversion * one character at a time. If a non-Windows platform doesn't have * either of these, fall back to the built-in UTF8 conversion. */ int archive_string_append_from_wcs(struct archive_string *as, const wchar_t *w, size_t len) { (void)as;/* UNUSED */ (void)w;/* UNUSED */ (void)len;/* UNUSED */ return (-1); } #endif /* HAVE_WCTOMB || HAVE_WCRTOMB */ /* * Find a string conversion object by a pair of 'from' charset name * and 'to' charset name from an archive object. * Return NULL if not found. */ static struct archive_string_conv * find_sconv_object(struct archive *a, const char *fc, const char *tc) { struct archive_string_conv *sc; if (a == NULL) return (NULL); for (sc = a->sconv; sc != NULL; sc = sc->next) { if (strcmp(sc->from_charset, fc) == 0 && strcmp(sc->to_charset, tc) == 0) break; } return (sc); } /* * Register a string object to an archive object. */ static void add_sconv_object(struct archive *a, struct archive_string_conv *sc) { struct archive_string_conv **psc; /* Add a new sconv to sconv list. */ psc = &(a->sconv); while (*psc != NULL) psc = &((*psc)->next); *psc = sc; } #if defined(__APPLE__) static int createUniInfo(struct archive_string_conv *sconv) { UnicodeMapping map; OSStatus err; map.unicodeEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, kUnicodeNoSubset, kUnicode16BitFormat); map.otherEncoding = CreateTextEncoding(kTextEncodingUnicodeDefault, kUnicodeHFSPlusDecompVariant, kUnicode16BitFormat); map.mappingVersion = kUnicodeUseLatestMapping; sconv->uniInfo = NULL; err = CreateUnicodeToTextInfo(&map, &(sconv->uniInfo)); return ((err == noErr)? 0: -1); } #endif /* __APPLE__ */ static void add_converter(struct archive_string_conv *sc, int (*converter) (struct archive_string *, const void *, size_t, struct archive_string_conv *)) { if (sc == NULL || sc->nconverter >= 2) __archive_errx(1, "Programing error"); sc->converter[sc->nconverter++] = converter; } static void setup_converter(struct archive_string_conv *sc) { /* Reset. */ sc->nconverter = 0; /* * Perform special sequence for the incorrect UTF-8 filenames * made by libarchive2.x. */ if (sc->flag & SCONV_UTF8_LIBARCHIVE_2) { add_converter(sc, strncat_from_utf8_libarchive2); return; } /* * Convert a string to UTF-16BE/LE. */ if (sc->flag & SCONV_TO_UTF16) { /* * If the current locale is UTF-8, we can translate * a UTF-8 string into a UTF-16BE string. */ if (sc->flag & SCONV_FROM_UTF8) { add_converter(sc, archive_string_append_unicode); return; } #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->flag & SCONV_WIN_CP) { if (sc->flag & SCONV_TO_UTF16BE) add_converter(sc, win_strncat_to_utf16be); else add_converter(sc, win_strncat_to_utf16le); return; } #endif #if defined(HAVE_ICONV) if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); return; } #endif if (sc->flag & SCONV_BEST_EFFORT) { if (sc->flag & SCONV_TO_UTF16BE) add_converter(sc, best_effort_strncat_to_utf16be); else add_converter(sc, best_effort_strncat_to_utf16le); } else /* Make sure we have no converter. */ sc->nconverter = 0; return; } /* * Convert a string from UTF-16BE/LE. */ if (sc->flag & SCONV_FROM_UTF16) { /* * At least we should normalize a UTF-16BE string. */ #if defined(__APPLE__) if (sc->flag & SCONV_NORMALIZATION_D) add_converter(sc,archive_string_normalize_D); else #endif if (sc->flag & SCONV_NORMALIZATION_C) add_converter(sc, archive_string_normalize_C); if (sc->flag & SCONV_TO_UTF8) { /* * If the current locale is UTF-8, we can translate * a UTF-16BE/LE string into a UTF-8 string directly. */ if (!(sc->flag & (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) add_converter(sc, archive_string_append_unicode); return; } #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->flag & SCONV_WIN_CP) { if (sc->flag & SCONV_FROM_UTF16BE) add_converter(sc, win_strncat_from_utf16be); else add_converter(sc, win_strncat_from_utf16le); return; } #endif #if defined(HAVE_ICONV) if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); return; } #endif if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16BE)) add_converter(sc, best_effort_strncat_from_utf16be); else if ((sc->flag & (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) == (SCONV_BEST_EFFORT | SCONV_FROM_UTF16LE)) add_converter(sc, best_effort_strncat_from_utf16le); else /* Make sure we have no converter. */ sc->nconverter = 0; return; } if (sc->flag & SCONV_FROM_UTF8) { /* * At least we should normalize a UTF-8 string. */ #if defined(__APPLE__) if (sc->flag & SCONV_NORMALIZATION_D) add_converter(sc,archive_string_normalize_D); else #endif if (sc->flag & SCONV_NORMALIZATION_C) add_converter(sc, archive_string_normalize_C); /* * Copy UTF-8 string with a check of CESU-8. * Apparently, iconv does not check surrogate pairs in UTF-8 * when both from-charset and to-charset are UTF-8, and then * we use our UTF-8 copy code. */ if (sc->flag & SCONV_TO_UTF8) { /* * If the current locale is UTF-8, we can translate * a UTF-16BE string into a UTF-8 string directly. */ if (!(sc->flag & (SCONV_NORMALIZATION_D |SCONV_NORMALIZATION_C))) add_converter(sc, strncat_from_utf8_to_utf8); return; } } #if defined(_WIN32) && !defined(__CYGWIN__) /* * On Windows we can use Windows API for a string conversion. */ if (sc->flag & SCONV_WIN_CP) { add_converter(sc, strncat_in_codepage); return; } #endif #if HAVE_ICONV if (sc->cd != (iconv_t)-1) { add_converter(sc, iconv_strncat_in_locale); return; } #endif /* * Try conversion in the best effort or no conversion. */ if ((sc->flag & SCONV_BEST_EFFORT) || sc->same) add_converter(sc, best_effort_strncat_in_locale); else /* Make sure we have no converter. */ sc->nconverter = 0; } /* * Return canonicalized charset-name but this supports just UTF-8, UTF-16BE * and CP932 which are referenced in create_sconv_object(). */ static const char * canonical_charset_name(const char *charset) { char cs[16]; char *p; const char *s; if (charset == NULL || charset[0] == '\0' || strlen(charset) > 15) return (charset); /* Copy name to uppercase. */ p = cs; s = charset; while (*s) { char c = *s++; if (c >= 'a' && c <= 'z') c -= 'a' - 'A'; *p++ = c; } *p++ = '\0'; if (strcmp(cs, "UTF-8") == 0 || strcmp(cs, "UTF8") == 0) return ("UTF-8"); if (strcmp(cs, "UTF-16BE") == 0 || strcmp(cs, "UTF16BE") == 0) return ("UTF-16BE"); if (strcmp(cs, "UTF-16LE") == 0 || strcmp(cs, "UTF16LE") == 0) return ("UTF-16LE"); if (strcmp(cs, "CP932") == 0) return ("CP932"); return (charset); } /* * Create a string conversion object. */ static struct archive_string_conv * create_sconv_object(const char *fc, const char *tc, unsigned current_codepage, int flag) { struct archive_string_conv *sc; sc = calloc(1, sizeof(*sc)); if (sc == NULL) return (NULL); sc->next = NULL; sc->from_charset = strdup(fc); if (sc->from_charset == NULL) { free(sc); return (NULL); } sc->to_charset = strdup(tc); if (sc->to_charset == NULL) { free(sc); free(sc->from_charset); return (NULL); } archive_string_init(&sc->utftmp); #if defined(__APPLE__) archive_string_init(&sc->utf16nfc); archive_string_init(&sc->utf16nfd); #endif if (flag & SCONV_TO_CHARSET) { /* * Convert characters from the current locale charset to * a specified charset. */ sc->from_cp = current_codepage; sc->to_cp = make_codepage_from_charset(tc); #if defined(_WIN32) && !defined(__CYGWIN__) if (IsValidCodePage(sc->to_cp)) flag |= SCONV_WIN_CP; #endif } else if (flag & SCONV_FROM_CHARSET) { /* * Convert characters from a specified charset to * the current locale charset. */ sc->to_cp = current_codepage; sc->from_cp = make_codepage_from_charset(fc); #if defined(_WIN32) && !defined(__CYGWIN__) if (IsValidCodePage(sc->from_cp)) flag |= SCONV_WIN_CP; #endif } /* * Check if "from charset" and "to charset" are the same. */ if (strcmp(fc, tc) == 0 || (sc->from_cp != (unsigned)-1 && sc->from_cp == sc->to_cp)) sc->same = 1; else sc->same = 0; /* * Mark if "from charset" or "to charset" are UTF-8 or UTF-16BE/LE. */ if (strcmp(tc, "UTF-8") == 0) flag |= SCONV_TO_UTF8; else if (strcmp(tc, "UTF-16BE") == 0) flag |= SCONV_TO_UTF16BE; else if (strcmp(tc, "UTF-16LE") == 0) flag |= SCONV_TO_UTF16LE; if (strcmp(fc, "UTF-8") == 0) flag |= SCONV_FROM_UTF8; else if (strcmp(fc, "UTF-16BE") == 0) flag |= SCONV_FROM_UTF16BE; else if (strcmp(fc, "UTF-16LE") == 0) flag |= SCONV_FROM_UTF16LE; #if defined(_WIN32) && !defined(__CYGWIN__) if (sc->to_cp == CP_UTF8) flag |= SCONV_TO_UTF8; else if (sc->to_cp == CP_UTF16BE) flag |= SCONV_TO_UTF16BE | SCONV_WIN_CP; else if (sc->to_cp == CP_UTF16LE) flag |= SCONV_TO_UTF16LE | SCONV_WIN_CP; if (sc->from_cp == CP_UTF8) flag |= SCONV_FROM_UTF8; else if (sc->from_cp == CP_UTF16BE) flag |= SCONV_FROM_UTF16BE | SCONV_WIN_CP; else if (sc->from_cp == CP_UTF16LE) flag |= SCONV_FROM_UTF16LE | SCONV_WIN_CP; #endif /* * Set a flag for Unicode NFD. Usually iconv cannot correctly * handle it. So we have to translate NFD characters to NFC ones * ourselves before iconv handles. Another reason is to prevent * that the same sight of two filenames, one is NFC and other * is NFD, would be in its directory. * On Mac OS X, although its filesystem layer automatically * convert filenames to NFD, it would be useful for filename * comparing to find out the same filenames that we normalize * that to be NFD ourselves. */ if ((flag & SCONV_FROM_CHARSET) && (flag & (SCONV_FROM_UTF16 | SCONV_FROM_UTF8))) { #if defined(__APPLE__) if (flag & SCONV_TO_UTF8) { if (createUniInfo(sc) == 0) flag |= SCONV_NORMALIZATION_D; } else #endif flag |= SCONV_NORMALIZATION_C; } #if defined(HAVE_ICONV) sc->cd_w = (iconv_t)-1; /* * Create an iconv object. */ if (((flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) && (flag & (SCONV_FROM_UTF8 | SCONV_FROM_UTF16))) || (flag & SCONV_WIN_CP)) { /* This case we won't use iconv. */ sc->cd = (iconv_t)-1; #if defined(__APPLE__) } else if ((flag & SCONV_FROM_CHARSET) && (flag & SCONV_TO_UTF8)) { /* * In case reading an archive file. * Translate non-Unicode filenames in an archive file to * UTF-8-MAC filenames. */ sc->cd = iconv_open("UTF-8-MAC", fc); if (sc->cd == (iconv_t)-1) { if ((sc->flag & SCONV_BEST_EFFORT) && strcmp(fc, "CP932") == 0) { sc->cd = iconv_open("UTF-8-MAC", "SJIS"); if (sc->cd == (iconv_t)-1) { sc->cd = iconv_open(tc, fc); if (sc->cd == (iconv_t)-1) sc->cd = iconv_open(tc, "SJIS"); } } else sc->cd = iconv_open(tc, fc); } } else if ((flag & SCONV_TO_CHARSET) && (flag & SCONV_FROM_UTF8)) { /* * In case writing an archive file. * Translate UTF-8-MAC filenames in HFS Plus to non-Unicode * filenames. */ sc->cd = iconv_open(tc, "UTF-8-MAC"); if (sc->cd == (iconv_t)-1) { if ((sc->flag & SCONV_BEST_EFFORT) && strcmp(tc, "CP932") == 0) { sc->cd = iconv_open("SJIS", "UTF-8-MAC"); if (sc->cd == (iconv_t)-1) { sc->cd = iconv_open(tc, fc); if (sc->cd == (iconv_t)-1) sc->cd = iconv_open("SJIS", fc); } } else sc->cd = iconv_open(tc, fc); } #endif } else { sc->cd = iconv_open(tc, fc); if (sc->cd == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { /* * Unfortunaly, all of iconv implements do support * "CP932" character-set, so we should use "SJIS" * instead if iconv_open failed. */ if (strcmp(tc, "CP932") == 0) sc->cd = iconv_open("SJIS", fc); else if (strcmp(fc, "CP932") == 0) sc->cd = iconv_open(tc, "SJIS"); } #if defined(_WIN32) && !defined(__CYGWIN__) /* * archive_mstring on Windows directly convert multi-bytes * into archive_wstring in order not to depend on locale * so that you can do a I18N programing. This will be * used only in archive_mstring_copy_mbs_len_l so far. */ if (flag & SCONV_FROM_CHARSET) { sc->cd_w = iconv_open("UTF-8", fc); if (sc->cd_w == (iconv_t)-1 && (sc->flag & SCONV_BEST_EFFORT)) { if (strcmp(fc, "CP932") == 0) sc->cd_w = iconv_open("UTF-8", "SJIS"); } } #endif /* _WIN32 && !__CYGWIN__ */ } #endif /* HAVE_ICONV */ sc->flag = flag; /* * Setup converters. */ setup_converter(sc); return (sc); } /* * Free a string conversion object. */ static void free_sconv_object(struct archive_string_conv *sc) { free(sc->from_charset); free(sc->to_charset); archive_string_free(&sc->utftmp); #if HAVE_ICONV if (sc->cd != (iconv_t)-1) iconv_close(sc->cd); if (sc->cd_w != (iconv_t)-1) iconv_close(sc->cd_w); #endif #if defined(__APPLE__) archive_string_free(&sc->utf16nfc); archive_string_free(&sc->utf16nfd); if (sc->uniInfo != NULL) DisposeUnicodeToTextInfo(&(sc->uniInfo)); #endif free(sc); } #if defined(_WIN32) && !defined(__CYGWIN__) static unsigned my_atoi(const char *p) { unsigned cp; cp = 0; while (*p) { if (*p >= '0' && *p <= '9') cp = cp * 10 + (*p - '0'); else return (-1); p++; } return (cp); } /* * Translate Charset name (as used by iconv) into CodePage (as used by Windows) * Return -1 if failed. * * Note: This translation code may be insufficient. */ static struct charset { const char *name; unsigned cp; } charsets[] = { /* MUST BE SORTED! */ {"ASCII", 1252}, {"ASMO-708", 708}, {"BIG5", 950}, {"CHINESE", 936}, {"CP367", 1252}, {"CP819", 1252}, {"CP1025", 21025}, {"DOS-720", 720}, {"DOS-862", 862}, {"EUC-CN", 51936}, {"EUC-JP", 51932}, {"EUC-KR", 949}, {"EUCCN", 51936}, {"EUCJP", 51932}, {"EUCKR", 949}, {"GB18030", 54936}, {"GB2312", 936}, {"HEBREW", 1255}, {"HZ-GB-2312", 52936}, {"IBM273", 20273}, {"IBM277", 20277}, {"IBM278", 20278}, {"IBM280", 20280}, {"IBM284", 20284}, {"IBM285", 20285}, {"IBM290", 20290}, {"IBM297", 20297}, {"IBM367", 1252}, {"IBM420", 20420}, {"IBM423", 20423}, {"IBM424", 20424}, {"IBM819", 1252}, {"IBM871", 20871}, {"IBM880", 20880}, {"IBM905", 20905}, {"IBM924", 20924}, {"ISO-8859-1", 28591}, {"ISO-8859-13", 28603}, {"ISO-8859-15", 28605}, {"ISO-8859-2", 28592}, {"ISO-8859-3", 28593}, {"ISO-8859-4", 28594}, {"ISO-8859-5", 28595}, {"ISO-8859-6", 28596}, {"ISO-8859-7", 28597}, {"ISO-8859-8", 28598}, {"ISO-8859-9", 28599}, {"ISO8859-1", 28591}, {"ISO8859-13", 28603}, {"ISO8859-15", 28605}, {"ISO8859-2", 28592}, {"ISO8859-3", 28593}, {"ISO8859-4", 28594}, {"ISO8859-5", 28595}, {"ISO8859-6", 28596}, {"ISO8859-7", 28597}, {"ISO8859-8", 28598}, {"ISO8859-9", 28599}, {"JOHAB", 1361}, {"KOI8-R", 20866}, {"KOI8-U", 21866}, {"KS_C_5601-1987", 949}, {"LATIN1", 1252}, {"LATIN2", 28592}, {"MACINTOSH", 10000}, {"SHIFT-JIS", 932}, {"SHIFT_JIS", 932}, {"SJIS", 932}, {"US", 1252}, {"US-ASCII", 1252}, {"UTF-16", 1200}, {"UTF-16BE", 1201}, {"UTF-16LE", 1200}, {"UTF-8", CP_UTF8}, {"X-EUROPA", 29001}, {"X-MAC-ARABIC", 10004}, {"X-MAC-CE", 10029}, {"X-MAC-CHINESEIMP", 10008}, {"X-MAC-CHINESETRAD", 10002}, {"X-MAC-CROATIAN", 10082}, {"X-MAC-CYRILLIC", 10007}, {"X-MAC-GREEK", 10006}, {"X-MAC-HEBREW", 10005}, {"X-MAC-ICELANDIC", 10079}, {"X-MAC-JAPANESE", 10001}, {"X-MAC-KOREAN", 10003}, {"X-MAC-ROMANIAN", 10010}, {"X-MAC-THAI", 10021}, {"X-MAC-TURKISH", 10081}, {"X-MAC-UKRAINIAN", 10017}, }; static unsigned make_codepage_from_charset(const char *charset) { char cs[16]; char *p; unsigned cp; int a, b; if (charset == NULL || strlen(charset) > 15) return -1; /* Copy name to uppercase. */ p = cs; while (*charset) { char c = *charset++; if (c >= 'a' && c <= 'z') c -= 'a' - 'A'; *p++ = c; } *p++ = '\0'; cp = -1; /* Look it up in the table first, so that we can easily * override CP367, which we map to 1252 instead of 367. */ a = 0; b = sizeof(charsets)/sizeof(charsets[0]); while (b > a) { int c = (b + a) / 2; int r = strcmp(charsets[c].name, cs); if (r < 0) a = c + 1; else if (r > 0) b = c; else return charsets[c].cp; } /* If it's not in the table, try to parse it. */ switch (*cs) { case 'C': if (cs[1] == 'P' && cs[2] >= '0' && cs[2] <= '9') { cp = my_atoi(cs + 2); } else if (strcmp(cs, "CP_ACP") == 0) cp = get_current_codepage(); else if (strcmp(cs, "CP_OEMCP") == 0) cp = get_current_oemcp(); break; case 'I': if (cs[1] == 'B' && cs[2] == 'M' && cs[3] >= '0' && cs[3] <= '9') { cp = my_atoi(cs + 3); } break; case 'W': if (strncmp(cs, "WINDOWS-", 8) == 0) { cp = my_atoi(cs + 8); if (cp != 874 && (cp < 1250 || cp > 1258)) cp = -1;/* This may invalid code. */ } break; } return (cp); } /* * Return ANSI Code Page of current locale set by setlocale(). */ static unsigned -get_current_codepage() +get_current_codepage(void) { char *locale, *p; unsigned cp; locale = setlocale(LC_CTYPE, NULL); if (locale == NULL) return (GetACP()); if (locale[0] == 'C' && locale[1] == '\0') return (CP_C_LOCALE); p = strrchr(locale, '.'); if (p == NULL) return (GetACP()); cp = my_atoi(p+1); if (cp <= 0) return (GetACP()); return (cp); } /* * Translation table between Locale Name and ACP/OEMCP. */ static struct { unsigned acp; unsigned ocp; const char *locale; } acp_ocp_map[] = { { 950, 950, "Chinese_Taiwan" }, { 936, 936, "Chinese_People's Republic of China" }, { 950, 950, "Chinese_Taiwan" }, { 1250, 852, "Czech_Czech Republic" }, { 1252, 850, "Danish_Denmark" }, { 1252, 850, "Dutch_Netherlands" }, { 1252, 850, "Dutch_Belgium" }, { 1252, 437, "English_United States" }, { 1252, 850, "English_Australia" }, { 1252, 850, "English_Canada" }, { 1252, 850, "English_New Zealand" }, { 1252, 850, "English_United Kingdom" }, { 1252, 437, "English_United States" }, { 1252, 850, "Finnish_Finland" }, { 1252, 850, "French_France" }, { 1252, 850, "French_Belgium" }, { 1252, 850, "French_Canada" }, { 1252, 850, "French_Switzerland" }, { 1252, 850, "German_Germany" }, { 1252, 850, "German_Austria" }, { 1252, 850, "German_Switzerland" }, { 1253, 737, "Greek_Greece" }, { 1250, 852, "Hungarian_Hungary" }, { 1252, 850, "Icelandic_Iceland" }, { 1252, 850, "Italian_Italy" }, { 1252, 850, "Italian_Switzerland" }, { 932, 932, "Japanese_Japan" }, { 949, 949, "Korean_Korea" }, { 1252, 850, "Norwegian (BokmOl)_Norway" }, { 1252, 850, "Norwegian (BokmOl)_Norway" }, { 1252, 850, "Norwegian-Nynorsk_Norway" }, { 1250, 852, "Polish_Poland" }, { 1252, 850, "Portuguese_Portugal" }, { 1252, 850, "Portuguese_Brazil" }, { 1251, 866, "Russian_Russia" }, { 1250, 852, "Slovak_Slovakia" }, { 1252, 850, "Spanish_Spain" }, { 1252, 850, "Spanish_Mexico" }, { 1252, 850, "Spanish_Spain" }, { 1252, 850, "Swedish_Sweden" }, { 1254, 857, "Turkish_Turkey" }, { 0, 0, NULL} }; /* * Return OEM Code Page of current locale set by setlocale(). */ static unsigned -get_current_oemcp() +get_current_oemcp(void) { int i; char *locale, *p; size_t len; locale = setlocale(LC_CTYPE, NULL); if (locale == NULL) return (GetOEMCP()); if (locale[0] == 'C' && locale[1] == '\0') return (CP_C_LOCALE); p = strrchr(locale, '.'); if (p == NULL) return (GetOEMCP()); len = p - locale; for (i = 0; acp_ocp_map[i].acp; i++) { if (strncmp(acp_ocp_map[i].locale, locale, len) == 0) return (acp_ocp_map[i].ocp); } return (GetOEMCP()); } #else /* * POSIX platform does not use CodePage. */ static unsigned -get_current_codepage() +get_current_codepage(void) { return (-1);/* Unknown */ } static unsigned make_codepage_from_charset(const char *charset) { (void)charset; /* UNUSED */ return (-1);/* Unknown */ } static unsigned -get_current_oemcp() +get_current_oemcp(void) { return (-1);/* Unknown */ } #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ /* * Return a string conversion object. */ static struct archive_string_conv * get_sconv_object(struct archive *a, const char *fc, const char *tc, int flag) { struct archive_string_conv *sc; unsigned current_codepage; /* Check if we have made the sconv object. */ sc = find_sconv_object(a, fc, tc); if (sc != NULL) return (sc); if (a == NULL) current_codepage = get_current_codepage(); else current_codepage = a->current_codepage; sc = create_sconv_object(canonical_charset_name(fc), canonical_charset_name(tc), current_codepage, flag); if (sc == NULL) { if (a != NULL) archive_set_error(a, ENOMEM, "Could not allocate memory for " "a string conversion object"); return (NULL); } /* * If there is no converter for current string conversion object, * we cannot handle this conversion. */ if (sc->nconverter == 0) { if (a != NULL) { #if HAVE_ICONV archive_set_error(a, ARCHIVE_ERRNO_MISC, "iconv_open failed : Cannot handle ``%s''", (flag & SCONV_TO_CHARSET)?tc:fc); #else archive_set_error(a, ARCHIVE_ERRNO_MISC, "A character-set conversion not fully supported " "on this platform"); #endif } /* Failed; free a sconv object. */ free_sconv_object(sc); return (NULL); } /* * Success! */ if (a != NULL) add_sconv_object(a, sc); return (sc); } static const char * get_current_charset(struct archive *a) { const char *cur_charset; if (a == NULL) cur_charset = default_iconv_charset(""); else { cur_charset = default_iconv_charset(a->current_code); if (a->current_code == NULL) { a->current_code = strdup(cur_charset); a->current_codepage = get_current_codepage(); a->current_oemcp = get_current_oemcp(); } } return (cur_charset); } /* * Make and Return a string conversion object. * Return NULL if the platform does not support the specified conversion * and best_effort is 0. * If best_effort is set, A string conversion object must be returned * unless memory allocation for the object fails, but the conversion * might fail when non-ASCII code is found. */ struct archive_string_conv * archive_string_conversion_to_charset(struct archive *a, const char *charset, int best_effort) { int flag = SCONV_TO_CHARSET; if (best_effort) flag |= SCONV_BEST_EFFORT; return (get_sconv_object(a, get_current_charset(a), charset, flag)); } struct archive_string_conv * archive_string_conversion_from_charset(struct archive *a, const char *charset, int best_effort) { int flag = SCONV_FROM_CHARSET; if (best_effort) flag |= SCONV_BEST_EFFORT; return (get_sconv_object(a, charset, get_current_charset(a), flag)); } /* * archive_string_default_conversion_*_archive() are provided for Windows * platform because other archiver application use CP_OEMCP for * MultiByteToWideChar() and WideCharToMultiByte() for the filenames * in tar or zip files. But mbstowcs/wcstombs(CRT) usually use CP_ACP * unless you use setlocale(LC_ALL, ".OCP")(specify CP_OEMCP). * So we should make a string conversion between CP_ACP and CP_OEMCP * for compatibillty. */ #if defined(_WIN32) && !defined(__CYGWIN__) struct archive_string_conv * archive_string_default_conversion_for_read(struct archive *a) { const char *cur_charset = get_current_charset(a); char oemcp[16]; /* NOTE: a check of cur_charset is unneeded but we need * that get_current_charset() has been surely called at * this time whatever C compiler optimized. */ if (cur_charset != NULL && (a->current_codepage == CP_C_LOCALE || a->current_codepage == a->current_oemcp)) return (NULL);/* no conversion. */ _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); /* Make sure a null termination must be set. */ oemcp[sizeof(oemcp)-1] = '\0'; return (get_sconv_object(a, oemcp, cur_charset, SCONV_FROM_CHARSET)); } struct archive_string_conv * archive_string_default_conversion_for_write(struct archive *a) { const char *cur_charset = get_current_charset(a); char oemcp[16]; /* NOTE: a check of cur_charset is unneeded but we need * that get_current_charset() has been surely called at * this time whatever C compiler optimized. */ if (cur_charset != NULL && (a->current_codepage == CP_C_LOCALE || a->current_codepage == a->current_oemcp)) return (NULL);/* no conversion. */ _snprintf(oemcp, sizeof(oemcp)-1, "CP%d", a->current_oemcp); /* Make sure a null termination must be set. */ oemcp[sizeof(oemcp)-1] = '\0'; return (get_sconv_object(a, cur_charset, oemcp, SCONV_TO_CHARSET)); } #else struct archive_string_conv * archive_string_default_conversion_for_read(struct archive *a) { (void)a; /* UNUSED */ return (NULL); } struct archive_string_conv * archive_string_default_conversion_for_write(struct archive *a) { (void)a; /* UNUSED */ return (NULL); } #endif /* * Dispose of all character conversion objects in the archive object. */ void archive_string_conversion_free(struct archive *a) { struct archive_string_conv *sc; struct archive_string_conv *sc_next; for (sc = a->sconv; sc != NULL; sc = sc_next) { sc_next = sc->next; free_sconv_object(sc); } a->sconv = NULL; free(a->current_code); a->current_code = NULL; } /* * Return a conversion charset name. */ const char * archive_string_conversion_charset_name(struct archive_string_conv *sc) { if (sc->flag & SCONV_TO_CHARSET) return (sc->to_charset); else return (sc->from_charset); } /* * Change the behavior of a string conversion. */ void archive_string_conversion_set_opt(struct archive_string_conv *sc, int opt) { switch (opt) { /* * A filename in UTF-8 was made with libarchive 2.x in a wrong * assumption that wchar_t was Unicode. * This option enables simulating the assumption in order to read * that filname correctly. */ case SCONV_SET_OPT_UTF8_LIBARCHIVE2X: #if (defined(_WIN32) && !defined(__CYGWIN__)) \ || defined(__STDC_ISO_10646__) || defined(__APPLE__) /* * Nothing to do for it since wchar_t on these platforms * is really Unicode. */ (void)sc; /* UNUSED */ #else if ((sc->flag & SCONV_UTF8_LIBARCHIVE_2) == 0) { sc->flag |= SCONV_UTF8_LIBARCHIVE_2; /* Re-setup string converters. */ setup_converter(sc); } #endif break; default: break; } } /* * * Copy one archive_string to another in locale conversion. * * archive_strncpy_in_locale(); * archive_strcpy_in_locale(); * */ static size_t mbsnbytes(const void *_p, size_t n) { size_t s; const char *p, *pp; if (_p == NULL) return (0); p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; while (s < n && *pp) { pp++; s++; } return (s); } static size_t utf16nbytes(const void *_p, size_t n) { size_t s; const char *p, *pp; if (_p == NULL) return (0); p = (const char *)_p; /* Like strlen(p), except won't examine positions beyond p[n]. */ s = 0; pp = p; n >>= 1; while (s < n && (pp[0] || pp[1])) { pp += 2; s++; } return (s<<1); } int archive_strncpy_in_locale(struct archive_string *as, const void *_p, size_t n, struct archive_string_conv *sc) { as->length = 0; return (archive_strncat_in_locale(as, _p, n, sc)); } int archive_strncat_in_locale(struct archive_string *as, const void *_p, size_t n, struct archive_string_conv *sc) { const void *s; size_t length; int i, r = 0, r2; /* We must allocate memory even if there is no data for conversion * or copy. This simulates archive_string_append behavior. */ if (_p == NULL || n == 0) { int tn = 1; if (sc != NULL && (sc->flag & SCONV_TO_UTF16)) tn = 2; if (archive_string_ensure(as, as->length + tn) == NULL) return (-1); as->s[as->length] = 0; if (tn == 2) as->s[as->length+1] = 0; return (0); } /* * If sc is NULL, we just make a copy. */ if (sc == NULL) { length = mbsnbytes(_p, n); if (archive_string_append(as, _p, length) == NULL) return (-1);/* No memory */ return (0); } if (sc->flag & SCONV_FROM_UTF16) length = utf16nbytes(_p, n); else length = mbsnbytes(_p, n); s = _p; i = 0; if (sc->nconverter > 1) { sc->utftmp.length = 0; r2 = sc->converter[0](&(sc->utftmp), s, length, sc); if (r2 != 0 && errno == ENOMEM) return (r2); if (r > r2) r = r2; s = sc->utftmp.s; length = sc->utftmp.length; ++i; } r2 = sc->converter[i](as, s, length, sc); if (r > r2) r = r2; return (r); } #if HAVE_ICONV /* * Return -1 if conversion failes. */ static int iconv_strncat_in_locale(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { ICONV_CONST char *inp; size_t remaining; iconv_t cd; char *outp; size_t avail, bs; int return_value = 0; /* success */ int to_size, from_size; if (sc->flag & SCONV_TO_UTF16) to_size = 2; else to_size = 1; if (sc->flag & SCONV_FROM_UTF16) from_size = 2; else from_size = 1; if (archive_string_ensure(as, as->length + length*2+to_size) == NULL) return (-1); cd = sc->cd; inp = (char *)(uintptr_t)_p; remaining = length; outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; while (remaining >= (size_t)from_size) { size_t result = iconv(cd, &inp, &remaining, &outp, &avail); if (result != (size_t)-1) break; /* Conversion completed. */ if (errno == EILSEQ || errno == EINVAL) { /* * If an output charset is UTF-8 or UTF-16BE/LE, * unknown character should be U+FFFD * (replacement character). */ if (sc->flag & (SCONV_TO_UTF8 | SCONV_TO_UTF16)) { size_t rbytes; if (sc->flag & SCONV_TO_UTF8) rbytes = UTF8_R_CHAR_SIZE; else rbytes = 2; if (avail < rbytes) { as->length = outp - as->s; bs = as->buffer_length + (remaining * to_size) + rbytes; if (NULL == archive_string_ensure(as, bs)) return (-1); outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; } if (sc->flag & SCONV_TO_UTF8) UTF8_SET_R_CHAR(outp); else if (sc->flag & SCONV_TO_UTF16BE) archive_be16enc(outp, UNICODE_R_CHAR); else archive_le16enc(outp, UNICODE_R_CHAR); outp += rbytes; avail -= rbytes; } else { /* Skip the illegal input bytes. */ *outp++ = '?'; avail--; } inp += from_size; remaining -= from_size; return_value = -1; /* failure */ } else { /* E2BIG no output buffer, * Increase an output buffer. */ as->length = outp - as->s; bs = as->buffer_length + remaining * 2; if (NULL == archive_string_ensure(as, bs)) return (-1); outp = as->s + as->length; avail = as->buffer_length - as->length - to_size; } } as->length = outp - as->s; as->s[as->length] = 0; if (to_size == 2) as->s[as->length+1] = 0; return (return_value); } #endif /* HAVE_ICONV */ #if defined(_WIN32) && !defined(__CYGWIN__) /* * Translate a string from a some CodePage to an another CodePage by * Windows APIs, and copy the result. Return -1 if conversion failes. */ static int strncat_in_codepage(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { const char *s = (const char *)_p; struct archive_wstring aws; size_t l; int r, saved_flag; archive_string_init(&aws); saved_flag = sc->flag; sc->flag &= ~(SCONV_NORMALIZATION_D | SCONV_NORMALIZATION_C); r = archive_wstring_append_from_mbs_in_codepage(&aws, s, length, sc); sc->flag = saved_flag; if (r != 0) { archive_wstring_free(&aws); if (errno != ENOMEM) archive_string_append(as, s, length); return (-1); } l = as->length; r = archive_string_append_from_wcs_in_codepage( as, aws.s, aws.length, sc); if (r != 0 && errno != ENOMEM && l == as->length) archive_string_append(as, s, length); archive_wstring_free(&aws); return (r); } /* * Test whether MBS ==> WCS is okay. */ static int invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) { const char *p = (const char *)_p; unsigned codepage; DWORD mbflag = MB_ERR_INVALID_CHARS; if (sc->flag & SCONV_FROM_CHARSET) codepage = sc->to_cp; else codepage = sc->from_cp; if (codepage == CP_C_LOCALE) return (0); if (codepage != CP_UTF8) mbflag |= MB_PRECOMPOSED; if (MultiByteToWideChar(codepage, mbflag, p, n, NULL, 0) == 0) return (-1); /* Invalid */ return (0); /* Okay */ } #else /* * Test whether MBS ==> WCS is okay. */ static int invalid_mbs(const void *_p, size_t n, struct archive_string_conv *sc) { const char *p = (const char *)_p; size_t r; (void)sc; /* UNUSED */ #if HAVE_MBRTOWC mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ mbtowc(NULL, NULL, 0); #endif while (n) { wchar_t wc; #if HAVE_MBRTOWC r = mbrtowc(&wc, p, n, &shift_state); #else r = mbtowc(&wc, p, n); #endif if (r == (size_t)-1 || r == (size_t)-2) return (-1);/* Invalid. */ if (r == 0) break; p += r; n -= r; } return (0); /* All Okey. */ } #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ /* * Basically returns -1 because we cannot make a conversion of charset * without iconv but in some cases this would return 0. * Returns 0 if all copied characters are ASCII. * Returns 0 if both from-locale and to-locale are the same and those * can be WCS with no error. */ static int best_effort_strncat_in_locale(struct archive_string *as, const void *_p, size_t length, struct archive_string_conv *sc) { size_t remaining; char *outp; const char *inp; size_t avail; int return_value = 0; /* success */ /* * If both from-locale and to-locale is the same, this makes a copy. * And then this checks all copied MBS can be WCS if so returns 0. */ if (sc->same) { if (archive_string_append(as, _p, length) == NULL) return (-1);/* No memory */ return (invalid_mbs(_p, length, sc)); } /* * If a character is ASCII, this just copies it. If not, this * assigns '?' charater instead but in UTF-8 locale this assigns * byte sequence 0xEF 0xBD 0xBD, which are code point U+FFFD, * a Replacement Character in Unicode. */ if (archive_string_ensure(as, as->length + length + 1) == NULL) return (-1); remaining = length; inp = (const char *)_p; outp = as->s + as->length; avail = as->buffer_length - as->length -1; while (*inp && remaining > 0) { if (*inp < 0 && (sc->flag & SCONV_TO_UTF8)) { if (avail < UTF8_R_CHAR_SIZE) { as->length = outp - as->s; if (NULL == archive_string_ensure(as, as->buffer_length + remaining + UTF8_R_CHAR_SIZE)) return (-1); outp = as->s + as->length; avail = as->buffer_length - as->length -1; } /* * When coping a string in UTF-8, unknown character * should be U+FFFD (replacement character). */ UTF8_SET_R_CHAR(outp); outp += UTF8_R_CHAR_SIZE; avail -= UTF8_R_CHAR_SIZE; inp++; remaining--; return_value = -1; } else if (*inp < 0) { *outp++ = '?'; inp++; remaining--; return_value = -1; } else { *outp++ = *inp++; remaining--; } } as->length = outp - as->s; as->s[as->length] = '\0'; return (return_value); } /* * Unicode conversion functions. * - UTF-8 <===> UTF-8 in removing surrogate pairs. * - UTF-8 NFD ===> UTF-8 NFC in removing surrogate pairs. * - UTF-8 made by libarchive 2.x ===> UTF-8. * - UTF-16BE <===> UTF-8. * */ /* * Utility to convert a single UTF-8 sequence. * * Usually return used bytes, return used byte in negative value when * a unicode character is replaced with U+FFFD. * See also http://unicode.org/review/pr-121.html Public Review Issue #121 * Recommended Practice for Replacement Characters. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch, i; int cnt; uint32_t wc; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if ((int)n < cnt) { cnt = n; for (i = 1; i < cnt; i++) { if ((s[i] & 0xc0) != 0x80) { cnt = i; break; } } goto invalid_sequence; } /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } if ((s[2] & 0xc0) != 0x80) { cnt = 2; goto invalid_sequence; } wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) goto invalid_sequence;/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) { cnt = 1; goto invalid_sequence; } if ((s[2] & 0xc0) != 0x80) { cnt = 2; goto invalid_sequence; } if ((s[3] & 0xc0) != 0x80) { cnt = 3; goto invalid_sequence; } wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) goto invalid_sequence;/* Overlong sequence. */ break; default: /* Others are all invalid sequence. */ if (ch == 0xc0 || ch == 0xc1) cnt = 2; else if (ch >= 0xf5 && ch <= 0xf7) cnt = 4; else if (ch >= 0xf8 && ch <= 0xfb) cnt = 5; else if (ch == 0xfc || ch == 0xfd) cnt = 6; else cnt = 1; if ((int)n < cnt) cnt = n; for (i = 1; i < cnt; i++) { if ((s[i] & 0xc0) != 0x80) { cnt = i; break; } } goto invalid_sequence; } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > UNICODE_MAX) goto invalid_sequence; /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); invalid_sequence: *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ return (cnt * -1); } static int utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { int cnt; cnt = _utf8_to_unicode(pwc, s, n); /* Any of Surrogate pair is not leagal Unicode values. */ if (cnt == 3 && IS_SURROGATE_PAIR_LA(*pwc)) return (-3); return (cnt); } static inline uint32_t combine_surrogate_pair(uint32_t uc, uint32_t uc2) { uc -= 0xD800; uc *= 0x400; uc += uc2 - 0xDC00; uc += 0x10000; return (uc); } /* * Convert a single UTF-8/CESU-8 sequence to a Unicode code point in * removing surrogate pairs. * * CESU-8: The Compatibility Encoding Scheme for UTF-16. * * Usually return used bytes, return used byte in negative value when * a unicode character is replaced with U+FFFD. */ static int cesu8_to_unicode(uint32_t *pwc, const char *s, size_t n) { uint32_t wc, wc2; int cnt; cnt = _utf8_to_unicode(&wc, s, n); if (cnt == 3 && IS_HIGH_SURROGATE_LA(wc)) { if (n - 3 < 3) { /* Invalid byte sequence. */ goto invalid_sequence; } cnt = _utf8_to_unicode(&wc2, s+3, n-3); if (cnt != 3 || !IS_LOW_SURROGATE_LA(wc2)) { /* Invalid byte sequence. */ goto invalid_sequence; } wc = combine_surrogate_pair(wc, wc2); cnt = 6; } else if (cnt == 3 && IS_LOW_SURROGATE_LA(wc)) { /* Invalid byte sequence. */ goto invalid_sequence; } *pwc = wc; return (cnt); invalid_sequence: *pwc = UNICODE_R_CHAR;/* set the Replacement Character instead. */ if (cnt > 0) cnt *= -1; return (cnt); } /* * Convert a Unicode code point to a single UTF-8 sequence. * * NOTE:This function does not check if the Unicode is leagal or not. * Please you definitely check it before calling this. */ static size_t unicode_to_utf8(char *p, size_t remaining, uint32_t uc) { char *_p = p; /* Translate code point to UTF8 */ if (uc <= 0x7f) { if (remaining == 0) return (0); *p++ = (char)uc; } else if (uc <= 0x7ff) { if (remaining < 2) return (0); *p++ = 0xc0 | ((uc >> 6) & 0x1f); *p++ = 0x80 | (uc & 0x3f); } else if (uc <= 0xffff) { if (remaining < 3) return (0); *p++ = 0xe0 | ((uc >> 12) & 0x0f); *p++ = 0x80 | ((uc >> 6) & 0x3f); *p++ = 0x80 | (uc & 0x3f); } else if (uc <= UNICODE_MAX) { if (remaining < 4) return (0); *p++ = 0xf0 | ((uc >> 18) & 0x07); *p++ = 0x80 | ((uc >> 12) & 0x3f); *p++ = 0x80 | ((uc >> 6) & 0x3f); *p++ = 0x80 | (uc & 0x3f); } else { /* * Undescribed code point should be U+FFFD * (replacement character). */ if (remaining < UTF8_R_CHAR_SIZE) return (0); UTF8_SET_R_CHAR(p); p += UTF8_R_CHAR_SIZE; } return (p - _p); } static int utf16be_to_unicode(uint32_t *pwc, const char *s, size_t n) { return (utf16_to_unicode(pwc, s, n, 1)); } static int utf16le_to_unicode(uint32_t *pwc, const char *s, size_t n) { return (utf16_to_unicode(pwc, s, n, 0)); } static int utf16_to_unicode(uint32_t *pwc, const char *s, size_t n, int be) { const char *utf16 = s; unsigned uc; if (n == 0) return (0); if (n == 1) { /* set the Replacement Character instead. */ *pwc = UNICODE_R_CHAR; return (-1); } if (be) uc = archive_be16dec(utf16); else uc = archive_le16dec(utf16); utf16 += 2; /* If this is a surrogate pair, assemble the full code point.*/ if (IS_HIGH_SURROGATE_LA(uc)) { unsigned uc2; if (n >= 4) { if (be) uc2 = archive_be16dec(utf16); else uc2 = archive_le16dec(utf16); } else uc2 = 0; if (IS_LOW_SURROGATE_LA(uc2)) { uc = combine_surrogate_pair(uc, uc2); utf16 += 2; } else { /* Undescribed code point should be U+FFFD * (replacement character). */ *pwc = UNICODE_R_CHAR; return (-2); } } /* * Surrogate pair values(0xd800 through 0xdfff) are only * used by UTF-16, so, after above culculation, the code * must not be surrogate values, and Unicode has no codes * larger than 0x10ffff. Thus, those are not leagal Unicode * values. */ if (IS_SURROGATE_PAIR_LA(uc) || uc > UNICODE_MAX) { /* Undescribed code point should be U+FFFD * (replacement character). */ *pwc = UNICODE_R_CHAR; return (((int)(utf16 - s)) * -1); } *pwc = uc; return ((int)(utf16 - s)); } static size_t unicode_to_utf16be(char *p, size_t remaining, uint32_t uc) { char *utf16 = p; if (uc > 0xffff) { /* We have a code point that won't fit into a * wchar_t; convert it to a surrogate pair. */ if (remaining < 4) return (0); uc -= 0x10000; archive_be16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); archive_be16enc(utf16+2, (uc & 0x3ff) + 0xDC00); return (4); } else { if (remaining < 2) return (0); archive_be16enc(utf16, uc); return (2); } } static size_t unicode_to_utf16le(char *p, size_t remaining, uint32_t uc) { char *utf16 = p; if (uc > 0xffff) { /* We have a code point that won't fit into a * wchar_t; convert it to a surrogate pair. */ if (remaining < 4) return (0); uc -= 0x10000; archive_le16enc(utf16, ((uc >> 10) & 0x3ff) + 0xD800); archive_le16enc(utf16+2, (uc & 0x3ff) + 0xDC00); return (4); } else { if (remaining < 2) return (0); archive_le16enc(utf16, uc); return (2); } } /* * Copy UTF-8 string in checking surrogate pair. * If any surrogate pair are found, it would be canonicalized. */ static int strncat_from_utf8_to_utf8(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; char *p, *endp; int n, ret = 0; (void)sc; /* UNUSED */ if (archive_string_ensure(as, as->length + len + 1) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; endp = as->s + as->buffer_length -1; do { uint32_t uc; const char *ss = s; size_t w; /* * Forward byte sequence until a conversion of that is needed. */ while ((n = utf8_to_unicode(&uc, s, len)) > 0) { s += n; len -= n; } if (ss < s) { if (p + (s - ss) > endp) { as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len + 1) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length -1; } memcpy(p, ss, s - ss); p += s - ss; } /* * If n is negative, current byte sequence needs a replacement. */ if (n < 0) { if (n == -3 && IS_SURROGATE_PAIR_LA(uc)) { /* Current byte sequence may be CESU-8. */ n = cesu8_to_unicode(&uc, s, len); } if (n < 0) { ret = -1; n *= -1;/* Use a replaced unicode character. */ } /* Rebuild UTF-8 byte sequence. */ while ((w = unicode_to_utf8(p, endp - p, uc)) == 0) { as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len + 1) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length -1; } p += w; s += n; len -= n; } } while (n > 0); as->length = p - as->s; as->s[as->length] = '\0'; return (ret); } static int archive_string_append_unicode(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; char *p, *endp; uint32_t uc; size_t w; int n, ret = 0, ts, tm; int (*parse)(uint32_t *, const char *, size_t); size_t (*unparse)(char *, size_t, uint32_t); if (sc->flag & SCONV_TO_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_TO_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else if (sc->flag & SCONV_TO_UTF8) { unparse = unicode_to_utf8; ts = 1; } else { /* * This case is going to be converted to another * character-set through iconv. */ if (sc->flag & SCONV_FROM_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_FROM_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else { unparse = unicode_to_utf8; ts = 1; } } if (sc->flag & SCONV_FROM_UTF16BE) { parse = utf16be_to_unicode; tm = 1; } else if (sc->flag & SCONV_FROM_UTF16LE) { parse = utf16le_to_unicode; tm = 1; } else { parse = cesu8_to_unicode; tm = ts; } if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; endp = as->s + as->buffer_length - ts; while ((n = parse(&uc, s, len)) != 0) { if (n < 0) { /* Use a replaced unicode character. */ n *= -1; ret = -1; } s += n; len -= n; while ((w = unparse(p, endp - p, uc)) == 0) { /* There is not enough output buffer so * we have to expand it. */ as->length = p - as->s; if (archive_string_ensure(as, as->buffer_length + len * tm + ts) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length - ts; } p += w; } as->length = p - as->s; as->s[as->length] = '\0'; if (ts == 2) as->s[as->length+1] = '\0'; return (ret); } /* * Following Constants for Hangul compositions this information comes from * Unicode Standard Annex #15 http://unicode.org/reports/tr15/ */ #define HC_SBASE 0xAC00 #define HC_LBASE 0x1100 #define HC_VBASE 0x1161 #define HC_TBASE 0x11A7 #define HC_LCOUNT 19 #define HC_VCOUNT 21 #define HC_TCOUNT 28 #define HC_NCOUNT (HC_VCOUNT * HC_TCOUNT) #define HC_SCOUNT (HC_LCOUNT * HC_NCOUNT) static uint32_t get_nfc(uint32_t uc, uint32_t uc2) { int t, b; t = 0; b = sizeof(u_composition_table)/sizeof(u_composition_table[0]) -1; while (b >= t) { int m = (t + b) / 2; if (u_composition_table[m].cp1 < uc) t = m + 1; else if (u_composition_table[m].cp1 > uc) b = m - 1; else if (u_composition_table[m].cp2 < uc2) t = m + 1; else if (u_composition_table[m].cp2 > uc2) b = m - 1; else return (u_composition_table[m].nfc); } return (0); } #define FDC_MAX 10 /* The maximum number of Following Decomposable * Characters. */ /* * Update first code point. */ #define UPDATE_UC(new_uc) do { \ uc = new_uc; \ ucptr = NULL; \ } while (0) /* * Replace first code point with second code point. */ #define REPLACE_UC_WITH_UC2() do { \ uc = uc2; \ ucptr = uc2ptr; \ n = n2; \ } while (0) #define EXPAND_BUFFER() do { \ as->length = p - as->s; \ if (archive_string_ensure(as, \ as->buffer_length + len * tm + ts) == NULL)\ return (-1); \ p = as->s + as->length; \ endp = as->s + as->buffer_length - ts; \ } while (0) #define UNPARSE(p, endp, uc) do { \ while ((w = unparse(p, (endp) - (p), uc)) == 0) {\ EXPAND_BUFFER(); \ } \ p += w; \ } while (0) /* * Write first code point. * If the code point has not be changed from its original code, * this just copies it from its original buffer pointer. * If not, this converts it to UTF-8 byte sequence and copies it. */ #define WRITE_UC() do { \ if (ucptr) { \ if (p + n > endp) \ EXPAND_BUFFER(); \ switch (n) { \ case 4: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 3: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 2: \ *p++ = *ucptr++; \ /* FALL THROUGH */ \ case 1: \ *p++ = *ucptr; \ break; \ } \ ucptr = NULL; \ } else { \ UNPARSE(p, endp, uc); \ } \ } while (0) /* * Collect following decomposable code points. */ #define COLLECT_CPS(start) do { \ int _i; \ for (_i = start; _i < FDC_MAX ; _i++) { \ nx = parse(&ucx[_i], s, len); \ if (nx <= 0) \ break; \ cx = CCC(ucx[_i]); \ if (cl >= cx && cl != 228 && cx != 228)\ break; \ s += nx; \ len -= nx; \ cl = cx; \ ccx[_i] = cx; \ } \ if (_i >= FDC_MAX) { \ ret = -1; \ ucx_size = FDC_MAX; \ } else \ ucx_size = _i; \ } while (0) /* * Normalize UTF-8/UTF-16BE characters to Form C and copy the result. * * TODO: Convert composition exclusions,which are never converted * from NFC,NFD,NFKC and NFKD, to Form C. */ static int archive_string_normalize_C(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s = (const char *)_p; char *p, *endp; uint32_t uc, uc2; size_t w; int always_replace, n, n2, ret = 0, spair, ts, tm; int (*parse)(uint32_t *, const char *, size_t); size_t (*unparse)(char *, size_t, uint32_t); always_replace = 1; ts = 1;/* text size. */ if (sc->flag & SCONV_TO_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; if (sc->flag & SCONV_FROM_UTF16BE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; if (sc->flag & SCONV_FROM_UTF16LE) always_replace = 0; } else if (sc->flag & SCONV_TO_UTF8) { unparse = unicode_to_utf8; if (sc->flag & SCONV_FROM_UTF8) always_replace = 0; } else { /* * This case is going to be converted to another * character-set through iconv. */ always_replace = 0; if (sc->flag & SCONV_FROM_UTF16BE) { unparse = unicode_to_utf16be; ts = 2; } else if (sc->flag & SCONV_FROM_UTF16LE) { unparse = unicode_to_utf16le; ts = 2; } else { unparse = unicode_to_utf8; } } if (sc->flag & SCONV_FROM_UTF16BE) { parse = utf16be_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else if (sc->flag & SCONV_FROM_UTF16LE) { parse = utf16le_to_unicode; tm = 1; spair = 4;/* surrogate pair size in UTF-16. */ } else { parse = cesu8_to_unicode; tm = ts; spair = 6;/* surrogate pair size in UTF-8. */ } if (archive_string_ensure(as, as->length + len * tm + ts) == NULL) return (-1); p = as->s + as->length; endp = as->s + as->buffer_length - ts; while ((n = parse(&uc, s, len)) != 0) { const char *ucptr, *uc2ptr; if (n < 0) { /* Use a replaced unicode character. */ UNPARSE(p, endp, uc); s += n*-1; len -= n*-1; ret = -1; continue; } else if (n == spair || always_replace) /* uc is converted from a surrogate pair. * this should be treated as a changed code. */ ucptr = NULL; else ucptr = s; s += n; len -= n; /* Read second code point. */ while ((n2 = parse(&uc2, s, len)) > 0) { uint32_t ucx[FDC_MAX]; int ccx[FDC_MAX]; int cl, cx, i, nx, ucx_size; int LIndex,SIndex; uint32_t nfc; if (n2 == spair || always_replace) /* uc2 is converted from a surrogate pair. * this should be treated as a changed code. */ uc2ptr = NULL; else uc2ptr = s; s += n2; len -= n2; /* * If current second code point is out of decomposable * code points, finding compositions is unneeded. */ if (!IS_DECOMPOSABLE_BLOCK(uc2)) { WRITE_UC(); REPLACE_UC_WITH_UC2(); continue; } /* * Try to combine current code points. */ /* * We have to combine Hangul characters according to * http://uniicode.org/reports/tr15/#Hangul */ if (0 <= (LIndex = uc - HC_LBASE) && LIndex < HC_LCOUNT) { /* * Hangul Composition. * 1. Two current code points are L and V. */ int VIndex = uc2 - HC_VBASE; if (0 <= VIndex && VIndex < HC_VCOUNT) { /* Make syllable of form LV. */ UPDATE_UC(HC_SBASE + (LIndex * HC_VCOUNT + VIndex) * HC_TCOUNT); } else { WRITE_UC(); REPLACE_UC_WITH_UC2(); } continue; } else if (0 <= (SIndex = uc - HC_SBASE) && SIndex < HC_SCOUNT && (SIndex % HC_TCOUNT) == 0) { /* * Hangul Composition. * 2. Two current code points are LV and T. */ int TIndex = uc2 - HC_TBASE; if (0 < TIndex && TIndex < HC_TCOUNT) { /* Make syllable of form LVT. */ UPDATE_UC(uc + TIndex); } else { WRITE_UC(); REPLACE_UC_WITH_UC2(); } continue; } else if ((nfc = get_nfc(uc, uc2)) != 0) { /* A composition to current code points * is found. */ UPDATE_UC(nfc); continue; } else if ((cl = CCC(uc2)) == 0) { /* Clearly 'uc2' the second code point is not * a decomposable code. */ WRITE_UC(); REPLACE_UC_WITH_UC2(); continue; } /* * Collect following decomposable code points. */ cx = 0; ucx[0] = uc2; ccx[0] = cl; COLLECT_CPS(1); /* * Find a composed code in the collected code points. */ i = 1; while (i < ucx_size) { int j; if ((nfc = get_nfc(uc, ucx[i])) == 0) { i++; continue; } /* * nfc is composed of uc and ucx[i]. */ UPDATE_UC(nfc); /* * Remove ucx[i] by shifting * following code points. */ for (j = i; j+1 < ucx_size; j++) { ucx[j] = ucx[j+1]; ccx[j] = ccx[j+1]; } ucx_size --; /* * Collect following code points blocked * by ucx[i] the removed code point. */ if (ucx_size > 0 && i == ucx_size && nx > 0 && cx == cl) { cl = ccx[ucx_size-1]; COLLECT_CPS(ucx_size); } /* * Restart finding a composed code with * the updated uc from the top of the * collected code points. */ i = 0; } /* * Apparently the current code points are not * decomposed characters or already composed. */ WRITE_UC(); for (i = 0; i < ucx_size; i++) UNPARSE(p, endp, ucx[i]); /* * Flush out remaining canonical combining characters. */ if (nx > 0 && cx == cl && len > 0) { while ((nx = parse(&ucx[0], s, len)) > 0) { cx = CCC(ucx[0]); if (cl > cx) break; s += nx; len -= nx; cl = cx; UNPARSE(p, endp, ucx[0]); } } break; } if (n2 < 0) { WRITE_UC(); /* Use a replaced unicode character. */ UNPARSE(p, endp, uc2); s += n2*-1; len -= n2*-1; ret = -1; continue; } else if (n2 == 0) { WRITE_UC(); break; } } as->length = p - as->s; as->s[as->length] = '\0'; if (ts == 2) as->s[as->length+1] = '\0'; return (ret); } #if defined(__APPLE__) /* * Normalize UTF-8 characters to Form D and copy the result. */ static int archive_string_normalize_D(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const UniChar *inp; char *outp; size_t newsize; ByteCount inCount, outCount; ByteCount inAvail, outAvail; OSStatus err; int ret, saved_flag; /* * Convert the current string to UTF-16LE for normalization. * The character-set of the current string must be UTF-16BE or * UTF-8. */ archive_string_empty(&(sc->utf16nfc)); saved_flag = sc->flag;/* save a flag. */ sc->flag &= ~(SCONV_TO_UTF16BE | SCONV_TO_UTF8); sc->flag |= SCONV_TO_UTF16LE; ret = archive_string_append_unicode(&(sc->utf16nfc), _p, len, sc); sc->flag = saved_flag;/* restore the saved flag */ if (archive_strlen(&(sc->utf16nfc)) == 0) { if (archive_string_ensure(as, as->length + 1) == NULL) return (-1); return (ret); } /* * Normalize an NFC string to be an NFD(HFS Plus version). */ newsize = sc->utf16nfc.length + 2; if (archive_string_ensure(&(sc->utf16nfd), newsize) == NULL) return (-1); inp = (UniChar *)sc->utf16nfc.s; inAvail = archive_strlen(&(sc->utf16nfc)); sc->utf16nfd.length = 0; outp = sc->utf16nfd.s; outAvail = sc->utf16nfd.buffer_length -2; do { /* Reinitialize all state information. */ if (ResetUnicodeToTextInfo(sc->uniInfo) != noErr) goto return_no_changed_data; inCount = outCount = 0; err = ConvertFromUnicodeToText(sc->uniInfo, inAvail, inp, kUnicodeDefaultDirectionMask, 0, NULL, NULL, NULL, outAvail, &inCount, &outCount, outp); if (err == noErr) { sc->utf16nfd.length = outCount; sc->utf16nfd.s[sc->utf16nfd.length] = 0; sc->utf16nfd.s[sc->utf16nfd.length+1] = 0; } else if (err == kTECOutputBufferFullStatus) { newsize = inAvail - inCount; if (newsize > inAvail) newsize = inAvail; newsize += sc->utf16nfd.buffer_length + 2; if (archive_string_ensure(&(sc->utf16nfd), newsize) == NULL) return (-1); outp = sc->utf16nfd.s; outAvail = sc->utf16nfd.buffer_length -2; } else goto return_no_changed_data; } while (err == kTECOutputBufferFullStatus); /* * If there is a next-step conversion, we should convert * a UTF-16LE(NFD) string back to the original Unicode type. */ saved_flag = sc->flag;/* save a flag. */ if (!(sc->flag & (SCONV_TO_UTF16BE | SCONV_TO_UTF16LE | SCONV_TO_UTF8))) { /* * This case is going to be converted to another * character-set through iconv. */ if (sc->flag & SCONV_FROM_UTF16BE) sc->flag |= SCONV_TO_UTF16BE; else if (sc->flag & SCONV_FROM_UTF16LE) sc->flag |= SCONV_TO_UTF16LE; else sc->flag |= SCONV_TO_UTF8; } sc->flag &= ~(SCONV_FROM_UTF16BE | SCONV_FROM_UTF8); sc->flag |= SCONV_FROM_UTF16LE; if (archive_string_append_unicode(as, sc->utf16nfd.s, sc->utf16nfd.length, sc) != 0) ret = -1; sc->flag = saved_flag;/* restore the saved flag */ return (ret); return_no_changed_data: /* * Something conversion error happened, so we return a no normalized * string with an error. */ (void)archive_string_append_unicode(as, _p, len, sc); return (-1); } #endif /* __APPLE__ */ /* * libarchive 2.x made incorrect UTF-8 strings in the wrong assumption * that WCS is Unicode. It is true for several platforms but some are false. * And then people who did not use UTF-8 locale on the non Unicode WCS * platform and made a tar file with libarchive(mostly bsdtar) 2.x. Those * now cannot get right filename from libarchive 3.x and later since we * fixed the wrong assumption and it is incompatible to older its versions. * So we provide special option, "compat-2x.x", for resolving it. * That option enable the string conversion of libarchive 2.x. * * Translates the wrong UTF-8 string made by libarchive 2.x into current * locale character set and appends to the archive_string. * Note: returns -1 if conversion fails. */ static int strncat_from_utf8_libarchive2(struct archive_string *as, const void *_p, size_t len, struct archive_string_conv *sc) { const char *s; int n; char *p; char *end; uint32_t unicode; #if HAVE_WCRTOMB mbstate_t shift_state; memset(&shift_state, 0, sizeof(shift_state)); #else /* Clear the shift state before starting. */ wctomb(NULL, L'\0'); #endif (void)sc; /* UNUSED */ /* * Allocate buffer for MBS. * We need this allocation here since it is possible that * as->s is still NULL. */ if (archive_string_ensure(as, as->length + len + 1) == NULL) return (-1); s = (const char *)_p; p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; while ((n = _utf8_to_unicode(&unicode, s, len)) != 0) { wchar_t wc; if (p >= end) { as->length = p - as->s; /* Re-allocate buffer for MBS. */ if (archive_string_ensure(as, as->length + len * 2 + 1) == NULL) return (-1); p = as->s + as->length; end = as->s + as->buffer_length - MB_CUR_MAX -1; } /* * As libarchie 2.x, translates the UTF-8 characters into * wide-characters in the assumption that WCS is Unicode. */ if (n < 0) { n *= -1; wc = L'?'; } else wc = (wchar_t)unicode; s += n; len -= n; /* * Translates the wide-character into the current locale MBS. */ #if HAVE_WCRTOMB n = wcrtomb(p, wc, &shift_state); #else n = wctomb(p, wc); #endif if (n == -1) return (-1); p += n; } as->length = p - as->s; as->s[as->length] = '\0'; return (0); } /* * Conversion functions between current locale dependent MBS and UTF-16BE. * strncat_from_utf16be() : UTF-16BE --> MBS * strncat_to_utf16be() : MBS --> UTF16BE */ #if defined(_WIN32) && !defined(__CYGWIN__) /* * Convert a UTF-16BE/LE string to current locale and copy the result. * Return -1 if conversion failes. */ static int win_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc, int be) { struct archive_string tmp; const char *u16; int ll; BOOL defchar; char *mbs; size_t mbs_size, b; int ret = 0; bytes &= ~1; if (archive_string_ensure(as, as->length + bytes +1) == NULL) return (-1); mbs = as->s + as->length; mbs_size = as->buffer_length - as->length -1; if (sc->to_cp == CP_C_LOCALE) { /* * "C" locale special process. */ u16 = _p; ll = 0; for (b = 0; b < bytes; b += 2) { uint16_t val; if (be) val = archive_be16dec(u16+b); else val = archive_le16dec(u16+b); if (val > 255) { *mbs++ = '?'; ret = -1; } else *mbs++ = (char)(val&0xff); ll++; } as->length += ll; as->s[as->length] = '\0'; return (ret); } archive_string_init(&tmp); if (be) { if (is_big_endian()) { u16 = _p; } else { if (archive_string_ensure(&tmp, bytes+2) == NULL) return (-1); memcpy(tmp.s, _p, bytes); for (b = 0; b < bytes; b += 2) { uint16_t val = archive_be16dec(tmp.s+b); archive_le16enc(tmp.s+b, val); } u16 = tmp.s; } } else { if (!is_big_endian()) { u16 = _p; } else { if (archive_string_ensure(&tmp, bytes+2) == NULL) return (-1); memcpy(tmp.s, _p, bytes); for (b = 0; b < bytes; b += 2) { uint16_t val = archive_le16dec(tmp.s+b); archive_be16enc(tmp.s+b, val); } u16 = tmp.s; } } do { defchar = 0; ll = WideCharToMultiByte(sc->to_cp, 0, (LPCWSTR)u16, bytes>>1, mbs, mbs_size, NULL, &defchar); if (ll == 0 && GetLastError() == ERROR_INSUFFICIENT_BUFFER) { /* Need more buffer for MBS. */ ll = WideCharToMultiByte(sc->to_cp, 0, (LPCWSTR)u16, bytes, NULL, 0, NULL, NULL); if (archive_string_ensure(as, ll +1) == NULL) return (-1); mbs = as->s + as->length; mbs_size = as->buffer_length - as->length -1; continue; } } while (0); archive_string_free(&tmp); as->length += ll; as->s[as->length] = '\0'; if (ll == 0 || defchar) ret = -1; return (ret); } static int win_strncat_from_utf16be(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (win_strncat_from_utf16(as, _p, bytes, sc, 1)); } static int win_strncat_from_utf16le(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (win_strncat_from_utf16(as, _p, bytes, sc, 0)); } static int is_big_endian(void) { uint16_t d = 1; return (archive_be16dec(&d) == 1); } /* * Convert a current locale string to UTF-16BE/LE and copy the result. * Return -1 if conversion failes. */ static int win_strncat_to_utf16(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc, int bigendian) { const char *s = (const char *)_p; char *u16; size_t count, avail; if (archive_string_ensure(as16, as16->length + (length + 1) * 2) == NULL) return (-1); u16 = as16->s + as16->length; avail = as16->buffer_length - 2; if (sc->from_cp == CP_C_LOCALE) { /* * "C" locale special process. */ count = 0; while (count < length && *s) { if (bigendian) archive_be16enc(u16, *s); else archive_le16enc(u16, *s); u16 += 2; s++; count++; } as16->length += count << 1; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; return (0); } do { count = MultiByteToWideChar(sc->from_cp, MB_PRECOMPOSED, s, length, (LPWSTR)u16, (int)avail>>1); if (count == 0 && GetLastError() == ERROR_INSUFFICIENT_BUFFER) { /* Need more buffer for UTF-16 string */ count = MultiByteToWideChar(sc->from_cp, MB_PRECOMPOSED, s, length, NULL, 0); if (archive_string_ensure(as16, (count +1) * 2) == NULL) return (-1); u16 = as16->s + as16->length; avail = as16->buffer_length - 2; continue; } } while (0); as16->length += count * 2; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; if (count == 0) return (-1); if (is_big_endian()) { if (!bigendian) { while (count > 0) { uint16_t v = archive_be16dec(u16); archive_le16enc(u16, v); u16 += 2; count--; } } } else { if (bigendian) { while (count > 0) { uint16_t v = archive_le16dec(u16); archive_be16enc(u16, v); u16 += 2; count--; } } } return (0); } static int win_strncat_to_utf16be(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (win_strncat_to_utf16(as16, _p, length, sc, 1)); } static int win_strncat_to_utf16le(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (win_strncat_to_utf16(as16, _p, length, sc, 0)); } #endif /* _WIN32 && !__CYGWIN__ */ /* * Do the best effort for conversions. * We cannot handle UTF-16BE character-set without such iconv, * but there is a chance if a string consists just ASCII code or * a current locale is UTF-8. */ /* * Convert a UTF-16BE string to current locale and copy the result. * Return -1 if conversion failes. */ static int best_effort_strncat_from_utf16(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc, int be) { const char *utf16 = (const char *)_p; char *mbs; uint32_t uc; int n, ret; (void)sc; /* UNUSED */ /* * Other case, we should do the best effort. * If all character are ASCII(<0x7f), we can convert it. * if not , we set a alternative character and return -1. */ ret = 0; if (archive_string_ensure(as, as->length + bytes +1) == NULL) return (-1); mbs = as->s + as->length; while ((n = utf16_to_unicode(&uc, utf16, bytes, be)) != 0) { if (n < 0) { n *= -1; ret = -1; } bytes -= n; utf16 += n; if (uc > 127) { /* We cannot handle it. */ *mbs++ = '?'; ret = -1; } else *mbs++ = (char)uc; } as->length = mbs - as->s; as->s[as->length] = '\0'; return (ret); } static int best_effort_strncat_from_utf16be(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 1)); } static int best_effort_strncat_from_utf16le(struct archive_string *as, const void *_p, size_t bytes, struct archive_string_conv *sc) { return (best_effort_strncat_from_utf16(as, _p, bytes, sc, 0)); } /* * Convert a current locale string to UTF-16BE/LE and copy the result. * Return -1 if conversion failes. */ static int best_effort_strncat_to_utf16(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc, int bigendian) { const char *s = (const char *)_p; char *utf16; size_t remaining; int ret; (void)sc; /* UNUSED */ /* * Other case, we should do the best effort. * If all character are ASCII(<0x7f), we can convert it. * if not , we set a alternative character and return -1. */ ret = 0; remaining = length; if (archive_string_ensure(as16, as16->length + (length + 1) * 2) == NULL) return (-1); utf16 = as16->s + as16->length; while (remaining--) { unsigned c = *s++; if (c > 127) { /* We cannot handle it. */ c = UNICODE_R_CHAR; ret = -1; } if (bigendian) archive_be16enc(utf16, c); else archive_le16enc(utf16, c); utf16 += 2; } as16->length = utf16 - as16->s; as16->s[as16->length] = 0; as16->s[as16->length+1] = 0; return (ret); } static int best_effort_strncat_to_utf16be(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (best_effort_strncat_to_utf16(as16, _p, length, sc, 1)); } static int best_effort_strncat_to_utf16le(struct archive_string *as16, const void *_p, size_t length, struct archive_string_conv *sc) { return (best_effort_strncat_to_utf16(as16, _p, length, sc, 0)); } /* * Multistring operations. */ void archive_mstring_clean(struct archive_mstring *aes) { archive_wstring_free(&(aes->aes_wcs)); archive_string_free(&(aes->aes_mbs)); archive_string_free(&(aes->aes_utf8)); archive_string_free(&(aes->aes_mbs_in_locale)); aes->aes_set = 0; } void archive_mstring_copy(struct archive_mstring *dest, struct archive_mstring *src) { dest->aes_set = src->aes_set; archive_string_copy(&(dest->aes_mbs), &(src->aes_mbs)); archive_string_copy(&(dest->aes_utf8), &(src->aes_utf8)); archive_wstring_copy(&(dest->aes_wcs), &(src->aes_wcs)); } int archive_mstring_get_utf8(struct archive *a, struct archive_mstring *aes, const char **p) { struct archive_string_conv *sc; int r; /* If we already have a UTF8 form, return that immediately. */ if (aes->aes_set & AES_SET_UTF8) { *p = aes->aes_utf8.s; return (0); } *p = NULL; if (aes->aes_set & AES_SET_MBS) { sc = archive_string_conversion_to_charset(a, "UTF-8", 1); if (sc == NULL) return (-1);/* Couldn't allocate memory for sc. */ r = archive_strncpy_in_locale(&(aes->aes_mbs), aes->aes_mbs.s, aes->aes_mbs.length, sc); if (a == NULL) free_sconv_object(sc); if (r == 0) { aes->aes_set |= AES_SET_UTF8; *p = aes->aes_utf8.s; return (0);/* success. */ } else return (-1);/* failure. */ } return (0);/* success. */ } int archive_mstring_get_mbs(struct archive *a, struct archive_mstring *aes, const char **p) { int r, ret = 0; (void)a; /* UNUSED */ /* If we already have an MBS form, return that immediately. */ if (aes->aes_set & AES_SET_MBS) { *p = aes->aes_mbs.s; return (ret); } *p = NULL; /* If there's a WCS form, try converting with the native locale. */ if (aes->aes_set & AES_SET_WCS) { archive_string_empty(&(aes->aes_mbs)); r = archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s, aes->aes_wcs.length); *p = aes->aes_mbs.s; if (r == 0) { aes->aes_set |= AES_SET_MBS; return (ret); } else ret = -1; } /* * Only a UTF-8 form cannot avail because its conversion already * failed at archive_mstring_update_utf8(). */ return (ret); } int archive_mstring_get_wcs(struct archive *a, struct archive_mstring *aes, const wchar_t **wp) { int r, ret = 0; (void)a;/* UNUSED */ /* Return WCS form if we already have it. */ if (aes->aes_set & AES_SET_WCS) { *wp = aes->aes_wcs.s; return (ret); } *wp = NULL; /* Try converting MBS to WCS using native locale. */ if (aes->aes_set & AES_SET_MBS) { archive_wstring_empty(&(aes->aes_wcs)); r = archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, aes->aes_mbs.length); if (r == 0) { aes->aes_set |= AES_SET_WCS; *wp = aes->aes_wcs.s; } else ret = -1;/* failure. */ } return (ret); } int archive_mstring_get_mbs_l(struct archive_mstring *aes, const char **p, size_t *length, struct archive_string_conv *sc) { int r, ret = 0; #if defined(_WIN32) && !defined(__CYGWIN__) /* * Internationalization programing on Windows must use Wide * characters because Windows platform cannot make locale UTF-8. */ if (sc != NULL && (aes->aes_set & AES_SET_WCS) != 0) { archive_string_empty(&(aes->aes_mbs_in_locale)); r = archive_string_append_from_wcs_in_codepage( &(aes->aes_mbs_in_locale), aes->aes_wcs.s, aes->aes_wcs.length, sc); if (r == 0) { *p = aes->aes_mbs_in_locale.s; if (length != NULL) *length = aes->aes_mbs_in_locale.length; return (0); } else if (errno == ENOMEM) return (-1); else ret = -1; } #endif /* If there is not an MBS form but is a WCS form, try converting * with the native locale to be used for translating it to specified * character-set. */ if ((aes->aes_set & AES_SET_MBS) == 0 && (aes->aes_set & AES_SET_WCS) != 0) { archive_string_empty(&(aes->aes_mbs)); r = archive_string_append_from_wcs(&(aes->aes_mbs), aes->aes_wcs.s, aes->aes_wcs.length); if (r == 0) aes->aes_set |= AES_SET_MBS; else if (errno == ENOMEM) return (-1); else ret = -1; } /* If we already have an MBS form, use it to be translated to * specified character-set. */ if (aes->aes_set & AES_SET_MBS) { if (sc == NULL) { /* Conversion is unneeded. */ *p = aes->aes_mbs.s; if (length != NULL) *length = aes->aes_mbs.length; return (0); } ret = archive_strncpy_in_locale(&(aes->aes_mbs_in_locale), aes->aes_mbs.s, aes->aes_mbs.length, sc); *p = aes->aes_mbs_in_locale.s; if (length != NULL) *length = aes->aes_mbs_in_locale.length; } else { *p = NULL; if (length != NULL) *length = 0; } return (ret); } int archive_mstring_copy_mbs(struct archive_mstring *aes, const char *mbs) { if (mbs == NULL) { aes->aes_set = 0; return (0); } return (archive_mstring_copy_mbs_len(aes, mbs, strlen(mbs))); } int archive_mstring_copy_mbs_len(struct archive_mstring *aes, const char *mbs, size_t len) { if (mbs == NULL) { aes->aes_set = 0; return (0); } aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ archive_strncpy(&(aes->aes_mbs), mbs, len); archive_string_empty(&(aes->aes_utf8)); archive_wstring_empty(&(aes->aes_wcs)); return (0); } int archive_mstring_copy_wcs(struct archive_mstring *aes, const wchar_t *wcs) { return archive_mstring_copy_wcs_len(aes, wcs, wcs == NULL ? 0 : wcslen(wcs)); } int archive_mstring_copy_wcs_len(struct archive_mstring *aes, const wchar_t *wcs, size_t len) { if (wcs == NULL) { aes->aes_set = 0; } aes->aes_set = AES_SET_WCS; /* Only WCS form set. */ archive_string_empty(&(aes->aes_mbs)); archive_string_empty(&(aes->aes_utf8)); archive_wstrncpy(&(aes->aes_wcs), wcs, len); return (0); } int archive_mstring_copy_mbs_len_l(struct archive_mstring *aes, const char *mbs, size_t len, struct archive_string_conv *sc) { int r; if (mbs == NULL) { aes->aes_set = 0; return (0); } archive_string_empty(&(aes->aes_mbs)); archive_wstring_empty(&(aes->aes_wcs)); archive_string_empty(&(aes->aes_utf8)); #if defined(_WIN32) && !defined(__CYGWIN__) /* * Internationalization programing on Windows must use Wide * characters because Windows platform cannot make locale UTF-8. */ if (sc == NULL) { if (archive_string_append(&(aes->aes_mbs), mbs, mbsnbytes(mbs, len)) == NULL) { aes->aes_set = 0; r = -1; } else { aes->aes_set = AES_SET_MBS; r = 0; } #if defined(HAVE_ICONV) } else if (sc != NULL && sc->cd_w != (iconv_t)-1) { /* * This case happens only when MultiByteToWideChar() cannot * handle sc->from_cp, and we have to iconv in order to * translate character-set to wchar_t,UTF-16. */ iconv_t cd = sc->cd; unsigned from_cp; int flag; /* * Translate multi-bytes from some character-set to UTF-8. */ sc->cd = sc->cd_w; r = archive_strncpy_in_locale(&(aes->aes_utf8), mbs, len, sc); sc->cd = cd; if (r != 0) { aes->aes_set = 0; return (r); } aes->aes_set = AES_SET_UTF8; /* * Append the UTF-8 string into wstring. */ flag = sc->flag; sc->flag &= ~(SCONV_NORMALIZATION_C | SCONV_TO_UTF16| SCONV_FROM_UTF16); from_cp = sc->from_cp; sc->from_cp = CP_UTF8; r = archive_wstring_append_from_mbs_in_codepage(&(aes->aes_wcs), aes->aes_utf8.s, aes->aes_utf8.length, sc); sc->flag = flag; sc->from_cp = from_cp; if (r == 0) aes->aes_set |= AES_SET_WCS; #endif } else { r = archive_wstring_append_from_mbs_in_codepage( &(aes->aes_wcs), mbs, len, sc); if (r == 0) aes->aes_set = AES_SET_WCS; else aes->aes_set = 0; } #else r = archive_strncpy_in_locale(&(aes->aes_mbs), mbs, len, sc); if (r == 0) aes->aes_set = AES_SET_MBS; /* Only MBS form is set now. */ else aes->aes_set = 0; #endif return (r); } /* * The 'update' form tries to proactively update all forms of * this string (WCS and MBS) and returns an error if any of * them fail. This is used by the 'pax' handler, for instance, * to detect and report character-conversion failures early while * still allowing clients to get potentially useful values from * the more tolerant lazy conversions. (get_mbs and get_wcs will * strive to give the user something useful, so you can get hopefully * usable values even if some of the character conversions are failing.) */ int archive_mstring_update_utf8(struct archive *a, struct archive_mstring *aes, const char *utf8) { struct archive_string_conv *sc; int r; if (utf8 == NULL) { aes->aes_set = 0; return (0); /* Succeeded in clearing everything. */ } /* Save the UTF8 string. */ archive_strcpy(&(aes->aes_utf8), utf8); /* Empty the mbs and wcs strings. */ archive_string_empty(&(aes->aes_mbs)); archive_wstring_empty(&(aes->aes_wcs)); aes->aes_set = AES_SET_UTF8; /* Only UTF8 is set now. */ /* Try converting UTF-8 to MBS, return false on failure. */ sc = archive_string_conversion_from_charset(a, "UTF-8", 1); if (sc == NULL) return (-1);/* Couldn't allocate memory for sc. */ r = archive_strcpy_in_locale(&(aes->aes_mbs), utf8, sc); if (a == NULL) free_sconv_object(sc); if (r != 0) return (-1); aes->aes_set = AES_SET_UTF8 | AES_SET_MBS; /* Both UTF8 and MBS set. */ /* Try converting MBS to WCS, return false on failure. */ if (archive_wstring_append_from_mbs(&(aes->aes_wcs), aes->aes_mbs.s, aes->aes_mbs.length)) return (-1); aes->aes_set = AES_SET_UTF8 | AES_SET_WCS | AES_SET_MBS; /* All conversions succeeded. */ return (0); } Index: vendor/libarchive/dist/libarchive/archive_write_disk_posix.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_write_disk_posix.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_write_disk_posix.c (revision 231297) @@ -1,2847 +1,2856 @@ /*- * Copyright (c) 2003-2010 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer * in this position and unchanged. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" __FBSDID("$FreeBSD$"); #if !defined(_WIN32) || defined(__CYGWIN__) #ifdef HAVE_SYS_TYPES_H #include #endif #ifdef HAVE_SYS_ACL_H #include #endif #ifdef HAVE_SYS_EXTATTR_H #include #endif #ifdef HAVE_SYS_XATTR_H #include #endif #ifdef HAVE_SYS_EA_H #include #endif #ifdef HAVE_ATTR_XATTR_H #include #endif #ifdef HAVE_SYS_IOCTL_H #include #endif #ifdef HAVE_SYS_STAT_H #include #endif #ifdef HAVE_SYS_TIME_H #include #endif #ifdef HAVE_SYS_UTIME_H #include #endif #ifdef HAVE_COPYFILE_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_GRP_H #include #endif #ifdef HAVE_LANGINFO_H #include #endif #ifdef HAVE_LINUX_FS_H #include /* for Linux file flags */ #endif /* * Some Linux distributions have both linux/ext2_fs.h and ext2fs/ext2_fs.h. * As the include guards don't agree, the order of include is important. */ #ifdef HAVE_LINUX_EXT2_FS_H #include /* for Linux file flags */ #endif #if defined(HAVE_EXT2FS_EXT2_FS_H) && !defined(__CYGWIN__) #include /* Linux file flags, broken on Cygwin */ #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_PWD_H #include #endif #include #ifdef HAVE_STDLIB_H #include #endif #ifdef HAVE_STRING_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_UTIME_H #include #endif #ifdef F_GETTIMES /* Tru64 specific */ #include #endif #if __APPLE__ #include #if TARGET_OS_MAC && !TARGET_OS_EMBEDDED && HAVE_QUARANTINE_H #include #define HAVE_QUARANTINE 1 #endif #endif /* TODO: Support Mac OS 'quarantine' feature. This is really just a * standard tag to mark files that have been downloaded as "tainted". * On Mac OS, we should mark the extracted files as tainted if the * archive being read was tainted. Windows has a similar feature; we * should investigate ways to support this generically. */ #include "archive.h" #include "archive_acl_private.h" #include "archive_string.h" #include "archive_entry.h" #include "archive_private.h" #ifndef O_BINARY #define O_BINARY 0 #endif struct fixup_entry { struct fixup_entry *next; struct archive_acl acl; mode_t mode; int64_t atime; int64_t birthtime; int64_t mtime; int64_t ctime; unsigned long atime_nanos; unsigned long birthtime_nanos; unsigned long mtime_nanos; unsigned long ctime_nanos; unsigned long fflags_set; size_t mac_metadata_size; void *mac_metadata; int fixup; /* bitmask of what needs fixing */ char *name; }; /* * We use a bitmask to track which operations remain to be done for * this file. In particular, this helps us avoid unnecessary * operations when it's possible to take care of one step as a * side-effect of another. For example, mkdir() can specify the mode * for the newly-created object but symlink() cannot. This means we * can skip chmod() if mkdir() succeeded, but we must explicitly * chmod() if we're trying to create a directory that already exists * (mkdir() failed) or if we're restoring a symlink. Similarly, we * need to verify UID/GID before trying to restore SUID/SGID bits; * that verification can occur explicitly through a stat() call or * implicitly because of a successful chown() call. */ #define TODO_MODE_FORCE 0x40000000 #define TODO_MODE_BASE 0x20000000 #define TODO_SUID 0x10000000 #define TODO_SUID_CHECK 0x08000000 #define TODO_SGID 0x04000000 #define TODO_SGID_CHECK 0x02000000 #define TODO_MODE (TODO_MODE_BASE|TODO_SUID|TODO_SGID) #define TODO_TIMES ARCHIVE_EXTRACT_TIME #define TODO_OWNER ARCHIVE_EXTRACT_OWNER #define TODO_FFLAGS ARCHIVE_EXTRACT_FFLAGS #define TODO_ACLS ARCHIVE_EXTRACT_ACL #define TODO_XATTR ARCHIVE_EXTRACT_XATTR #define TODO_MAC_METADATA ARCHIVE_EXTRACT_MAC_METADATA struct archive_write_disk { struct archive archive; mode_t user_umask; struct fixup_entry *fixup_list; struct fixup_entry *current_fixup; int64_t user_uid; int skip_file_set; dev_t skip_file_dev; ino_t skip_file_ino; time_t start_time; int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid); void (*cleanup_gid)(void *private); void *lookup_gid_data; int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid); void (*cleanup_uid)(void *private); void *lookup_uid_data; /* * Full path of last file to satisfy symlink checks. */ struct archive_string path_safe; /* * Cached stat data from disk for the current entry. * If this is valid, pst points to st. Otherwise, * pst is null. */ struct stat st; struct stat *pst; /* Information about the object being restored right now. */ struct archive_entry *entry; /* Entry being extracted. */ char *name; /* Name of entry, possibly edited. */ struct archive_string _name_data; /* backing store for 'name' */ /* Tasks remaining for this object. */ int todo; /* Tasks deferred until end-of-archive. */ int deferred; /* Options requested by the client. */ int flags; /* Handle for the file we're restoring. */ int fd; /* Current offset for writing data to the file. */ int64_t offset; /* Last offset actually written to disk. */ int64_t fd_offset; /* Total bytes actually written to files. */ int64_t total_bytes_written; /* Maximum size of file, -1 if unknown. */ int64_t filesize; /* Dir we were in before this restore; only for deep paths. */ int restore_pwd; /* Mode we should use for this entry; affected by _PERM and umask. */ mode_t mode; /* UID/GID to use in restoring this entry. */ int64_t uid; int64_t gid; }; /* * Default mode for dirs created automatically (will be modified by umask). * Note that POSIX specifies 0777 for implicitly-created dirs, "modified * by the process' file creation mask." */ #define DEFAULT_DIR_MODE 0777 /* * Dir modes are restored in two steps: During the extraction, the permissions * in the archive are modified to match the following limits. During * the post-extract fixup pass, the permissions from the archive are * applied. */ #define MINIMUM_DIR_MODE 0700 #define MAXIMUM_DIR_MODE 0775 static int check_symlinks(struct archive_write_disk *); static int create_filesystem_object(struct archive_write_disk *); static struct fixup_entry *current_fixup(struct archive_write_disk *, const char *pathname); #if defined(HAVE_FCHDIR) && defined(PATH_MAX) static void edit_deep_directories(struct archive_write_disk *ad); #endif static int cleanup_pathname(struct archive_write_disk *); static int create_dir(struct archive_write_disk *, char *); static int create_parent_dir(struct archive_write_disk *, char *); static int older(struct stat *, struct archive_entry *); static int restore_entry(struct archive_write_disk *); #ifdef HAVE_POSIX_ACL static int set_acl(struct archive_write_disk *, int fd, const char *, struct archive_acl *, acl_type_t, int archive_entry_acl_type, const char *tn); #endif static int set_acls(struct archive_write_disk *, int fd, const char *, struct archive_acl *); static int set_mac_metadata(struct archive_write_disk *, const char *, const void *, size_t); static int set_xattrs(struct archive_write_disk *); static int set_fflags(struct archive_write_disk *); static int set_fflags_platform(struct archive_write_disk *, int fd, const char *name, mode_t mode, unsigned long fflags_set, unsigned long fflags_clear); static int set_ownership(struct archive_write_disk *); static int set_mode(struct archive_write_disk *, int mode); static int set_time(int, int, const char *, time_t, long, time_t, long); static int set_times(struct archive_write_disk *, int, int, const char *, time_t, long, time_t, long, time_t, long, time_t, long); static int set_times_from_entry(struct archive_write_disk *); static struct fixup_entry *sort_dir_list(struct fixup_entry *p); static ssize_t write_data_block(struct archive_write_disk *, const char *, size_t); static struct archive_vtable *archive_write_disk_vtable(void); static int _archive_write_disk_close(struct archive *); static int _archive_write_disk_free(struct archive *); static int _archive_write_disk_header(struct archive *, struct archive_entry *); static int64_t _archive_write_disk_filter_bytes(struct archive *, int); static int _archive_write_disk_finish_entry(struct archive *); static ssize_t _archive_write_disk_data(struct archive *, const void *, size_t); static ssize_t _archive_write_disk_data_block(struct archive *, const void *, size_t, int64_t); static int lazy_stat(struct archive_write_disk *a) { if (a->pst != NULL) { /* Already have stat() data available. */ return (ARCHIVE_OK); } #ifdef HAVE_FSTAT if (a->fd >= 0 && fstat(a->fd, &a->st) == 0) { a->pst = &a->st; return (ARCHIVE_OK); } #endif /* * XXX At this point, symlinks should not be hit, otherwise * XXX a race occurred. Do we want to check explicitly for that? */ if (lstat(a->name, &a->st) == 0) { a->pst = &a->st; return (ARCHIVE_OK); } archive_set_error(&a->archive, errno, "Couldn't stat file"); return (ARCHIVE_WARN); } static struct archive_vtable * archive_write_disk_vtable(void) { static struct archive_vtable av; static int inited = 0; if (!inited) { av.archive_close = _archive_write_disk_close; av.archive_filter_bytes = _archive_write_disk_filter_bytes; av.archive_free = _archive_write_disk_free; av.archive_write_header = _archive_write_disk_header; av.archive_write_finish_entry = _archive_write_disk_finish_entry; av.archive_write_data = _archive_write_disk_data; av.archive_write_data_block = _archive_write_disk_data_block; inited = 1; } return (&av); } static int64_t _archive_write_disk_filter_bytes(struct archive *_a, int n) { struct archive_write_disk *a = (struct archive_write_disk *)_a; (void)n; /* UNUSED */ if (n == -1 || n == 0) return (a->total_bytes_written); return (-1); } int archive_write_disk_set_options(struct archive *_a, int flags) { struct archive_write_disk *a = (struct archive_write_disk *)_a; a->flags = flags; return (ARCHIVE_OK); } /* * Extract this entry to disk. * * TODO: Validate hardlinks. According to the standards, we're * supposed to check each extracted hardlink and squawk if it refers * to a file that we didn't restore. I'm not entirely convinced this * is a good idea, but more importantly: Is there any way to validate * hardlinks without keeping a complete list of filenames from the * entire archive?? Ugh. * */ static int _archive_write_disk_header(struct archive *_a, struct archive_entry *entry) { struct archive_write_disk *a = (struct archive_write_disk *)_a; struct fixup_entry *fe; int ret, r; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_disk_header"); archive_clear_error(&a->archive); if (a->archive.state & ARCHIVE_STATE_DATA) { r = _archive_write_disk_finish_entry(&a->archive); if (r == ARCHIVE_FATAL) return (r); } /* Set up for this particular entry. */ a->pst = NULL; a->current_fixup = NULL; a->deferred = 0; if (a->entry) { archive_entry_free(a->entry); a->entry = NULL; } a->entry = archive_entry_clone(entry); a->fd = -1; a->fd_offset = 0; a->offset = 0; a->restore_pwd = -1; a->uid = a->user_uid; a->mode = archive_entry_mode(a->entry); if (archive_entry_size_is_set(a->entry)) a->filesize = archive_entry_size(a->entry); else a->filesize = -1; archive_strcpy(&(a->_name_data), archive_entry_pathname(a->entry)); a->name = a->_name_data.s; archive_clear_error(&a->archive); /* * Clean up the requested path. This is necessary for correct * dir restores; the dir restore logic otherwise gets messed * up by nonsense like "dir/.". */ ret = cleanup_pathname(a); if (ret != ARCHIVE_OK) return (ret); /* * Query the umask so we get predictable mode settings. * This gets done on every call to _write_header in case the * user edits their umask during the extraction for some * reason. */ umask(a->user_umask = umask(0)); /* Figure out what we need to do for this entry. */ a->todo = TODO_MODE_BASE; if (a->flags & ARCHIVE_EXTRACT_PERM) { a->todo |= TODO_MODE_FORCE; /* Be pushy about permissions. */ /* * SGID requires an extra "check" step because we * cannot easily predict the GID that the system will * assign. (Different systems assign GIDs to files * based on a variety of criteria, including process * credentials and the gid of the enclosing * directory.) We can only restore the SGID bit if * the file has the right GID, and we only know the * GID if we either set it (see set_ownership) or if * we've actually called stat() on the file after it * was restored. Since there are several places at * which we might verify the GID, we need a TODO bit * to keep track. */ if (a->mode & S_ISGID) a->todo |= TODO_SGID | TODO_SGID_CHECK; /* * Verifying the SUID is simpler, but can still be * done in multiple ways, hence the separate "check" bit. */ if (a->mode & S_ISUID) a->todo |= TODO_SUID | TODO_SUID_CHECK; } else { /* * User didn't request full permissions, so don't * restore SUID, SGID bits and obey umask. */ a->mode &= ~S_ISUID; a->mode &= ~S_ISGID; a->mode &= ~S_ISVTX; a->mode &= ~a->user_umask; } if (a->flags & ARCHIVE_EXTRACT_OWNER) a->todo |= TODO_OWNER; if (a->flags & ARCHIVE_EXTRACT_TIME) a->todo |= TODO_TIMES; if (a->flags & ARCHIVE_EXTRACT_ACL) { if (archive_entry_filetype(a->entry) == AE_IFDIR) a->deferred |= TODO_ACLS; else a->todo |= TODO_ACLS; } if (a->flags & ARCHIVE_EXTRACT_MAC_METADATA) { if (archive_entry_filetype(a->entry) == AE_IFDIR) a->deferred |= TODO_MAC_METADATA; else a->todo |= TODO_MAC_METADATA; } if (a->flags & ARCHIVE_EXTRACT_XATTR) a->todo |= TODO_XATTR; if (a->flags & ARCHIVE_EXTRACT_FFLAGS) a->todo |= TODO_FFLAGS; if (a->flags & ARCHIVE_EXTRACT_SECURE_SYMLINKS) { ret = check_symlinks(a); if (ret != ARCHIVE_OK) return (ret); } #if defined(HAVE_FCHDIR) && defined(PATH_MAX) /* If path exceeds PATH_MAX, shorten the path. */ edit_deep_directories(a); #endif ret = restore_entry(a); /* * TODO: There are rumours that some extended attributes must * be restored before file data is written. If this is true, * then we either need to write all extended attributes both * before and after restoring the data, or find some rule for * determining which must go first and which last. Due to the * many ways people are using xattrs, this may prove to be an * intractable problem. */ #ifdef HAVE_FCHDIR /* If we changed directory above, restore it here. */ if (a->restore_pwd >= 0) { r = fchdir(a->restore_pwd); if (r != 0) { archive_set_error(&a->archive, errno, "chdir() failure"); ret = ARCHIVE_FATAL; } close(a->restore_pwd); a->restore_pwd = -1; } #endif /* * Fixup uses the unedited pathname from archive_entry_pathname(), * because it is relative to the base dir and the edited path * might be relative to some intermediate dir as a result of the * deep restore logic. */ if (a->deferred & TODO_MODE) { fe = current_fixup(a, archive_entry_pathname(entry)); fe->fixup |= TODO_MODE_BASE; fe->mode = a->mode; } if ((a->deferred & TODO_TIMES) && (archive_entry_mtime_is_set(entry) || archive_entry_atime_is_set(entry))) { fe = current_fixup(a, archive_entry_pathname(entry)); fe->mode = a->mode; fe->fixup |= TODO_TIMES; if (archive_entry_atime_is_set(entry)) { fe->atime = archive_entry_atime(entry); fe->atime_nanos = archive_entry_atime_nsec(entry); } else { /* If atime is unset, use start time. */ fe->atime = a->start_time; fe->atime_nanos = 0; } if (archive_entry_mtime_is_set(entry)) { fe->mtime = archive_entry_mtime(entry); fe->mtime_nanos = archive_entry_mtime_nsec(entry); } else { /* If mtime is unset, use start time. */ fe->mtime = a->start_time; fe->mtime_nanos = 0; } if (archive_entry_birthtime_is_set(entry)) { fe->birthtime = archive_entry_birthtime(entry); fe->birthtime_nanos = archive_entry_birthtime_nsec(entry); } else { /* If birthtime is unset, use mtime. */ fe->birthtime = fe->mtime; fe->birthtime_nanos = fe->mtime_nanos; } } if (a->deferred & TODO_ACLS) { fe = current_fixup(a, archive_entry_pathname(entry)); archive_acl_copy(&fe->acl, archive_entry_acl(entry)); } if (a->deferred & TODO_MAC_METADATA) { const void *metadata; size_t metadata_size; metadata = archive_entry_mac_metadata(a->entry, &metadata_size); if (metadata != NULL && metadata_size > 0) { fe = current_fixup(a, archive_entry_pathname(entry)); fe->mac_metadata = malloc(metadata_size); if (fe->mac_metadata != NULL) { memcpy(fe->mac_metadata, metadata, metadata_size); fe->mac_metadata_size = metadata_size; fe->fixup |= TODO_MAC_METADATA; } } } if (a->deferred & TODO_FFLAGS) { fe = current_fixup(a, archive_entry_pathname(entry)); fe->fixup |= TODO_FFLAGS; /* TODO: Complete this.. defer fflags from below. */ } /* We've created the object and are ready to pour data into it. */ if (ret >= ARCHIVE_WARN) a->archive.state = ARCHIVE_STATE_DATA; /* * If it's not open, tell our client not to try writing. * In particular, dirs, links, etc, don't get written to. */ if (a->fd < 0) { archive_entry_set_size(entry, 0); a->filesize = 0; } return (ret); } int archive_write_disk_set_skip_file(struct archive *_a, int64_t d, int64_t i) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_skip_file"); a->skip_file_set = 1; a->skip_file_dev = d; a->skip_file_ino = i; return (ARCHIVE_OK); } static ssize_t write_data_block(struct archive_write_disk *a, const char *buff, size_t size) { uint64_t start_size = size; ssize_t bytes_written = 0; ssize_t block_size = 0, bytes_to_write; if (size == 0) return (ARCHIVE_OK); if (a->filesize == 0 || a->fd < 0) { archive_set_error(&a->archive, 0, "Attempt to write to an empty file"); return (ARCHIVE_WARN); } if (a->flags & ARCHIVE_EXTRACT_SPARSE) { #if HAVE_STRUCT_STAT_ST_BLKSIZE int r; if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); block_size = a->pst->st_blksize; #else /* XXX TODO XXX Is there a more appropriate choice here ? */ /* This needn't match the filesystem allocation size. */ block_size = 16*1024; #endif } /* If this write would run beyond the file size, truncate it. */ if (a->filesize >= 0 && (int64_t)(a->offset + size) > a->filesize) start_size = size = (size_t)(a->filesize - a->offset); /* Write the data. */ while (size > 0) { if (block_size == 0) { bytes_to_write = size; } else { /* We're sparsifying the file. */ const char *p, *end; int64_t block_end; /* Skip leading zero bytes. */ for (p = buff, end = buff + size; p < end; ++p) { if (*p != '\0') break; } a->offset += p - buff; size -= p - buff; buff = p; if (size == 0) break; /* Calculate next block boundary after offset. */ block_end = (a->offset / block_size + 1) * block_size; /* If the adjusted write would cross block boundary, * truncate it to the block boundary. */ bytes_to_write = size; if (a->offset + bytes_to_write > block_end) bytes_to_write = block_end - a->offset; } /* Seek if necessary to the specified offset. */ if (a->offset != a->fd_offset) { if (lseek(a->fd, a->offset, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek failed"); return (ARCHIVE_FATAL); } a->fd_offset = a->offset; } bytes_written = write(a->fd, buff, bytes_to_write); if (bytes_written < 0) { archive_set_error(&a->archive, errno, "Write failed"); return (ARCHIVE_WARN); } buff += bytes_written; size -= bytes_written; a->total_bytes_written += bytes_written; a->offset += bytes_written; a->fd_offset = a->offset; } return (start_size - size); } static ssize_t _archive_write_disk_data_block(struct archive *_a, const void *buff, size_t size, int64_t offset) { struct archive_write_disk *a = (struct archive_write_disk *)_a; ssize_t r; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_write_data_block"); a->offset = offset; r = write_data_block(a, buff, size); if (r < ARCHIVE_OK) return (r); if ((size_t)r < size) { archive_set_error(&a->archive, 0, "Write request too large"); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } static ssize_t _archive_write_disk_data(struct archive *_a, const void *buff, size_t size) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_DATA, "archive_write_data"); return (write_data_block(a, buff, size)); } static int _archive_write_disk_finish_entry(struct archive *_a) { struct archive_write_disk *a = (struct archive_write_disk *)_a; int ret = ARCHIVE_OK; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_finish_entry"); if (a->archive.state & ARCHIVE_STATE_HEADER) return (ARCHIVE_OK); archive_clear_error(&a->archive); /* Pad or truncate file to the right size. */ if (a->fd < 0) { /* There's no file. */ } else if (a->filesize < 0) { /* File size is unknown, so we can't set the size. */ } else if (a->fd_offset == a->filesize) { /* Last write ended at exactly the filesize; we're done. */ /* Hopefully, this is the common case. */ } else { #if HAVE_FTRUNCATE if (ftruncate(a->fd, a->filesize) == -1 && a->filesize == 0) { archive_set_error(&a->archive, errno, "File size could not be restored"); return (ARCHIVE_FAILED); } #endif /* * Not all platforms implement the XSI option to * extend files via ftruncate. Stat() the file again * to see what happened. */ a->pst = NULL; if ((ret = lazy_stat(a)) != ARCHIVE_OK) return (ret); /* We can use lseek()/write() to extend the file if * ftruncate didn't work or isn't available. */ if (a->st.st_size < a->filesize) { const char nul = '\0'; if (lseek(a->fd, a->filesize - 1, SEEK_SET) < 0) { archive_set_error(&a->archive, errno, "Seek failed"); return (ARCHIVE_FATAL); } if (write(a->fd, &nul, 1) < 0) { archive_set_error(&a->archive, errno, "Write to restore size failed"); return (ARCHIVE_FATAL); } a->pst = NULL; } } /* Restore metadata. */ /* * Look up the "real" UID only if we're going to need it. * TODO: the TODO_SGID condition can be dropped here, can't it? */ if (a->todo & (TODO_OWNER | TODO_SUID | TODO_SGID)) { a->uid = archive_write_disk_uid(&a->archive, archive_entry_uname(a->entry), archive_entry_uid(a->entry)); } /* Look up the "real" GID only if we're going to need it. */ /* TODO: the TODO_SUID condition can be dropped here, can't it? */ if (a->todo & (TODO_OWNER | TODO_SGID | TODO_SUID)) { a->gid = archive_write_disk_gid(&a->archive, archive_entry_gname(a->entry), archive_entry_gid(a->entry)); } /* * Restore ownership before set_mode tries to restore suid/sgid * bits. If we set the owner, we know what it is and can skip * a stat() call to examine the ownership of the file on disk. */ if (a->todo & TODO_OWNER) ret = set_ownership(a); /* * set_mode must precede ACLs on systems such as Solaris and * FreeBSD where setting the mode implicitly clears extended ACLs */ if (a->todo & TODO_MODE) { int r2 = set_mode(a, a->mode); if (r2 < ret) ret = r2; } /* * Security-related extended attributes (such as * security.capability on Linux) have to be restored last, * since they're implicitly removed by other file changes. */ if (a->todo & TODO_XATTR) { int r2 = set_xattrs(a); if (r2 < ret) ret = r2; } /* * Some flags prevent file modification; they must be restored after * file contents are written. */ if (a->todo & TODO_FFLAGS) { int r2 = set_fflags(a); if (r2 < ret) ret = r2; } /* * Time must follow most other metadata; * otherwise atime will get changed. */ if (a->todo & TODO_TIMES) { int r2 = set_times_from_entry(a); if (r2 < ret) ret = r2; } /* * Mac extended metadata includes ACLs. */ if (a->todo & TODO_MAC_METADATA) { const void *metadata; size_t metadata_size; metadata = archive_entry_mac_metadata(a->entry, &metadata_size); if (metadata != NULL && metadata_size > 0) { int r2 = set_mac_metadata(a, archive_entry_pathname(a->entry), metadata, metadata_size); if (r2 < ret) ret = r2; } } /* * ACLs must be restored after timestamps because there are * ACLs that prevent attribute changes (including time). */ if (a->todo & TODO_ACLS) { int r2 = set_acls(a, a->fd, archive_entry_pathname(a->entry), archive_entry_acl(a->entry)); if (r2 < ret) ret = r2; } /* If there's an fd, we can close it now. */ if (a->fd >= 0) { close(a->fd); a->fd = -1; } /* If there's an entry, we can release it now. */ if (a->entry) { archive_entry_free(a->entry); a->entry = NULL; } a->archive.state = ARCHIVE_STATE_HEADER; return (ret); } int archive_write_disk_set_group_lookup(struct archive *_a, void *private_data, int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid), void (*cleanup_gid)(void *private)) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_group_lookup"); if (a->cleanup_gid != NULL && a->lookup_gid_data != NULL) (a->cleanup_gid)(a->lookup_gid_data); a->lookup_gid = lookup_gid; a->cleanup_gid = cleanup_gid; a->lookup_gid_data = private_data; return (ARCHIVE_OK); } int archive_write_disk_set_user_lookup(struct archive *_a, void *private_data, int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid), void (*cleanup_uid)(void *private)) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_set_user_lookup"); if (a->cleanup_uid != NULL && a->lookup_uid_data != NULL) (a->cleanup_uid)(a->lookup_uid_data); a->lookup_uid = lookup_uid; a->cleanup_uid = cleanup_uid; a->lookup_uid_data = private_data; return (ARCHIVE_OK); } int64_t archive_write_disk_gid(struct archive *_a, const char *name, int64_t id) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_gid"); if (a->lookup_gid) return (a->lookup_gid)(a->lookup_gid_data, name, id); return (id); } int64_t archive_write_disk_uid(struct archive *_a, const char *name, int64_t id) { struct archive_write_disk *a = (struct archive_write_disk *)_a; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY, "archive_write_disk_uid"); if (a->lookup_uid) return (a->lookup_uid)(a->lookup_uid_data, name, id); return (id); } /* * Create a new archive_write_disk object and initialize it with global state. */ struct archive * archive_write_disk_new(void) { struct archive_write_disk *a; a = (struct archive_write_disk *)malloc(sizeof(*a)); if (a == NULL) return (NULL); memset(a, 0, sizeof(*a)); a->archive.magic = ARCHIVE_WRITE_DISK_MAGIC; /* We're ready to write a header immediately. */ a->archive.state = ARCHIVE_STATE_HEADER; a->archive.vtable = archive_write_disk_vtable(); a->start_time = time(NULL); /* Query and restore the umask. */ umask(a->user_umask = umask(0)); #ifdef HAVE_GETEUID a->user_uid = geteuid(); #endif /* HAVE_GETEUID */ if (archive_string_ensure(&a->path_safe, 512) == NULL) { free(a); return (NULL); } return (&a->archive); } /* * If pathname is longer than PATH_MAX, chdir to a suitable * intermediate dir and edit the path down to a shorter suffix. Note * that this routine never returns an error; if the chdir() attempt * fails for any reason, we just go ahead with the long pathname. The * object creation is likely to fail, but any error will get handled * at that time. */ #if defined(HAVE_FCHDIR) && defined(PATH_MAX) static void edit_deep_directories(struct archive_write_disk *a) { int ret; char *tail = a->name; /* If path is short, avoid the open() below. */ if (strlen(tail) <= PATH_MAX) return; /* Try to record our starting dir. */ a->restore_pwd = open(".", O_RDONLY | O_BINARY); if (a->restore_pwd < 0) return; /* As long as the path is too long... */ while (strlen(tail) > PATH_MAX) { /* Locate a dir prefix shorter than PATH_MAX. */ tail += PATH_MAX - 8; while (tail > a->name && *tail != '/') tail--; /* Exit if we find a too-long path component. */ if (tail <= a->name) return; /* Create the intermediate dir and chdir to it. */ *tail = '\0'; /* Terminate dir portion */ ret = create_dir(a, a->name); if (ret == ARCHIVE_OK && chdir(a->name) != 0) ret = ARCHIVE_FAILED; *tail = '/'; /* Restore the / we removed. */ if (ret != ARCHIVE_OK) return; tail++; /* The chdir() succeeded; we've now shortened the path. */ a->name = tail; } return; } #endif /* * The main restore function. */ static int restore_entry(struct archive_write_disk *a) { int ret = ARCHIVE_OK, en; if (a->flags & ARCHIVE_EXTRACT_UNLINK && !S_ISDIR(a->mode)) { /* * TODO: Fix this. Apparently, there are platforms * that still allow root to hose the entire filesystem * by unlinking a dir. The S_ISDIR() test above * prevents us from using unlink() here if the new * object is a dir, but that doesn't mean the old * object isn't a dir. */ if (unlink(a->name) == 0) { /* We removed it, reset cached stat. */ a->pst = NULL; } else if (errno == ENOENT) { /* File didn't exist, that's just as good. */ } else if (rmdir(a->name) == 0) { /* It was a dir, but now it's gone. */ a->pst = NULL; } else { /* We tried, but couldn't get rid of it. */ archive_set_error(&a->archive, errno, "Could not unlink"); return(ARCHIVE_FAILED); } } /* Try creating it first; if this fails, we'll try to recover. */ en = create_filesystem_object(a); if ((en == ENOTDIR || en == ENOENT) && !(a->flags & ARCHIVE_EXTRACT_NO_AUTODIR)) { /* If the parent dir doesn't exist, try creating it. */ create_parent_dir(a, a->name); /* Now try to create the object again. */ en = create_filesystem_object(a); } if ((en == EISDIR || en == EEXIST) && (a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { /* If we're not overwriting, we're done. */ archive_entry_unset_size(a->entry); return (ARCHIVE_OK); } /* * Some platforms return EISDIR if you call * open(O_WRONLY | O_EXCL | O_CREAT) on a directory, some * return EEXIST. POSIX is ambiguous, requiring EISDIR * for open(O_WRONLY) on a dir and EEXIST for open(O_EXCL | O_CREAT) * on an existing item. */ if (en == EISDIR) { /* A dir is in the way of a non-dir, rmdir it. */ if (rmdir(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't remove already-existing dir"); return (ARCHIVE_FAILED); } a->pst = NULL; /* Try again. */ en = create_filesystem_object(a); } else if (en == EEXIST) { /* * We know something is in the way, but we don't know what; * we need to find out before we go any further. */ int r = 0; /* * The SECURE_SYMLINKS logic has already removed a * symlink to a dir if the client wants that. So * follow the symlink if we're creating a dir. */ if (S_ISDIR(a->mode)) r = stat(a->name, &a->st); /* * If it's not a dir (or it's a broken symlink), * then don't follow it. */ if (r != 0 || !S_ISDIR(a->mode)) r = lstat(a->name, &a->st); if (r != 0) { archive_set_error(&a->archive, errno, "Can't stat existing object"); return (ARCHIVE_FAILED); } /* * NO_OVERWRITE_NEWER doesn't apply to directories. */ if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE_NEWER) && !S_ISDIR(a->st.st_mode)) { if (!older(&(a->st), a->entry)) { archive_entry_unset_size(a->entry); return (ARCHIVE_OK); } } /* If it's our archive, we're done. */ if (a->skip_file_set && a->st.st_dev == a->skip_file_dev && a->st.st_ino == a->skip_file_ino) { archive_set_error(&a->archive, 0, "Refusing to overwrite archive"); return (ARCHIVE_FAILED); } if (!S_ISDIR(a->st.st_mode)) { /* A non-dir is in the way, unlink it. */ if (unlink(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't unlink already-existing object"); return (ARCHIVE_FAILED); } a->pst = NULL; /* Try again. */ en = create_filesystem_object(a); } else if (!S_ISDIR(a->mode)) { /* A dir is in the way of a non-dir, rmdir it. */ if (rmdir(a->name) != 0) { archive_set_error(&a->archive, errno, "Can't remove already-existing dir"); return (ARCHIVE_FAILED); } /* Try again. */ en = create_filesystem_object(a); } else { /* * There's a dir in the way of a dir. Don't * waste time with rmdir()/mkdir(), just fix * up the permissions on the existing dir. * Note that we don't change perms on existing * dirs unless _EXTRACT_PERM is specified. */ if ((a->mode != a->st.st_mode) && (a->todo & TODO_MODE_FORCE)) a->deferred |= (a->todo & TODO_MODE); /* Ownership doesn't need deferred fixup. */ en = 0; /* Forget the EEXIST. */ } } if (en) { /* Everything failed; give up here. */ archive_set_error(&a->archive, en, "Can't create '%s'", a->name); return (ARCHIVE_FAILED); } a->pst = NULL; /* Cached stat data no longer valid. */ return (ret); } /* * Returns 0 if creation succeeds, or else returns errno value from * the failed system call. Note: This function should only ever perform * a single system call. */ static int create_filesystem_object(struct archive_write_disk *a) { /* Create the entry. */ const char *linkname; mode_t final_mode, mode; int r; /* We identify hard/symlinks according to the link names. */ /* Since link(2) and symlink(2) don't handle modes, we're done here. */ linkname = archive_entry_hardlink(a->entry); if (linkname != NULL) { #if !HAVE_LINK return (EPERM); #else r = link(linkname, a->name) ? errno : 0; /* * New cpio and pax formats allow hardlink entries * to carry data, so we may have to open the file * for hardlink entries. * * If the hardlink was successfully created and * the archive doesn't have carry data for it, * consider it to be non-authoritative for meta data. * This is consistent with GNU tar and BSD pax. * If the hardlink does carry data, let the last * archive entry decide ownership. */ if (r == 0 && a->filesize <= 0) { a->todo = 0; a->deferred = 0; } else if (r == 0 && a->filesize > 0) { a->fd = open(a->name, O_WRONLY | O_TRUNC | O_BINARY); if (a->fd < 0) r = errno; } return (r); #endif } linkname = archive_entry_symlink(a->entry); if (linkname != NULL) { #if HAVE_SYMLINK return symlink(linkname, a->name) ? errno : 0; #else return (EPERM); #endif } /* * The remaining system calls all set permissions, so let's * try to take advantage of that to avoid an extra chmod() * call. (Recall that umask is set to zero right now!) */ /* Mode we want for the final restored object (w/o file type bits). */ final_mode = a->mode & 07777; /* * The mode that will actually be restored in this step. Note * that SUID, SGID, etc, require additional work to ensure * security, so we never restore them at this point. */ mode = final_mode & 0777 & a->user_umask; switch (a->mode & AE_IFMT) { default: /* POSIX requires that we fall through here. */ /* FALLTHROUGH */ case AE_IFREG: a->fd = open(a->name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY, mode); r = (a->fd < 0); break; case AE_IFCHR: #ifdef HAVE_MKNOD /* Note: we use AE_IFCHR for the case label, and * S_IFCHR for the mknod() call. This is correct. */ r = mknod(a->name, mode | S_IFCHR, archive_entry_rdev(a->entry)); break; #else /* TODO: Find a better way to warn about our inability * to restore a char device node. */ return (EINVAL); #endif /* HAVE_MKNOD */ case AE_IFBLK: #ifdef HAVE_MKNOD r = mknod(a->name, mode | S_IFBLK, archive_entry_rdev(a->entry)); break; #else /* TODO: Find a better way to warn about our inability * to restore a block device node. */ return (EINVAL); #endif /* HAVE_MKNOD */ case AE_IFDIR: mode = (mode | MINIMUM_DIR_MODE) & MAXIMUM_DIR_MODE; r = mkdir(a->name, mode); if (r == 0) { /* Defer setting dir times. */ a->deferred |= (a->todo & TODO_TIMES); a->todo &= ~TODO_TIMES; /* Never use an immediate chmod(). */ /* We can't avoid the chmod() entirely if EXTRACT_PERM * because of SysV SGID inheritance. */ if ((mode != final_mode) || (a->flags & ARCHIVE_EXTRACT_PERM)) a->deferred |= (a->todo & TODO_MODE); a->todo &= ~TODO_MODE; } break; case AE_IFIFO: #ifdef HAVE_MKFIFO r = mkfifo(a->name, mode); break; #else /* TODO: Find a better way to warn about our inability * to restore a fifo. */ return (EINVAL); #endif /* HAVE_MKFIFO */ } /* All the system calls above set errno on failure. */ if (r) return (errno); /* If we managed to set the final mode, we've avoided a chmod(). */ if (mode == final_mode) a->todo &= ~TODO_MODE; return (0); } /* * Cleanup function for archive_extract. Mostly, this involves processing * the fixup list, which is used to address a number of problems: * * Dir permissions might prevent us from restoring a file in that * dir, so we restore the dir with minimum 0700 permissions first, * then correct the mode at the end. * * Similarly, the act of restoring a file touches the directory * and changes the timestamp on the dir, so we have to touch-up dir * timestamps at the end as well. * * Some file flags can interfere with the restore by, for example, * preventing the creation of hardlinks to those files. * * Mac OS extended metadata includes ACLs, so must be deferred on dirs. * * Note that tar/cpio do not require that archives be in a particular * order; there is no way to know when the last file has been restored * within a directory, so there's no way to optimize the memory usage * here by fixing up the directory any earlier than the * end-of-archive. * * XXX TODO: Directory ACLs should be restored here, for the same * reason we set directory perms here. XXX */ static int _archive_write_disk_close(struct archive *_a) { struct archive_write_disk *a = (struct archive_write_disk *)_a; struct fixup_entry *next, *p; int ret; archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, "archive_write_disk_close"); ret = _archive_write_disk_finish_entry(&a->archive); /* Sort dir list so directories are fixed up in depth-first order. */ p = sort_dir_list(a->fixup_list); while (p != NULL) { a->pst = NULL; /* Mark stat cache as out-of-date. */ if (p->fixup & TODO_TIMES) { set_times(a, -1, p->mode, p->name, p->atime, p->atime_nanos, p->birthtime, p->birthtime_nanos, p->mtime, p->mtime_nanos, p->ctime, p->ctime_nanos); } if (p->fixup & TODO_MODE_BASE) chmod(p->name, p->mode); if (p->fixup & TODO_ACLS) set_acls(a, -1, p->name, &p->acl); if (p->fixup & TODO_FFLAGS) set_fflags_platform(a, -1, p->name, p->mode, p->fflags_set, 0); if (p->fixup & TODO_MAC_METADATA) set_mac_metadata(a, p->name, p->mac_metadata, p->mac_metadata_size); next = p->next; archive_acl_clear(&p->acl); free(p->mac_metadata); free(p->name); free(p); p = next; } a->fixup_list = NULL; return (ret); } static int _archive_write_disk_free(struct archive *_a) { struct archive_write_disk *a; int ret; if (_a == NULL) return (ARCHIVE_OK); archive_check_magic(_a, ARCHIVE_WRITE_DISK_MAGIC, ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_write_disk_free"); a = (struct archive_write_disk *)_a; ret = _archive_write_disk_close(&a->archive); archive_write_disk_set_group_lookup(&a->archive, NULL, NULL, NULL); archive_write_disk_set_user_lookup(&a->archive, NULL, NULL, NULL); if (a->entry) archive_entry_free(a->entry); archive_string_free(&a->_name_data); archive_string_free(&a->archive.error_string); archive_string_free(&a->path_safe); a->archive.magic = 0; __archive_clean(&a->archive); free(a); return (ret); } /* * Simple O(n log n) merge sort to order the fixup list. In * particular, we want to restore dir timestamps depth-first. */ static struct fixup_entry * sort_dir_list(struct fixup_entry *p) { struct fixup_entry *a, *b, *t; if (p == NULL) return (NULL); /* A one-item list is already sorted. */ if (p->next == NULL) return (p); /* Step 1: split the list. */ t = p; a = p->next->next; while (a != NULL) { /* Step a twice, t once. */ a = a->next; if (a != NULL) a = a->next; t = t->next; } /* Now, t is at the mid-point, so break the list here. */ b = t->next; t->next = NULL; a = p; /* Step 2: Recursively sort the two sub-lists. */ a = sort_dir_list(a); b = sort_dir_list(b); /* Step 3: Merge the returned lists. */ /* Pick the first element for the merged list. */ if (strcmp(a->name, b->name) > 0) { t = p = a; a = a->next; } else { t = p = b; b = b->next; } /* Always put the later element on the list first. */ while (a != NULL && b != NULL) { if (strcmp(a->name, b->name) > 0) { t->next = a; a = a->next; } else { t->next = b; b = b->next; } t = t->next; } /* Only one list is non-empty, so just splice it on. */ if (a != NULL) t->next = a; if (b != NULL) t->next = b; return (p); } /* * Returns a new, initialized fixup entry. * * TODO: Reduce the memory requirements for this list by using a tree * structure rather than a simple list of names. */ static struct fixup_entry * new_fixup(struct archive_write_disk *a, const char *pathname) { struct fixup_entry *fe; fe = (struct fixup_entry *)calloc(1, sizeof(struct fixup_entry)); if (fe == NULL) return (NULL); fe->next = a->fixup_list; a->fixup_list = fe; fe->fixup = 0; fe->name = strdup(pathname); return (fe); } /* * Returns a fixup structure for the current entry. */ static struct fixup_entry * current_fixup(struct archive_write_disk *a, const char *pathname) { if (a->current_fixup == NULL) a->current_fixup = new_fixup(a, pathname); return (a->current_fixup); } /* TODO: Make this work. */ /* * TODO: The deep-directory support bypasses this; disable deep directory * support if we're doing symlink checks. */ /* * TODO: Someday, integrate this with the deep dir support; they both * scan the path and both can be optimized by comparing against other * recent paths. */ /* TODO: Extend this to support symlinks on Windows Vista and later. */ static int check_symlinks(struct archive_write_disk *a) { #if !defined(HAVE_LSTAT) /* Platform doesn't have lstat, so we can't look for symlinks. */ (void)a; /* UNUSED */ return (ARCHIVE_OK); #else char *pn; char c; int r; struct stat st; /* * Guard against symlink tricks. Reject any archive entry whose * destination would be altered by a symlink. */ /* Whatever we checked last time doesn't need to be re-checked. */ pn = a->name; if (archive_strlen(&(a->path_safe)) > 0) { char *p = a->path_safe.s; while ((*pn != '\0') && (*p == *pn)) ++p, ++pn; } c = pn[0]; /* Keep going until we've checked the entire name. */ while (pn[0] != '\0' && (pn[0] != '/' || pn[1] != '\0')) { /* Skip the next path element. */ while (*pn != '\0' && *pn != '/') ++pn; c = pn[0]; pn[0] = '\0'; /* Check that we haven't hit a symlink. */ r = lstat(a->name, &st); if (r != 0) { /* We've hit a dir that doesn't exist; stop now. */ if (errno == ENOENT) break; } else if (S_ISLNK(st.st_mode)) { if (c == '\0') { /* * Last element is symlink; remove it * so we can overwrite it with the * item being extracted. */ if (unlink(a->name)) { archive_set_error(&a->archive, errno, "Could not remove symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } a->pst = NULL; /* * Even if we did remove it, a warning * is in order. The warning is silly, * though, if we're just replacing one * symlink with another symlink. */ if (!S_ISLNK(a->mode)) { archive_set_error(&a->archive, 0, "Removing symlink %s", a->name); } /* Symlink gone. No more problem! */ pn[0] = c; return (0); } else if (a->flags & ARCHIVE_EXTRACT_UNLINK) { /* User asked us to remove problems. */ if (unlink(a->name) != 0) { archive_set_error(&a->archive, 0, "Cannot remove intervening symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } a->pst = NULL; } else { archive_set_error(&a->archive, 0, "Cannot extract through symlink %s", a->name); pn[0] = c; return (ARCHIVE_FAILED); } } } pn[0] = c; /* We've checked and/or cleaned the whole path, so remember it. */ archive_strcpy(&a->path_safe, a->name); return (ARCHIVE_OK); #endif } #if defined(__CYGWIN__) /* * 1. Convert a path separator from '\' to '/' . * We shouldn't check multibyte character directly because some * character-set have been using the '\' character for a part of * its multibyte character code. * 2. Replace unusable characters in Windows with underscore('_'). * See also : http://msdn.microsoft.com/en-us/library/aa365247.aspx */ static void cleanup_pathname_win(struct archive_write_disk *a) { wchar_t wc; char *p; size_t alen, l; int mb, complete, utf8; alen = 0; mb = 0; complete = 1; utf8 = (strcmp(nl_langinfo(CODESET), "UTF-8") == 0)? 1: 0; for (p = a->name; *p != '\0'; p++) { ++alen; if (*p == '\\') { /* If previous byte is smaller than 128, * this is not second byte of multibyte characters, * so we can replace '\' with '/'. */ if (utf8 || !mb) *p = '/'; else complete = 0;/* uncompleted. */ } else if (*(unsigned char *)p > 127) mb = 1; else mb = 0; /* Rewrite the path name if its next character is unusable. */ if (*p == ':' || *p == '*' || *p == '?' || *p == '"' || *p == '<' || *p == '>' || *p == '|') *p = '_'; } if (complete) return; /* * Convert path separator in wide-character. */ p = a->name; while (*p != '\0' && alen) { l = mbtowc(&wc, p, alen); - if (l == -1) { + if (l == (size_t)-1) { while (*p != '\0') { if (*p == '\\') *p = '/'; ++p; } break; } if (l == 1 && wc == L'\\') *p = '/'; p += l; alen -= l; } } #endif /* * Canonicalize the pathname. In particular, this strips duplicate * '/' characters, '.' elements, and trailing '/'. It also raises an * error for an empty path, a trailing '..' or (if _SECURE_NODOTDOT is * set) any '..' in the path. */ static int cleanup_pathname(struct archive_write_disk *a) { char *dest, *src; char separator = '\0'; dest = src = a->name; if (*src == '\0') { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid empty pathname"); return (ARCHIVE_FAILED); } #if defined(__CYGWIN__) cleanup_pathname_win(a); #endif /* Skip leading '/'. */ if (*src == '/') separator = *src++; /* Scan the pathname one element at a time. */ for (;;) { /* src points to first char after '/' */ if (src[0] == '\0') { break; } else if (src[0] == '/') { /* Found '//', ignore second one. */ src++; continue; } else if (src[0] == '.') { if (src[1] == '\0') { /* Ignore trailing '.' */ break; } else if (src[1] == '/') { /* Skip './'. */ src += 2; continue; } else if (src[1] == '.') { if (src[2] == '/' || src[2] == '\0') { /* Conditionally warn about '..' */ if (a->flags & ARCHIVE_EXTRACT_SECURE_NODOTDOT) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Path contains '..'"); return (ARCHIVE_FAILED); } } /* * Note: Under no circumstances do we * remove '..' elements. In * particular, restoring * '/foo/../bar/' should create the * 'foo' dir as a side-effect. */ } } /* Copy current element, including leading '/'. */ if (separator) *dest++ = '/'; while (*src != '\0' && *src != '/') { *dest++ = *src++; } if (*src == '\0') break; /* Skip '/' separator. */ separator = *src++; } /* * We've just copied zero or more path elements, not including the * final '/'. */ if (dest == a->name) { /* * Nothing got copied. The path must have been something * like '.' or '/' or './' or '/././././/./'. */ if (separator) *dest++ = '/'; else *dest++ = '.'; } /* Terminate the result. */ *dest = '\0'; return (ARCHIVE_OK); } /* * Create the parent directory of the specified path, assuming path * is already in mutable storage. */ static int create_parent_dir(struct archive_write_disk *a, char *path) { char *slash; int r; /* Remove tail element to obtain parent name. */ slash = strrchr(path, '/'); if (slash == NULL) return (ARCHIVE_OK); *slash = '\0'; r = create_dir(a, path); *slash = '/'; return (r); } /* * Create the specified dir, recursing to create parents as necessary. * * Returns ARCHIVE_OK if the path exists when we're done here. * Otherwise, returns ARCHIVE_FAILED. * Assumes path is in mutable storage; path is unchanged on exit. */ static int create_dir(struct archive_write_disk *a, char *path) { struct stat st; struct fixup_entry *le; char *slash, *base; mode_t mode_final, mode; int r; /* Check for special names and just skip them. */ slash = strrchr(path, '/'); if (slash == NULL) base = path; else base = slash + 1; if (base[0] == '\0' || (base[0] == '.' && base[1] == '\0') || (base[0] == '.' && base[1] == '.' && base[2] == '\0')) { /* Don't bother trying to create null path, '.', or '..'. */ if (slash != NULL) { *slash = '\0'; r = create_dir(a, path); *slash = '/'; return (r); } return (ARCHIVE_OK); } /* * Yes, this should be stat() and not lstat(). Using lstat() * here loses the ability to extract through symlinks. Also note * that this should not use the a->st cache. */ if (stat(path, &st) == 0) { if (S_ISDIR(st.st_mode)) return (ARCHIVE_OK); if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { archive_set_error(&a->archive, EEXIST, "Can't create directory '%s'", path); return (ARCHIVE_FAILED); } if (unlink(path) != 0) { archive_set_error(&a->archive, errno, "Can't create directory '%s': " "Conflicting file cannot be removed", path); return (ARCHIVE_FAILED); } } else if (errno != ENOENT && errno != ENOTDIR) { /* Stat failed? */ archive_set_error(&a->archive, errno, "Can't test directory '%s'", path); return (ARCHIVE_FAILED); } else if (slash != NULL) { *slash = '\0'; r = create_dir(a, path); *slash = '/'; if (r != ARCHIVE_OK) return (r); } /* * Mode we want for the final restored directory. Per POSIX, * implicitly-created dirs must be created obeying the umask. * There's no mention whether this is different for privileged * restores (which the rest of this code handles by pretending * umask=0). I've chosen here to always obey the user's umask for * implicit dirs, even if _EXTRACT_PERM was specified. */ mode_final = DEFAULT_DIR_MODE & ~a->user_umask; /* Mode we want on disk during the restore process. */ mode = mode_final; mode |= MINIMUM_DIR_MODE; mode &= MAXIMUM_DIR_MODE; if (mkdir(path, mode) == 0) { if (mode != mode_final) { le = new_fixup(a, path); le->fixup |=TODO_MODE_BASE; le->mode = mode_final; } return (ARCHIVE_OK); } /* * Without the following check, a/b/../b/c/d fails at the * second visit to 'b', so 'd' can't be created. Note that we * don't add it to the fixup list here, as it's already been * added. */ if (stat(path, &st) == 0 && S_ISDIR(st.st_mode)) return (ARCHIVE_OK); archive_set_error(&a->archive, errno, "Failed to create dir '%s'", path); return (ARCHIVE_FAILED); } /* * Note: Although we can skip setting the user id if the desired user * id matches the current user, we cannot skip setting the group, as * many systems set the gid based on the containing directory. So * we have to perform a chown syscall if we want to set the SGID * bit. (The alternative is to stat() and then possibly chown(); it's * more efficient to skip the stat() and just always chown().) Note * that a successful chown() here clears the TODO_SGID_CHECK bit, which * allows set_mode to skip the stat() check for the GID. */ static int set_ownership(struct archive_write_disk *a) { #ifndef __CYGWIN__ /* unfortunately, on win32 there is no 'root' user with uid 0, so we just have to try the chown and see if it works */ /* If we know we can't change it, don't bother trying. */ if (a->user_uid != 0 && a->user_uid != a->uid) { archive_set_error(&a->archive, errno, "Can't set UID=%jd", (intmax_t)a->uid); return (ARCHIVE_WARN); } #endif #ifdef HAVE_FCHOWN /* If we have an fd, we can avoid a race. */ if (a->fd >= 0 && fchown(a->fd, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #endif /* We prefer lchown() but will use chown() if that's all we have. */ /* Of course, if we have neither, this will always fail. */ #ifdef HAVE_LCHOWN if (lchown(a->name, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #elif HAVE_CHOWN if (!S_ISLNK(a->mode) && chown(a->name, a->uid, a->gid) == 0) { /* We've set owner and know uid/gid are correct. */ a->todo &= ~(TODO_OWNER | TODO_SGID_CHECK | TODO_SUID_CHECK); return (ARCHIVE_OK); } #endif archive_set_error(&a->archive, errno, "Can't set user=%jd/group=%jd for %s", (intmax_t)a->uid, (intmax_t)a->gid, a->name); return (ARCHIVE_WARN); } /* * Note: Returns 0 on success, non-zero on failure. */ static int set_time(int fd, int mode, const char *name, time_t atime, long atime_nsec, time_t mtime, long mtime_nsec) { /* Select the best implementation for this platform. */ #if defined(HAVE_UTIMENSAT) && defined(HAVE_FUTIMENS) /* * utimensat() and futimens() are defined in * POSIX.1-2008. They support ns resolution and setting times * on fds and symlinks. */ struct timespec ts[2]; + (void)mode; /* UNUSED */ ts[0].tv_sec = atime; ts[0].tv_nsec = atime_nsec; ts[1].tv_sec = mtime; ts[1].tv_nsec = mtime_nsec; if (fd >= 0) return futimens(fd, ts); return utimensat(AT_FDCWD, name, ts, AT_SYMLINK_NOFOLLOW); #elif HAVE_UTIMES /* * The utimes()-family functions support µs-resolution and * setting times fds and symlinks. utimes() is documented as * LEGACY by POSIX, futimes() and lutimes() are not described * in POSIX. */ struct timeval times[2]; times[0].tv_sec = atime; times[0].tv_usec = atime_nsec / 1000; times[1].tv_sec = mtime; times[1].tv_usec = mtime_nsec / 1000; #ifdef HAVE_FUTIMES if (fd >= 0) return (futimes(fd, times)); #else (void)fd; /* UNUSED */ #endif #ifdef HAVE_LUTIMES (void)mode; /* UNUSED */ return (lutimes(name, times)); #else if (S_ISLNK(mode)) return (0); return (utimes(name, times)); #endif #elif defined(HAVE_UTIME) /* * utime() is POSIX-standard but only supports 1s resolution and * does not support fds or symlinks. */ struct utimbuf times; (void)fd; /* UNUSED */ (void)name; /* UNUSED */ (void)atime_nsec; /* UNUSED */ (void)mtime_nsec; /* UNUSED */ times.actime = atime; times.modtime = mtime; if (S_ISLNK(mode)) return (ARCHIVE_OK); return (utime(name, ×)); #else /* * We don't know how to set the time on this platform. */ + (void)fd; /* UNUSED */ + (void)mode; /* UNUSED */ + (void)name; /* UNUSED */ + (void)atime_nsec; /* UNUSED */ + (void)mtime_nsec; /* UNUSED */ return (ARCHIVE_WARN); #endif } #ifdef F_SETTIMES /* Tru64 */ static int set_time_tru64(int fd, int mode, const char *name, time_t atime, long atime_nsec, time_t mtime, long mtime_nsec, time_t ctime, long ctime_nsec) { struct attr_timbuf tstamp; struct timeval times[3]; times[0].tv_sec = atime; times[0].tv_usec = atime_nsec / 1000; times[1].tv_sec = mtime; times[1].tv_usec = mtime_nsec / 1000; times[2].tv_sec = ctime; times[2].tv_usec = ctime_nsec / 1000; tstamp.atime = times[0]; tstamp.mtime = times[1]; tstamp.ctime = times[2]; return (fcntl(fd,F_SETTIMES,&tstamp)); } #endif /* Tru64 */ static int set_times(struct archive_write_disk *a, int fd, int mode, const char *name, time_t atime, long atime_nanos, time_t birthtime, long birthtime_nanos, time_t mtime, long mtime_nanos, time_t cctime, long ctime_nanos) { /* Note: set_time doesn't use libarchive return conventions! * It uses syscall conventions. So 0 here instead of ARCHIVE_OK. */ int r1 = 0, r2 = 0; #ifdef F_SETTIMES /* * on Tru64 try own fcntl first which can restore even the * ctime, fall back to default code path below if it fails * or if we are not running as root */ if (a->user_uid == 0 && set_time_tru64(fd, mode, name, atime, atime_nanos, mtime, mtime_nanos, cctime, ctime_nanos) == 0) { return (ARCHIVE_OK); } #else /* Tru64 */ (void)cctime; /* UNUSED */ (void)ctime_nanos; /* UNUSED */ #endif /* Tru64 */ #ifdef HAVE_STRUCT_STAT_ST_BIRTHTIME /* * If you have struct stat.st_birthtime, we assume BSD * birthtime semantics, in which {f,l,}utimes() updates * birthtime to earliest mtime. So we set the time twice, * first using the birthtime, then using the mtime. If * birthtime == mtime, this isn't necessary, so we skip it. * If birthtime > mtime, then this won't work, so we skip it. */ if (birthtime < mtime || (birthtime == mtime && birthtime_nanos < mtime_nanos)) r1 = set_time(fd, mode, name, atime, atime_nanos, birthtime, birthtime_nanos); +#else + (void)birthtime; /* UNUSED */ + (void)birthtime_nanos; /* UNUSED */ #endif r2 = set_time(fd, mode, name, atime, atime_nanos, mtime, mtime_nanos); if (r1 != 0 || r2 != 0) { archive_set_error(&a->archive, errno, "Can't restore time"); return (ARCHIVE_WARN); } return (ARCHIVE_OK); } static int set_times_from_entry(struct archive_write_disk *a) { time_t atime, birthtime, mtime, cctime; long atime_nsec, birthtime_nsec, mtime_nsec, ctime_nsec; /* Suitable defaults. */ atime = birthtime = mtime = cctime = a->start_time; atime_nsec = birthtime_nsec = mtime_nsec = ctime_nsec = 0; /* If no time was provided, we're done. */ if (!archive_entry_atime_is_set(a->entry) #if HAVE_STRUCT_STAT_ST_BIRTHTIME && !archive_entry_birthtime_is_set(a->entry) #endif && !archive_entry_mtime_is_set(a->entry)) return (ARCHIVE_OK); if (archive_entry_atime_is_set(a->entry)) { atime = archive_entry_atime(a->entry); atime_nsec = archive_entry_atime_nsec(a->entry); } if (archive_entry_birthtime_is_set(a->entry)) { birthtime = archive_entry_birthtime(a->entry); birthtime_nsec = archive_entry_birthtime_nsec(a->entry); } if (archive_entry_mtime_is_set(a->entry)) { mtime = archive_entry_mtime(a->entry); mtime_nsec = archive_entry_mtime_nsec(a->entry); } if (archive_entry_ctime_is_set(a->entry)) { cctime = archive_entry_ctime(a->entry); ctime_nsec = archive_entry_ctime_nsec(a->entry); } return set_times(a, a->fd, a->mode, a->name, atime, atime_nsec, birthtime, birthtime_nsec, mtime, mtime_nsec, cctime, ctime_nsec); } static int set_mode(struct archive_write_disk *a, int mode) { int r = ARCHIVE_OK; mode &= 07777; /* Strip off file type bits. */ if (a->todo & TODO_SGID_CHECK) { /* * If we don't know the GID is right, we must stat() * to verify it. We can't just check the GID of this * process, since systems sometimes set GID from * the enclosing dir or based on ACLs. */ if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); if (a->pst->st_gid != a->gid) { mode &= ~ S_ISGID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { /* * This is only an error if you * requested owner restore. If you * didn't, we'll try to restore * sgid/suid, but won't consider it a * problem if we can't. */ archive_set_error(&a->archive, -1, "Can't restore SGID bit"); r = ARCHIVE_WARN; } } /* While we're here, double-check the UID. */ if (a->pst->st_uid != a->uid && (a->todo & TODO_SUID)) { mode &= ~ S_ISUID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { archive_set_error(&a->archive, -1, "Can't restore SUID bit"); r = ARCHIVE_WARN; } } a->todo &= ~TODO_SGID_CHECK; a->todo &= ~TODO_SUID_CHECK; } else if (a->todo & TODO_SUID_CHECK) { /* * If we don't know the UID is right, we can just check * the user, since all systems set the file UID from * the process UID. */ if (a->user_uid != a->uid) { mode &= ~ S_ISUID; if (a->flags & ARCHIVE_EXTRACT_OWNER) { archive_set_error(&a->archive, -1, "Can't make file SUID"); r = ARCHIVE_WARN; } } a->todo &= ~TODO_SUID_CHECK; } if (S_ISLNK(a->mode)) { #ifdef HAVE_LCHMOD /* * If this is a symlink, use lchmod(). If the * platform doesn't support lchmod(), just skip it. A * platform that doesn't provide a way to set * permissions on symlinks probably ignores * permissions on symlinks, so a failure here has no * impact. */ if (lchmod(a->name, mode) != 0) { archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } #endif } else if (!S_ISDIR(a->mode)) { /* * If it's not a symlink and not a dir, then use * fchmod() or chmod(), depending on whether we have * an fd. Dirs get their perms set during the * post-extract fixup, which is handled elsewhere. */ #ifdef HAVE_FCHMOD if (a->fd >= 0) { if (fchmod(a->fd, mode) != 0) { archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } } else #endif /* If this platform lacks fchmod(), then * we'll just use chmod(). */ if (chmod(a->name, mode) != 0) { archive_set_error(&a->archive, errno, "Can't set permissions to 0%o", (int)mode); r = ARCHIVE_WARN; } } return (r); } static int set_fflags(struct archive_write_disk *a) { struct fixup_entry *le; unsigned long set, clear; int r; int critical_flags; mode_t mode = archive_entry_mode(a->entry); /* * Make 'critical_flags' hold all file flags that can't be * immediately restored. For example, on BSD systems, * SF_IMMUTABLE prevents hardlinks from being created, so * should not be set until after any hardlinks are created. To * preserve some semblance of portability, this uses #ifdef * extensively. Ugly, but it works. * * Yes, Virginia, this does create a security race. It's mitigated * somewhat by the practice of creating dirs 0700 until the extract * is done, but it would be nice if we could do more than that. * People restoring critical file systems should be wary of * other programs that might try to muck with files as they're * being restored. */ /* Hopefully, the compiler will optimize this mess into a constant. */ critical_flags = 0; #ifdef SF_IMMUTABLE critical_flags |= SF_IMMUTABLE; #endif #ifdef UF_IMMUTABLE critical_flags |= UF_IMMUTABLE; #endif #ifdef SF_APPEND critical_flags |= SF_APPEND; #endif #ifdef UF_APPEND critical_flags |= UF_APPEND; #endif #ifdef EXT2_APPEND_FL critical_flags |= EXT2_APPEND_FL; #endif #ifdef EXT2_IMMUTABLE_FL critical_flags |= EXT2_IMMUTABLE_FL; #endif if (a->todo & TODO_FFLAGS) { archive_entry_fflags(a->entry, &set, &clear); /* * The first test encourages the compiler to eliminate * all of this if it's not necessary. */ if ((critical_flags != 0) && (set & critical_flags)) { le = current_fixup(a, a->name); le->fixup |= TODO_FFLAGS; le->fflags_set = set; /* Store the mode if it's not already there. */ if ((le->fixup & TODO_MODE) == 0) le->mode = mode; } else { r = set_fflags_platform(a, a->fd, a->name, mode, set, clear); if (r != ARCHIVE_OK) return (r); } } return (ARCHIVE_OK); } #if ( defined(HAVE_LCHFLAGS) || defined(HAVE_CHFLAGS) || defined(HAVE_FCHFLAGS) ) && defined(HAVE_STRUCT_STAT_ST_FLAGS) /* * BSD reads flags using stat() and sets them with one of {f,l,}chflags() */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { int r; (void)mode; /* UNUSED */ if (set == 0 && clear == 0) return (ARCHIVE_OK); /* * XXX Is the stat here really necessary? Or can I just use * the 'set' flags directly? In particular, I'm not sure * about the correct approach if we're overwriting an existing * file that already has flags on it. XXX */ if ((r = lazy_stat(a)) != ARCHIVE_OK) return (r); a->st.st_flags &= ~clear; a->st.st_flags |= set; #ifdef HAVE_FCHFLAGS /* If platform has fchflags() and we were given an fd, use it. */ if (fd >= 0 && fchflags(fd, a->st.st_flags) == 0) return (ARCHIVE_OK); #endif /* * If we can't use the fd to set the flags, we'll use the * pathname to set flags. We prefer lchflags() but will use * chflags() if we must. */ #ifdef HAVE_LCHFLAGS if (lchflags(name, a->st.st_flags) == 0) return (ARCHIVE_OK); #elif defined(HAVE_CHFLAGS) if (S_ISLNK(a->st.st_mode)) { archive_set_error(&a->archive, errno, "Can't set file flags on symlink."); return (ARCHIVE_WARN); } if (chflags(name, a->st.st_flags) == 0) return (ARCHIVE_OK); #endif archive_set_error(&a->archive, errno, "Failed to set file flags"); return (ARCHIVE_WARN); } #elif defined(EXT2_IOC_GETFLAGS) && defined(EXT2_IOC_SETFLAGS) && defined(HAVE_WORKING_EXT2_IOC_GETFLAGS) /* * Linux uses ioctl() to read and write file flags. */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { int ret; int myfd = fd; unsigned long newflags, oldflags; unsigned long sf_mask = 0; if (set == 0 && clear == 0) return (ARCHIVE_OK); /* Only regular files and dirs can have flags. */ if (!S_ISREG(mode) && !S_ISDIR(mode)) return (ARCHIVE_OK); /* If we weren't given an fd, open it ourselves. */ if (myfd < 0) myfd = open(name, O_RDONLY | O_NONBLOCK | O_BINARY); if (myfd < 0) return (ARCHIVE_OK); /* * Linux has no define for the flags that are only settable by * the root user. This code may seem a little complex, but * there seem to be some Linux systems that lack these * defines. (?) The code below degrades reasonably gracefully * if sf_mask is incomplete. */ #ifdef EXT2_IMMUTABLE_FL sf_mask |= EXT2_IMMUTABLE_FL; #endif #ifdef EXT2_APPEND_FL sf_mask |= EXT2_APPEND_FL; #endif /* * XXX As above, this would be way simpler if we didn't have * to read the current flags from disk. XXX */ ret = ARCHIVE_OK; /* Read the current file flags. */ if (ioctl(myfd, EXT2_IOC_GETFLAGS, &oldflags) < 0) goto fail; /* Try setting the flags as given. */ newflags = (oldflags & ~clear) | set; if (ioctl(myfd, EXT2_IOC_SETFLAGS, &newflags) >= 0) goto cleanup; if (errno != EPERM) goto fail; /* If we couldn't set all the flags, try again with a subset. */ newflags &= ~sf_mask; oldflags &= sf_mask; newflags |= oldflags; if (ioctl(myfd, EXT2_IOC_SETFLAGS, &newflags) >= 0) goto cleanup; /* We couldn't set the flags, so report the failure. */ fail: archive_set_error(&a->archive, errno, "Failed to set file flags"); ret = ARCHIVE_WARN; cleanup: if (fd < 0) close(myfd); return (ret); } #else /* * Of course, some systems have neither BSD chflags() nor Linux' flags * support through ioctl(). */ static int set_fflags_platform(struct archive_write_disk *a, int fd, const char *name, mode_t mode, unsigned long set, unsigned long clear) { (void)a; /* UNUSED */ (void)fd; /* UNUSED */ (void)name; /* UNUSED */ (void)mode; /* UNUSED */ (void)set; /* UNUSED */ (void)clear; /* UNUSED */ return (ARCHIVE_OK); } #endif /* __linux */ #ifndef HAVE_COPYFILE_H /* Default is to simply drop Mac extended metadata. */ static int set_mac_metadata(struct archive_write_disk *a, const char *pathname, const void *metadata, size_t metadata_size) { (void)a; /* UNUSED */ (void)pathname; /* UNUSED */ (void)metadata; /* UNUSED */ (void)metadata_size; /* UNUSED */ return (ARCHIVE_OK); } #else /* * On Mac OS, we use copyfile() to unpack the metadata and * apply it to the target file. */ static int set_mac_metadata(struct archive_write_disk *a, const char *pathname, const void *metadata, size_t metadata_size) { struct archive_string tmp; ssize_t written; int fd; int ret = ARCHIVE_OK; /* This would be simpler if copyfile() could just accept the * metadata as a block of memory; then we could sidestep this * silly dance of writing the data to disk just so that * copyfile() can read it back in again. */ archive_string_init(&tmp); archive_strcpy(&tmp, pathname); archive_strcat(&tmp, ".XXXXXX"); fd = mkstemp(tmp.s); if (fd < 0) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); return (ARCHIVE_WARN); } written = write(fd, metadata, metadata_size); close(fd); if (written != metadata_size || copyfile(tmp.s, pathname, 0, COPYFILE_UNPACK | COPYFILE_NOFOLLOW | COPYFILE_ACL | COPYFILE_XATTR)) { archive_set_error(&a->archive, errno, "Failed to restore metadata"); ret = ARCHIVE_WARN; } unlink(tmp.s); return (ret); } #endif #ifndef HAVE_POSIX_ACL /* Default empty function body to satisfy mainline code. */ static int set_acls(struct archive_write_disk *a, int fd, const char *name, - struct archive_acl *acl) + struct archive_acl *aacl) { (void)a; /* UNUSED */ (void)fd; /* UNUSED */ (void)name; /* UNUSED */ - (void)acl; /* UNUSED */ + (void)aacl; /* UNUSED */ return (ARCHIVE_OK); } #else /* * XXX TODO: What about ACL types other than ACCESS and DEFAULT? */ static int set_acls(struct archive_write_disk *a, int fd, const char *name, struct archive_acl *abstract_acl) { int ret; ret = set_acl(a, fd, name, abstract_acl, ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, "access"); if (ret != ARCHIVE_OK) return (ret); ret = set_acl(a, fd, name, abstract_acl, ACL_TYPE_DEFAULT, ARCHIVE_ENTRY_ACL_TYPE_DEFAULT, "default"); return (ret); } static int set_acl(struct archive_write_disk *a, int fd, const char *name, struct archive_acl *abstract_acl, acl_type_t acl_type, int ae_requested_type, const char *tname) { acl_t acl; acl_entry_t acl_entry; acl_permset_t acl_permset; int ret; int ae_type, ae_permset, ae_tag, ae_id; uid_t ae_uid; gid_t ae_gid; const char *ae_name; int entries; ret = ARCHIVE_OK; entries = archive_acl_reset(abstract_acl, ae_requested_type); if (entries == 0) return (ARCHIVE_OK); acl = acl_init(entries); while (archive_acl_next(&a->archive, abstract_acl, ae_requested_type, &ae_type, &ae_permset, &ae_tag, &ae_id, &ae_name) == ARCHIVE_OK) { acl_create_entry(&acl, &acl_entry); switch (ae_tag) { case ARCHIVE_ENTRY_ACL_USER: acl_set_tag_type(acl_entry, ACL_USER); ae_uid = archive_write_disk_uid(&a->archive, ae_name, ae_id); acl_set_qualifier(acl_entry, &ae_uid); break; case ARCHIVE_ENTRY_ACL_GROUP: acl_set_tag_type(acl_entry, ACL_GROUP); ae_gid = archive_write_disk_gid(&a->archive, ae_name, ae_id); acl_set_qualifier(acl_entry, &ae_gid); break; case ARCHIVE_ENTRY_ACL_USER_OBJ: acl_set_tag_type(acl_entry, ACL_USER_OBJ); break; case ARCHIVE_ENTRY_ACL_GROUP_OBJ: acl_set_tag_type(acl_entry, ACL_GROUP_OBJ); break; case ARCHIVE_ENTRY_ACL_MASK: acl_set_tag_type(acl_entry, ACL_MASK); break; case ARCHIVE_ENTRY_ACL_OTHER: acl_set_tag_type(acl_entry, ACL_OTHER); break; default: /* XXX */ break; } acl_get_permset(acl_entry, &acl_permset); acl_clear_perms(acl_permset); if (ae_permset & ARCHIVE_ENTRY_ACL_EXECUTE) acl_add_perm(acl_permset, ACL_EXECUTE); if (ae_permset & ARCHIVE_ENTRY_ACL_WRITE) acl_add_perm(acl_permset, ACL_WRITE); if (ae_permset & ARCHIVE_ENTRY_ACL_READ) acl_add_perm(acl_permset, ACL_READ); } /* Try restoring the ACL through 'fd' if we can. */ #if HAVE_ACL_SET_FD if (fd >= 0 && acl_type == ACL_TYPE_ACCESS && acl_set_fd(fd, acl) == 0) ret = ARCHIVE_OK; else #else #if HAVE_ACL_SET_FD_NP if (fd >= 0 && acl_set_fd_np(fd, acl, acl_type) == 0) ret = ARCHIVE_OK; else #endif #endif if (acl_set_file(name, acl_type, acl) != 0) { archive_set_error(&a->archive, errno, "Failed to set %s acl", tname); ret = ARCHIVE_WARN; } acl_free(acl); return (ret); } #endif #if HAVE_LSETXATTR || HAVE_LSETEA /* * Restore extended attributes - Linux and AIX implementations: * AIX' ea interface is syntaxwise identical to the Linux xattr interface. */ static int set_xattrs(struct archive_write_disk *a) { struct archive_entry *entry = a->entry; static int warning_done = 0; int ret = ARCHIVE_OK; int i = archive_entry_xattr_reset(entry); while (i--) { const char *name; const void *value; size_t size; archive_entry_xattr_next(entry, &name, &value, &size); if (name != NULL && strncmp(name, "xfsroot.", 8) != 0 && strncmp(name, "system.", 7) != 0) { int e; #if HAVE_FSETXATTR if (a->fd >= 0) e = fsetxattr(a->fd, name, value, size, 0); else #elif HAVE_FSETEA if (a->fd >= 0) e = fsetea(a->fd, name, value, size, 0); else #endif { #if HAVE_LSETXATTR e = lsetxattr(archive_entry_pathname(entry), name, value, size, 0); #elif HAVE_LSETEA e = lsetea(archive_entry_pathname(entry), name, value, size, 0); #endif } if (e == -1) { if (errno == ENOTSUP || errno == ENOSYS) { if (!warning_done) { warning_done = 1; archive_set_error(&a->archive, errno, "Cannot restore extended " "attributes on this file " "system"); } } else archive_set_error(&a->archive, errno, "Failed to set extended attribute"); ret = ARCHIVE_WARN; } } else { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Invalid extended attribute encountered"); ret = ARCHIVE_WARN; } } return (ret); } #elif HAVE_EXTATTR_SET_FILE && HAVE_DECL_EXTATTR_NAMESPACE_USER /* * Restore extended attributes - FreeBSD implementation */ static int set_xattrs(struct archive_write_disk *a) { struct archive_entry *entry = a->entry; static int warning_done = 0; int ret = ARCHIVE_OK; int i = archive_entry_xattr_reset(entry); while (i--) { const char *name; const void *value; size_t size; archive_entry_xattr_next(entry, &name, &value, &size); if (name != NULL) { int e; int namespace; if (strncmp(name, "user.", 5) == 0) { /* "user." attributes go to user namespace */ name += 5; namespace = EXTATTR_NAMESPACE_USER; } else { /* Warn about other extended attributes. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Can't restore extended attribute ``%s''", name); ret = ARCHIVE_WARN; continue; } errno = 0; #if HAVE_EXTATTR_SET_FD if (a->fd >= 0) e = extattr_set_fd(a->fd, namespace, name, value, size); else #endif /* TODO: should we use extattr_set_link() instead? */ { e = extattr_set_file(archive_entry_pathname(entry), namespace, name, value, size); } if (e != (int)size) { if (errno == ENOTSUP || errno == ENOSYS) { if (!warning_done) { warning_done = 1; archive_set_error(&a->archive, errno, "Cannot restore extended " "attributes on this file " "system"); } } else { archive_set_error(&a->archive, errno, "Failed to set extended attribute"); } ret = ARCHIVE_WARN; } } } return (ret); } #else /* * Restore extended attributes - stub implementation for unsupported systems */ static int set_xattrs(struct archive_write_disk *a) { static int warning_done = 0; /* If there aren't any extended attributes, then it's okay not * to extract them, otherwise, issue a single warning. */ if (archive_entry_xattr_count(a->entry) != 0 && !warning_done) { warning_done = 1; archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Cannot restore extended attributes on this system"); return (ARCHIVE_WARN); } /* Warning was already emitted; suppress further warnings. */ return (ARCHIVE_OK); } #endif /* * Test if file on disk is older than entry. */ static int older(struct stat *st, struct archive_entry *entry) { /* First, test the seconds and return if we have a definite answer. */ /* Definitely older. */ if (st->st_mtime < archive_entry_mtime(entry)) return (1); /* Definitely younger. */ if (st->st_mtime > archive_entry_mtime(entry)) return (0); /* If this platform supports fractional seconds, try those. */ #if HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC /* Definitely older. */ if (st->st_mtimespec.tv_nsec < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIM_TV_NSEC /* Definitely older. */ if (st->st_mtim.tv_nsec < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIME_N /* older. */ if (st->st_mtime_n < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_UMTIME /* older. */ if (st->st_umtime * 1000 < archive_entry_mtime_nsec(entry)) return (1); #elif HAVE_STRUCT_STAT_ST_MTIME_USEC /* older. */ if (st->st_mtime_usec * 1000 < archive_entry_mtime_nsec(entry)) return (1); #else /* This system doesn't have high-res timestamps. */ #endif /* Same age or newer, so not older. */ return (0); } #endif /* !_WIN32 || __CYGWIN__ */ Index: vendor/libarchive/dist/libarchive/archive_write_set_format_iso9660.c =================================================================== --- vendor/libarchive/dist/libarchive/archive_write_set_format_iso9660.c (revision 231296) +++ vendor/libarchive/dist/libarchive/archive_write_set_format_iso9660.c (revision 231297) @@ -1,8113 +1,8112 @@ /*- * Copyright (c) 2009-2011 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "archive_platform.h" #ifdef HAVE_SYS_TYPES_H #include #endif #ifdef HAVE_SYS_UTSNAME_H #include #endif #ifdef HAVE_ERRNO_H #include #endif #ifdef HAVE_LIMITS_H #include #endif #include #include #ifdef HAVE_STDLIB_H #include #endif #include #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_ZLIB_H #include #endif #include "archive.h" #include "archive_endian.h" #include "archive_entry.h" #include "archive_entry_locale.h" #include "archive_private.h" #include "archive_rb.h" #include "archive_write_private.h" #if defined(_WIN32) && !defined(__CYGWIN__) #define getuid() 0 #define getgid() 0 #endif /*#define DEBUG 1*/ #ifdef DEBUG /* To compare to the ISO image file made by mkisofs. */ #define COMPAT_MKISOFS 1 #endif #define LOGICAL_BLOCK_BITS 11 #define LOGICAL_BLOCK_SIZE 2048 #define PATH_TABLE_BLOCK_SIZE 4096 #define SYSTEM_AREA_BLOCK 16 #define PRIMARY_VOLUME_DESCRIPTOR_BLOCK 1 #define SUPPLEMENTARY_VOLUME_DESCRIPTOR_BLOCK 1 #define BOOT_RECORD_DESCRIPTOR_BLOCK 1 #define VOLUME_DESCRIPTOR_SET_TERMINATOR_BLOCK 1 #define NON_ISO_FILE_SYSTEM_INFORMATION_BLOCK 1 #define RRIP_ER_BLOCK 1 #define PADDING_BLOCK 150 #define FD_1_2M_SIZE (1024 * 1200) #define FD_1_44M_SIZE (1024 * 1440) #define FD_2_88M_SIZE (1024 * 2880) #define MULTI_EXTENT_SIZE (ARCHIVE_LITERAL_LL(1) << 32) /* 4Gi bytes. */ #define MAX_DEPTH 8 #define RR_CE_SIZE 28 /* SUSP "CE" extension size */ #define FILE_FLAG_EXISTENCE 0x01 #define FILE_FLAG_DIRECTORY 0x02 #define FILE_FLAG_ASSOCIATED 0x04 #define FILE_FLAG_RECORD 0x08 #define FILE_FLAG_PROTECTION 0x10 #define FILE_FLAG_MULTI_EXTENT 0x80 static const char rrip_identifier[] = "RRIP_1991A"; static const char rrip_descriptor[] = "THE ROCK RIDGE INTERCHANGE PROTOCOL PROVIDES SUPPORT FOR " "POSIX FILE SYSTEM SEMANTICS"; static const char rrip_source[] = "PLEASE CONTACT DISC PUBLISHER FOR SPECIFICATION SOURCE. " "SEE PUBLISHER IDENTIFIER IN PRIMARY VOLUME DESCRIPTOR FOR " "CONTACT INFORMATION."; #define RRIP_ER_ID_SIZE (sizeof(rrip_identifier)-1) #define RRIP_ER_DSC_SIZE (sizeof(rrip_descriptor)-1) #define RRIP_ER_SRC_SIZE (sizeof(rrip_source)-1) #define RRIP_ER_SIZE (8 + RRIP_ER_ID_SIZE + \ RRIP_ER_DSC_SIZE + RRIP_ER_SRC_SIZE) static const unsigned char zisofs_magic[8] = { 0x37, 0xE4, 0x53, 0x96, 0xC9, 0xDB, 0xD6, 0x07 }; #define ZF_HEADER_SIZE 16 /* zisofs header size. */ #define ZF_LOG2_BS 15 /* log2 block size; 32K bytes. */ #define ZF_BLOCK_SIZE (1UL << ZF_LOG2_BS) /* * Manage extra records. */ struct extr_rec { int location; int offset; unsigned char buf[LOGICAL_BLOCK_SIZE]; struct extr_rec *next; }; struct ctl_extr_rec { int use_extr; unsigned char *bp; struct isoent *isoent; unsigned char *ce_ptr; int cur_len; int dr_len; int limit; int extr_off; int extr_loc; }; #define DR_SAFETY RR_CE_SIZE #define DR_LIMIT (254 - DR_SAFETY) /* * The relation of struct isofile and isoent and archive_entry. * * Primary volume tree --> struct isoent * | * v * struct isofile --> archive_entry * ^ * | * Joliet volume tree --> struct isoent * * struct isoent has specific information for volume. */ struct isofile { /* Used for managing struct isofile list. */ struct isofile *allnext; struct isofile *datanext; /* Used for managing a hardlined struct isofile list. */ struct isofile *hlnext; struct isofile *hardlink_target; struct archive_entry *entry; /* * Used for making a directory tree. */ struct archive_string parentdir; struct archive_string basename; struct archive_string basename_utf16; struct archive_string symlink; int dircnt; /* The number of elements of * its parent directory */ /* * Used for a Directory Record. */ struct content { int64_t offset_of_temp; int64_t size; int blocks; uint32_t location; /* * One extent equals one content. * If this entry has multi extent, `next' variable points * next content data. */ struct content *next; /* next content */ } content, *cur_content; int write_content; enum { NO = 0, BOOT_CATALOG, BOOT_IMAGE, } boot; /* * Used for a zisofs. */ struct { unsigned char header_size; unsigned char log2_bs; uint32_t uncompressed_size; } zisofs; }; struct isoent { /* Keep `rbnode' at the first member of struct isoent. */ struct archive_rb_node rbnode; struct isofile *file; struct isoent *parent; /* A list of children.(use chnext) */ struct { struct isoent *first; struct isoent **last; int cnt; } children; struct archive_rb_tree rbtree; /* A list of sub directories.(use drnext) */ struct { struct isoent *first; struct isoent **last; int cnt; } subdirs; /* A sorted list of sub directories. */ struct isoent **children_sorted; /* Used for managing struct isoent list. */ struct isoent *chnext; struct isoent *drnext; struct isoent *ptnext; /* * Used for making a Directory Record. */ int dir_number; struct { int vd; int self; int parent; int normal; } dr_len; uint32_t dir_location; int dir_block; /* * Identifier: * on primary, ISO9660 file/directory name. * on joliet, UCS2 file/directory name. * ext_off : offset of identifier extension. * ext_len : length of identifier extension. * id_len : byte size of identifier. * on primary, this is ext_off + ext_len + version length. * on joliet, this is ext_off + ext_len. * mb_len : length of multibyte-character of identifier. * on primary, mb_len and id_len are always the same. * on joliet, mb_len and id_len are different. */ char *identifier; int ext_off; int ext_len; int id_len; int mb_len; /* * Used for making a Rockridge extension. * This is a part of Directory Records. */ struct isoent *rr_parent; struct isoent *rr_child; /* Extra Record.(which we call in this source file) * A maximum size of the Directory Record is 254. * so, if generated RRIP data of a file cannot into a Directory * Record because of its size, that surplus data relocate this * Extra Record. */ struct { struct extr_rec *first; struct extr_rec **last; struct extr_rec *current; } extr_rec_list; int virtual:1; /* If set to one, this file type is a directory. * A convenience flag to be used as * "archive_entry_filetype(isoent->file->entry) == AE_IFDIR". */ int dir:1; }; struct hardlink { struct archive_rb_node rbnode; int nlink; struct { struct isofile *first; struct isofile **last; } file_list; }; /* * ISO writer options */ struct iso_option { /* * Usage : abstract-file= * Type : string, max 37 bytes * Default: Not specified * COMPAT : mkisofs -abstract * * Specifies Abstract Filename. * This file shall be described in the Root Directory * and containing a abstract statement. */ unsigned int abstract_file:1; #define OPT_ABSTRACT_FILE_DEFAULT 0 /* Not specified */ #define ABSTRACT_FILE_SIZE 37 /* * Usage : application-id= * Type : string, max 128 bytes * Default: Not specified * COMPAT : mkisofs -A/-appid . * * Specifies Application Identifier. * If the first byte is set to '_'(5F), the remaining * bytes of this option shall specify an identifier * for a file containing the identification of the * application. * This file shall be described in the Root Directory. */ unsigned int application_id:1; #define OPT_APPLICATION_ID_DEFAULT 0 /* Use default identifier */ #define APPLICATION_IDENTIFIER_SIZE 128 /* * Usage : !allow-vernum * Type : boolean * Default: Enabled * : Violates the ISO9660 standard if disable. * COMPAT: mkisofs -N * * Allow filenames to use version numbers. */ unsigned int allow_vernum:1; #define OPT_ALLOW_VERNUM_DEFAULT 1 /* Enabled */ /* * Usage : biblio-file= * Type : string, max 37 bytes * Default: Not specified * COMPAT : mkisofs -biblio * * Specifies Bibliographic Filename. * This file shall be described in the Root Directory * and containing bibliographic records. */ unsigned int biblio_file:1; #define OPT_BIBLIO_FILE_DEFAULT 0 /* Not specified */ #define BIBLIO_FILE_SIZE 37 /* * Usage : boot= * Type : string * Default: Not specified * COMPAT : mkisofs -b/-eltorito-boot * * Specifies "El Torito" boot image file to make * a bootable CD. */ unsigned int boot:1; #define OPT_BOOT_DEFAULT 0 /* Not specified */ /* * Usage : boot-catalog= * Type : string * Default: "boot.catalog" * COMPAT : mkisofs -c/-eltorito-catalog * * Specifies a fullpath of El Torito boot catalog. */ unsigned int boot_catalog:1; #define OPT_BOOT_CATALOG_DEFAULT 0 /* Not specified */ /* * Usage : boot-info-table * Type : boolean * Default: Disabled * COMPAT : mkisofs -boot-info-table * * Modify the boot image file specified by `boot' * option; ISO writer stores boot file information * into the boot file in ISO image at offset 8 * through offset 64. */ unsigned int boot_info_table:1; #define OPT_BOOT_INFO_TABLE_DEFAULT 0 /* Disabled */ /* * Usage : boot-load-seg= * Type : hexadecimal * Default: Not specified * COMPAT : mkisofs -boot-load-seg * * Specifies a load segment for boot image. * This is used with no-emulation mode. */ unsigned int boot_load_seg:1; #define OPT_BOOT_LOAD_SEG_DEFAULT 0 /* Not specified */ /* * Usage : boot-load-size= * Type : decimal * Default: Not specified * COMPAT : mkisofs -boot-load-size * * Specifies a sector count for boot image. * This is used with no-emulation mode. */ unsigned int boot_load_size:1; #define OPT_BOOT_LOAD_SIZE_DEFAULT 0 /* Not specified */ /* * Usage : boot-type= * : 'no-emulation' : 'no emulation' image * : 'fd' : floppy disk image * : 'hard-disk' : hard disk image * Type : string * Default: Auto detect * : We check a size of boot image; * : If ths size is just 1.22M/1.44M/2.88M, * : we assume boot_type is 'fd'; * : otherwise boot_type is 'no-emulation'. * COMPAT : * boot=no-emulation * mkisofs -no-emul-boot * boot=fd * This is a default on the mkisofs. * boot=hard-disk * mkisofs -hard-disk-boot * * Specifies a type of "El Torito" boot image. */ unsigned int boot_type:2; #define OPT_BOOT_TYPE_AUTO 0 /* auto detect */ #define OPT_BOOT_TYPE_NO_EMU 1 /* ``no emulation'' image */ #define OPT_BOOT_TYPE_FD 2 /* floppy disk image */ #define OPT_BOOT_TYPE_HARD_DISK 3 /* hard disk image */ #define OPT_BOOT_TYPE_DEFAULT OPT_BOOT_TYPE_AUTO /* * Usage : compression-level= * Type : decimal * Default: Not specified * COMPAT : NONE * * Specifies compression level for option zisofs=direct. */ unsigned int compression_level:1; #define OPT_COMPRESSION_LEVEL_DEFAULT 0 /* Not specified */ /* * Usage : copyright-file= * Type : string, max 37 bytes * Default: Not specified * COMPAT : mkisofs -copyright * * Specifies Copyright Filename. * This file shall be described in the Root Directory * and containing a copyright statement. */ unsigned int copyright_file:1; #define OPT_COPYRIGHT_FILE_DEFAULT 0 /* Not specified */ #define COPYRIGHT_FILE_SIZE 37 /* * Usage : gid= * Type : decimal * Default: Not specified * COMPAT : mkisofs -gid * * Specifies a group id to rewrite the group id of all files. */ unsigned int gid:1; #define OPT_GID_DEFAULT 0 /* Not specified */ /* * Usage : iso-level=[1234] * Type : decimal * Default: 1 * COMPAT : mkisofs -iso-level * * Specifies ISO9600 Level. * Level 1: [DEFAULT] * - limits each file size less than 4Gi bytes; * - a File Name shall not contain more than eight * d-characters or eight d1-characters; * - a File Name Extension shall not contain more than * three d-characters or three d1-characters; * - a Directory Identifier shall not contain more * than eight d-characters or eight d1-characters. * Level 2: * - limits each file size less than 4Giga bytes; * - a File Name shall not contain more than thirty * d-characters or thirty d1-characters; * - a File Name Extension shall not contain more than * thirty d-characters or thirty d1-characters; * - a Directory Identifier shall not contain more * than thirty-one d-characters or thirty-one * d1-characters. * Level 3: * - no limit of file size; use multi extent. * Level 4: * - this level 4 simulates mkisofs option * '-iso-level 4'; * - crate a enhanced volume as mkisofs doing; * - allow a File Name to have leading dot; * - allow a File Name to have all ASCII letters; * - allow a File Name to have multiple dots; * - allow more then 8 depths of directory trees; * - disable a version number to a File Name; * - disable a forced period to the tail of a File Name; * - the maxinum length of files and directories is raised to 193. * if rockridge option is disabled, raised to 207. */ unsigned int iso_level:3; #define OPT_ISO_LEVEL_DEFAULT 1 /* ISO Level 1 */ /* * Usage : joliet[=long] * : !joliet * : Do not generate Joliet Volume and Records. * : joliet [DEFAULT] * : Generates Joliet Volume and Directory Records. * : [COMPAT: mkisofs -J/-joliet] * : joliet=long * : The joliet filenames are up to 103 Unicode * : characters. * : This option breaks the Joliet specification. * : [COMPAT: mkisofs -J -joliet-long] * Type : boolean/string * Default: Enabled * COMPAT : mkisofs -J / -joliet-long * * Generates Joliet Volume and Directory Records. */ unsigned int joliet:2; #define OPT_JOLIET_DISABLE 0 /* Not generate Joliet Records. */ #define OPT_JOLIET_ENABLE 1 /* Generate Joliet Records. */ #define OPT_JOLIET_LONGNAME 2 /* Use long joliet filenames.*/ #define OPT_JOLIET_DEFAULT OPT_JOLIET_ENABLE /* * Usage : !limit-depth * Type : boolean * Default: Enabled * : Violates the ISO9660 standard if disable. * COMPAT : mkisofs -D/-disable-deep-relocation * * The number of levels in hierarchy cannot exceed eight. */ unsigned int limit_depth:1; #define OPT_LIMIT_DEPTH_DEFAULT 1 /* Enabled */ /* * Usage : !limit-dirs * Type : boolean * Default: Enabled * : Violates the ISO9660 standard if disable. * COMPAT : mkisofs -no-limit-pathtables * * Limits the number of directories less than 65536 due * to the size of the Parent Directory Number of Path * Table. */ unsigned int limit_dirs:1; #define OPT_LIMIT_DIRS_DEFAULT 1 /* Enabled */ /* * Usage : !pad * Type : boolean * Default: Enabled * COMPAT : -pad/-no-pad * * Pads the end of the ISO image by null of 300Ki bytes. */ unsigned int pad:1; #define OPT_PAD_DEFAULT 1 /* Enabled */ /* * Usage : publisher= * Type : string, max 128 bytes * Default: Not specified * COMPAT : mkisofs -publisher * * Specifies Publisher Identifier. * If the first byte is set to '_'(5F), the remaining * bytes of this option shall specify an identifier * for a file containing the identification of the user. * This file shall be described in the Root Directory. */ unsigned int publisher:1; #define OPT_PUBLISHER_DEFAULT 0 /* Not specified */ #define PUBLISHER_IDENTIFIER_SIZE 128 /* * Usage : rockridge * : !rockridge * : disable to generate SUSP and RR records. * : rockridge * : the same as 'rockridge=useful'. * : rockridge=strict * : generate SUSP and RR records. * : [COMPAT: mkisofs -R] * : rockridge=useful [DEFAULT] * : generate SUSP and RR records. * : [COMPAT: mkisofs -r] * : NOTE Our rockridge=useful option does not set a zero * : to uid and gid, you should use application * : option such as --gid,--gname,--uid and --uname * : badtar options instead. * Type : boolean/string * Default: Enabled as rockridge=useful * COMPAT : mkisofs -r / -R * * Generates SUSP and RR records. */ unsigned int rr:2; #define OPT_RR_DISABLED 0 #define OPT_RR_STRICT 1 #define OPT_RR_USEFUL 2 #define OPT_RR_DEFAULT OPT_RR_USEFUL /* * Usage : volume-id= * Type : string, max 32 bytes * Default: Not specified * COMPAT : mkisofs -V * * Specifies Volume Identifier. */ unsigned int volume_id:1; #define OPT_VOLUME_ID_DEFAULT 0 /* Use default identifier */ #define VOLUME_IDENTIFIER_SIZE 32 /* * Usage : !zisofs [DEFAULT] * : Disable to generate RRIP 'ZF' extension. * : zisofs * : Make files zisofs file and generate RRIP 'ZF' * : extension. So you do not need mkzftree utility * : for making zisofs. * : When the file size is less than one Logical Block * : size, that file will not zisofs'ed since it does * : reduece an ISO-image size. * : * : When you specify option 'boot=', that * : 'boot-image' file won't be converted to zisofs file. * Type : boolean * Default: Disabled * * Generates RRIP 'ZF' System Use Entry. */ unsigned int zisofs:1; #define OPT_ZISOFS_DISABLED 0 #define OPT_ZISOFS_DIRECT 1 #define OPT_ZISOFS_DEFAULT OPT_ZISOFS_DISABLED }; struct iso9660 { /* The creation time of ISO image. */ time_t birth_time; /* A file stream of a temporary file, which file contents * save to until ISO iamge can be created. */ int temp_fd; struct isofile *cur_file; struct isoent *cur_dirent; struct archive_string cur_dirstr; uint64_t bytes_remaining; int need_multi_extent; /* Temporary string buffer for Joliet extension. */ struct archive_string utf16be; struct archive_string mbs; struct archive_string_conv *sconv_to_utf16be; struct archive_string_conv *sconv_from_utf16be; /* A list of all of struct isofile entries. */ struct { struct isofile *first; struct isofile **last; } all_file_list; /* A list of struct isofile entries which have its * contents and are not a directory, a hardlined file * and a symlink file. */ struct { struct isofile *first; struct isofile **last; } data_file_list; /* Used for managing to find hardlinking files. */ struct archive_rb_tree hardlink_rbtree; /* Used for making the Path Table Record. */ struct vdd { /* the root of entry tree. */ struct isoent *rootent; enum vdd_type { VDD_PRIMARY, VDD_JOLIET, VDD_ENHANCED } vdd_type; struct path_table { struct isoent *first; struct isoent **last; struct isoent **sorted; int cnt; } *pathtbl; int max_depth; int path_table_block; int path_table_size; int location_type_L_path_table; int location_type_M_path_table; int total_dir_block; } primary, joliet; /* Used for making a Volume Descriptor. */ int volume_space_size; int volume_sequence_number; int total_file_block; struct archive_string volume_identifier; struct archive_string publisher_identifier; struct archive_string data_preparer_identifier; struct archive_string application_identifier; struct archive_string copyright_file_identifier; struct archive_string abstract_file_identifier; struct archive_string bibliographic_file_identifier; /* Used for making rockridge extensions. */ int location_rrip_er; /* Used for making zisofs. */ struct { int detect_magic:1; int making:1; int allzero:1; unsigned char magic_buffer[64]; int magic_cnt; #ifdef HAVE_ZLIB_H /* * Copy a compressed file to iso9660.zisofs.temp_fd * and also copy a uncompressed file(original file) to * iso9660.temp_fd . If the number of logical block * of the compressed file is less than the number of * logical block of the uncompressed file, use it and * remove the copy of the uncompressed file. * but if not, we use uncompressed file and remove * the copy of the compressed file. */ uint32_t *block_pointers; size_t block_pointers_allocated; int block_pointers_cnt; int block_pointers_idx; int64_t total_size; int64_t block_offset; z_stream stream; int stream_valid; int64_t remaining; int compression_level; #endif } zisofs; struct isoent *directories_too_deep; int dircnt_max; /* Write buffer. */ #define wb_buffmax() (LOGICAL_BLOCK_SIZE * 32) #define wb_remaining(a) (((struct iso9660 *)(a)->format_data)->wbuff_remaining) #define wb_offset(a) (((struct iso9660 *)(a)->format_data)->wbuff_offset \ + wb_buffmax() - wb_remaining(a)) unsigned char wbuff[LOGICAL_BLOCK_SIZE * 32]; size_t wbuff_remaining; enum { WB_TO_STREAM, WB_TO_TEMP } wbuff_type; int64_t wbuff_offset; int64_t wbuff_written; int64_t wbuff_tail; /* 'El Torito' boot data. */ struct { /* boot catalog file */ struct archive_string catalog_filename; struct isoent *catalog; /* boot image file */ struct archive_string boot_filename; struct isoent *boot; unsigned char platform_id; #define BOOT_PLATFORM_X86 0 #define BOOT_PLATFORM_PPC 1 #define BOOT_PLATFORM_MAC 2 struct archive_string id; unsigned char media_type; #define BOOT_MEDIA_NO_EMULATION 0 #define BOOT_MEDIA_1_2M_DISKETTE 1 #define BOOT_MEDIA_1_44M_DISKETTE 2 #define BOOT_MEDIA_2_88M_DISKETTE 3 #define BOOT_MEDIA_HARD_DISK 4 unsigned char system_type; uint16_t boot_load_seg; uint16_t boot_load_size; #define BOOT_LOAD_SIZE 4 } el_torito; struct iso_option opt; }; /* * Types of Volume Descriptor */ enum VD_type { VDT_BOOT_RECORD=0, /* Boot Record Volume Descriptor */ VDT_PRIMARY=1, /* Primary Volume Descriptor */ VDT_SUPPLEMENTARY=2, /* Supplementary Volume Descriptor */ VDT_TERMINATOR=255 /* Volume Descriptor Set Terminator */ }; /* * Types of Directory Record */ enum dir_rec_type { DIR_REC_VD, /* Stored in Volume Descriptor. */ DIR_REC_SELF, /* Stored as Current Directory. */ DIR_REC_PARENT, /* Stored as Parent Directory. */ DIR_REC_NORMAL, /* Stored as Child. */ }; /* * Kinds of Volume Descriptor Character */ enum vdc { VDC_STD, VDC_LOWERCASE, VDC_UCS2, VDC_UCS2_DIRECT, }; /* * IDentifier Resolver. * Used for resolving duplicated filenames. */ struct idr { struct idrent { struct archive_rb_node rbnode; /* Used in wait_list. */ struct idrent *wnext; struct idrent *avail; struct isoent *isoent; int weight; int noff; int rename_num; } *idrent_pool; struct archive_rb_tree rbtree; struct { struct idrent *first; struct idrent **last; } wait_list; int pool_size; int pool_idx; int num_size; int null_size; char char_map[0x80]; }; enum char_type { A_CHAR, D_CHAR, }; static int iso9660_options(struct archive_write *, const char *, const char *); static int iso9660_write_header(struct archive_write *, struct archive_entry *); static ssize_t iso9660_write_data(struct archive_write *, const void *, size_t); static int iso9660_finish_entry(struct archive_write *); static int iso9660_close(struct archive_write *); static int iso9660_free(struct archive_write *); static void get_system_identitier(char *, size_t); static void set_str(unsigned char *, const char *, size_t, char, const char *); static inline int joliet_allowed_char(unsigned char, unsigned char); static int set_str_utf16be(struct archive_write *, unsigned char *, const char *, size_t, uint16_t, enum vdc); static int set_str_a_characters_bp(struct archive_write *, unsigned char *, int, int, const char *, enum vdc); static int set_str_d_characters_bp(struct archive_write *, unsigned char *, int, int, const char *, enum vdc); static void set_VD_bp(unsigned char *, enum VD_type, unsigned char); static inline void set_unused_field_bp(unsigned char *, int, int); static unsigned char *extra_open_record(unsigned char *, int, struct isoent *, struct ctl_extr_rec *); static void extra_close_record(struct ctl_extr_rec *, int); static unsigned char * extra_next_record(struct ctl_extr_rec *, int); static unsigned char *extra_get_record(struct isoent *, int *, int *, int *); static void extra_tell_used_size(struct ctl_extr_rec *, int); static int extra_setup_location(struct isoent *, int); static int set_directory_record_rr(unsigned char *, int, struct isoent *, struct iso9660 *, enum dir_rec_type); static int set_directory_record(unsigned char *, size_t, struct isoent *, struct iso9660 *, enum dir_rec_type, enum vdd_type); static inline int get_dir_rec_size(struct iso9660 *, struct isoent *, enum dir_rec_type, enum vdd_type); static inline unsigned char *wb_buffptr(struct archive_write *); static int wb_write_out(struct archive_write *); static int wb_consume(struct archive_write *, size_t); #ifdef HAVE_ZLIB_H static int wb_set_offset(struct archive_write *, int64_t); #endif static int write_null(struct archive_write *, size_t); static int write_VD_terminator(struct archive_write *); static int set_file_identifier(unsigned char *, int, int, enum vdc, struct archive_write *, struct vdd *, struct archive_string *, const char *, int, enum char_type); static int write_VD(struct archive_write *, struct vdd *); static int write_VD_boot_record(struct archive_write *); static int write_information_block(struct archive_write *); static int write_path_table(struct archive_write *, int, struct vdd *); static int write_directory_descriptors(struct archive_write *, struct vdd *); static int write_file_descriptors(struct archive_write *); static int write_rr_ER(struct archive_write *); static void calculate_path_table_size(struct vdd *); static void isofile_init_entry_list(struct iso9660 *); static void isofile_add_entry(struct iso9660 *, struct isofile *); static void isofile_free_all_entries(struct iso9660 *); static void isofile_init_entry_data_file_list(struct iso9660 *); static void isofile_add_data_file(struct iso9660 *, struct isofile *); static struct isofile * isofile_new(struct archive_write *, struct archive_entry *); static void isofile_free(struct isofile *); static int isofile_gen_utility_names(struct archive_write *, struct isofile *); static int isofile_register_hardlink(struct archive_write *, struct isofile *); static void isofile_connect_hardlink_files(struct iso9660 *); static void isofile_init_hardlinks(struct iso9660 *); static void isofile_free_hardlinks(struct iso9660 *); static struct isoent *isoent_new(struct isofile *); static int isoent_clone_tree(struct archive_write *, struct isoent **, struct isoent *); static void _isoent_free(struct isoent *isoent); static void isoent_free_all(struct isoent *); static struct isoent * isoent_create_virtual_dir(struct archive_write *, struct iso9660 *, const char *); static int isoent_cmp_node(const struct archive_rb_node *, const struct archive_rb_node *); static int isoent_cmp_key(const struct archive_rb_node *, const void *); static int isoent_add_child_head(struct isoent *, struct isoent *); static int isoent_add_child_tail(struct isoent *, struct isoent *); static void isoent_remove_child(struct isoent *, struct isoent *); static void isoent_setup_directory_location(struct iso9660 *, int, struct vdd *); static void isoent_setup_file_location(struct iso9660 *, int); static int get_path_component(char *, int, const char *); static int isoent_tree(struct archive_write *, struct isoent **); static struct isoent *isoent_find_child(struct isoent *, const char *); static struct isoent *isoent_find_entry(struct isoent *, const char *); static void idr_relaxed_filenames(char *); static void idr_init(struct iso9660 *, struct vdd *, struct idr *); static void idr_cleanup(struct idr *); static int idr_ensure_poolsize(struct archive_write *, struct idr *, int); static int idr_start(struct archive_write *, struct idr *, int, int, int, int, const struct archive_rb_tree_ops *); static void idr_register(struct idr *, struct isoent *, int, int); static void idr_extend_identifier(struct idrent *, int, int); static void idr_resolve(struct idr *, void (*)(unsigned char *, int)); static void idr_set_num(unsigned char *, int); static void idr_set_num_beutf16(unsigned char *, int); static int isoent_gen_iso9660_identifier(struct archive_write *, struct isoent *, struct idr *); static int isoent_gen_joliet_identifier(struct archive_write *, struct isoent *, struct idr *); static int isoent_cmp_iso9660_identifier(const struct isoent *, const struct isoent *); static int isoent_cmp_node_iso9660(const struct archive_rb_node *, const struct archive_rb_node *); static int isoent_cmp_key_iso9660(const struct archive_rb_node *, const void *); static int isoent_cmp_joliet_identifier(const struct isoent *, const struct isoent *); static int isoent_cmp_node_joliet(const struct archive_rb_node *, const struct archive_rb_node *); static int isoent_cmp_key_joliet(const struct archive_rb_node *, const void *); static inline void path_table_add_entry(struct path_table *, struct isoent *); static inline struct isoent * path_table_last_entry(struct path_table *); static int isoent_make_path_table(struct archive_write *); static int isoent_find_out_boot_file(struct archive_write *, struct isoent *); static int isoent_create_boot_catalog(struct archive_write *, struct isoent *); static size_t fd_boot_image_size(int); static int make_boot_catalog(struct archive_write *); static int setup_boot_information(struct archive_write *); static int zisofs_init(struct archive_write *, struct isofile *); static void zisofs_detect_magic(struct archive_write *, const void *, size_t); static int zisofs_write_to_temp(struct archive_write *, const void *, size_t); static int zisofs_finish_entry(struct archive_write *); static int zisofs_rewind_boot_file(struct archive_write *); static int zisofs_free(struct archive_write *); int archive_write_set_format_iso9660(struct archive *_a) { struct archive_write *a = (struct archive_write *)_a; struct iso9660 *iso9660; archive_check_magic(_a, ARCHIVE_WRITE_MAGIC, ARCHIVE_STATE_NEW, "archive_write_set_format_iso9660"); /* If another format was already registered, unregister it. */ if (a->format_free != NULL) (a->format_free)(a); iso9660 = calloc(1, sizeof(*iso9660)); if (iso9660 == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate iso9660 data"); return (ARCHIVE_FATAL); } iso9660->birth_time = 0; iso9660->temp_fd = -1; iso9660->cur_file = NULL; iso9660->primary.max_depth = 0; iso9660->primary.vdd_type = VDD_PRIMARY; iso9660->primary.pathtbl = NULL; iso9660->joliet.rootent = NULL; iso9660->joliet.max_depth = 0; iso9660->joliet.vdd_type = VDD_JOLIET; iso9660->joliet.pathtbl = NULL; isofile_init_entry_list(iso9660); isofile_init_entry_data_file_list(iso9660); isofile_init_hardlinks(iso9660); iso9660->directories_too_deep = NULL; iso9660->dircnt_max = 1; iso9660->wbuff_remaining = wb_buffmax(); iso9660->wbuff_type = WB_TO_TEMP; iso9660->wbuff_offset = 0; iso9660->wbuff_written = 0; iso9660->wbuff_tail = 0; archive_string_init(&(iso9660->utf16be)); archive_string_init(&(iso9660->mbs)); /* * Init Identifiers used for PVD and SVD. */ archive_string_init(&(iso9660->volume_identifier)); archive_strcpy(&(iso9660->volume_identifier), "CDROM"); archive_string_init(&(iso9660->publisher_identifier)); archive_string_init(&(iso9660->data_preparer_identifier)); archive_string_init(&(iso9660->application_identifier)); archive_strcpy(&(iso9660->application_identifier), archive_version_string()); archive_string_init(&(iso9660->copyright_file_identifier)); archive_string_init(&(iso9660->abstract_file_identifier)); archive_string_init(&(iso9660->bibliographic_file_identifier)); /* * Init El Torito bootable CD variables. */ archive_string_init(&(iso9660->el_torito.catalog_filename)); iso9660->el_torito.catalog = NULL; /* Set default file name of boot catalog */ archive_strcpy(&(iso9660->el_torito.catalog_filename), "boot.catalog"); archive_string_init(&(iso9660->el_torito.boot_filename)); iso9660->el_torito.boot = NULL; iso9660->el_torito.platform_id = BOOT_PLATFORM_X86; archive_string_init(&(iso9660->el_torito.id)); iso9660->el_torito.boot_load_seg = 0; iso9660->el_torito.boot_load_size = BOOT_LOAD_SIZE; /* * Init zisofs variables. */ #ifdef HAVE_ZLIB_H iso9660->zisofs.block_pointers = NULL; iso9660->zisofs.block_pointers_allocated = 0; iso9660->zisofs.stream_valid = 0; iso9660->zisofs.compression_level = 9; memset(&(iso9660->zisofs.stream), 0, sizeof(iso9660->zisofs.stream)); #endif /* * Set default value of iso9660 options. */ iso9660->opt.abstract_file = OPT_ABSTRACT_FILE_DEFAULT; iso9660->opt.application_id = OPT_APPLICATION_ID_DEFAULT; iso9660->opt.allow_vernum = OPT_ALLOW_VERNUM_DEFAULT; iso9660->opt.biblio_file = OPT_BIBLIO_FILE_DEFAULT; iso9660->opt.boot = OPT_BOOT_DEFAULT; iso9660->opt.boot_catalog = OPT_BOOT_CATALOG_DEFAULT; iso9660->opt.boot_info_table = OPT_BOOT_INFO_TABLE_DEFAULT; iso9660->opt.boot_load_seg = OPT_BOOT_LOAD_SEG_DEFAULT; iso9660->opt.boot_load_size = OPT_BOOT_LOAD_SIZE_DEFAULT; iso9660->opt.boot_type = OPT_BOOT_TYPE_DEFAULT; iso9660->opt.compression_level = OPT_COMPRESSION_LEVEL_DEFAULT; iso9660->opt.copyright_file = OPT_COPYRIGHT_FILE_DEFAULT; iso9660->opt.iso_level = OPT_ISO_LEVEL_DEFAULT; iso9660->opt.joliet = OPT_JOLIET_DEFAULT; iso9660->opt.limit_depth = OPT_LIMIT_DEPTH_DEFAULT; iso9660->opt.limit_dirs = OPT_LIMIT_DIRS_DEFAULT; iso9660->opt.pad = OPT_PAD_DEFAULT; iso9660->opt.publisher = OPT_PUBLISHER_DEFAULT; iso9660->opt.rr = OPT_RR_DEFAULT; iso9660->opt.volume_id = OPT_VOLUME_ID_DEFAULT; iso9660->opt.zisofs = OPT_ZISOFS_DEFAULT; /* Create the root directory. */ iso9660->primary.rootent = isoent_create_virtual_dir(a, iso9660, ""); if (iso9660->primary.rootent == NULL) { free(iso9660); archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } iso9660->primary.rootent->parent = iso9660->primary.rootent; iso9660->cur_dirent = iso9660->primary.rootent; archive_string_init(&(iso9660->cur_dirstr)); archive_string_ensure(&(iso9660->cur_dirstr), 1); iso9660->cur_dirstr.s[0] = 0; iso9660->sconv_to_utf16be = NULL; iso9660->sconv_from_utf16be = NULL; a->format_data = iso9660; a->format_name = "iso9660"; a->format_options = iso9660_options; a->format_write_header = iso9660_write_header; a->format_write_data = iso9660_write_data; a->format_finish_entry = iso9660_finish_entry; a->format_close = iso9660_close; a->format_free = iso9660_free; a->archive.archive_format = ARCHIVE_FORMAT_ISO9660; a->archive.archive_format_name = "ISO9660"; return (ARCHIVE_OK); } static int get_str_opt(struct archive_write *a, struct archive_string *s, size_t maxsize, const char *key, const char *value) { if (strlen(value) > maxsize) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Value is longer than %zu characters " "for option ``%s''", maxsize, key); return (ARCHIVE_FATAL); } archive_strcpy(s, value); return (ARCHIVE_OK); } static int get_num_opt(struct archive_write *a, int *num, int high, int low, const char *key, const char *value) { const char *p = value; int data = 0; int neg = 0; if (p == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value(empty) for option ``%s''", key); return (ARCHIVE_FATAL); } if (*p == '-') { neg = 1; p++; } while (*p) { if (*p >= '0' && *p <= '9') data = data * 10 + *p - '0'; else { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value for option ``%s''", key); return (ARCHIVE_FATAL); } if (data > high) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value(over %d) for " "option ``%s''", high, key); return (ARCHIVE_FATAL); } if (data < low) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value(under %d) for " "option ``%s''", low, key); return (ARCHIVE_FATAL); } p++; } if (neg) data *= -1; *num = data; return (ARCHIVE_OK); } static int iso9660_options(struct archive_write *a, const char *key, const char *value) { struct iso9660 *iso9660 = a->format_data; const char *p; int r; switch (key[0]) { case 'a': if (strcmp(key, "abstract-file") == 0) { r = get_str_opt(a, &(iso9660->abstract_file_identifier), ABSTRACT_FILE_SIZE, key, value); iso9660->opt.abstract_file = r == ARCHIVE_OK; return (r); } if (strcmp(key, "application-id") == 0) { r = get_str_opt(a, &(iso9660->application_identifier), APPLICATION_IDENTIFIER_SIZE, key, value); iso9660->opt.application_id = r == ARCHIVE_OK; return (r); } if (strcmp(key, "allow-vernum") == 0) { iso9660->opt.allow_vernum = value != NULL; return (ARCHIVE_OK); } break; case 'b': if (strcmp(key, "biblio-file") == 0) { r = get_str_opt(a, &(iso9660->bibliographic_file_identifier), BIBLIO_FILE_SIZE, key, value); iso9660->opt.biblio_file = r == ARCHIVE_OK; return (r); } if (strcmp(key, "boot") == 0) { if (value == NULL) iso9660->opt.boot = 0; else { iso9660->opt.boot = 1; archive_strcpy( &(iso9660->el_torito.boot_filename), value); } return (ARCHIVE_OK); } if (strcmp(key, "boot-catalog") == 0) { r = get_str_opt(a, &(iso9660->el_torito.catalog_filename), 1024, key, value); iso9660->opt.boot_catalog = r == ARCHIVE_OK; return (r); } if (strcmp(key, "boot-info-table") == 0) { iso9660->opt.boot_info_table = value != NULL; return (ARCHIVE_OK); } if (strcmp(key, "boot-load-seg") == 0) { uint32_t seg; iso9660->opt.boot_load_seg = 0; if (value == NULL) goto invalid_value; seg = 0; p = value; if (p[0] == '0' && (p[1] == 'x' || p[1] == 'X')) p += 2; while (*p) { if (seg) seg <<= 4; if (*p >= 'A' && *p <= 'F') seg += *p - 'A' + 0x0a; else if (*p >= 'a' && *p <= 'f') seg += *p - 'a' + 0x0a; else if (*p >= '0' && *p <= '9') seg += *p - '0'; else goto invalid_value; if (seg > 0xffff) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value(over 0xffff) for " "option ``%s''", key); return (ARCHIVE_FATAL); } p++; } iso9660->el_torito.boot_load_seg = (uint16_t)seg; iso9660->opt.boot_load_seg = 1; return (ARCHIVE_OK); } if (strcmp(key, "boot-load-size") == 0) { int num = 0; r = get_num_opt(a, &num, 0xffff, 1, key, value); iso9660->opt.boot_load_size = r == ARCHIVE_OK; if (r != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->el_torito.boot_load_size = (uint16_t)num; return (ARCHIVE_OK); } if (strcmp(key, "boot-type") == 0) { if (value == NULL) goto invalid_value; if (strcmp(value, "no-emulation") == 0) iso9660->opt.boot_type = OPT_BOOT_TYPE_NO_EMU; else if (strcmp(value, "fd") == 0) iso9660->opt.boot_type = OPT_BOOT_TYPE_FD; else if (strcmp(value, "hard-disk") == 0) iso9660->opt.boot_type = OPT_BOOT_TYPE_HARD_DISK; else goto invalid_value; return (ARCHIVE_OK); } break; case 'c': if (strcmp(key, "compression-level") == 0) { #ifdef HAVE_ZLIB_H if (value == NULL || !(value[0] >= '0' && value[0] <= '9') || value[1] != '\0') goto invalid_value; iso9660->zisofs.compression_level = value[0] - '0'; iso9660->opt.compression_level = 1; return (ARCHIVE_OK); #else archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Option ``%s'' " "is not supported on this platform.", key); return (ARCHIVE_FATAL); #endif } if (strcmp(key, "copyright-file") == 0) { r = get_str_opt(a, &(iso9660->copyright_file_identifier), COPYRIGHT_FILE_SIZE, key, value); iso9660->opt.copyright_file = r == ARCHIVE_OK; return (r); } #ifdef DEBUG /* Specifies Volume creation date and time; * year(4),month(2),day(2),hour(2),minute(2),second(2). * e.g. "20090929033757" */ if (strcmp(key, "creation") == 0) { struct tm tm; char buf[5]; p = value; if (p == NULL || strlen(p) < 14) goto invalid_value; memset(&tm, 0, sizeof(tm)); memcpy(buf, p, 4); buf[4] = '\0'; p += 4; tm.tm_year = strtol(buf, NULL, 10) - 1900; memcpy(buf, p, 2); buf[2] = '\0'; p += 2; tm.tm_mon = strtol(buf, NULL, 10) - 1; memcpy(buf, p, 2); buf[2] = '\0'; p += 2; tm.tm_mday = strtol(buf, NULL, 10); memcpy(buf, p, 2); buf[2] = '\0'; p += 2; tm.tm_hour = strtol(buf, NULL, 10); memcpy(buf, p, 2); buf[2] = '\0'; p += 2; tm.tm_min = strtol(buf, NULL, 10); memcpy(buf, p, 2); buf[2] = '\0'; tm.tm_sec = strtol(buf, NULL, 10); iso9660->birth_time = mktime(&tm); return (ARCHIVE_OK); } #endif break; case 'i': if (strcmp(key, "iso-level") == 0) { if (value != NULL && value[1] == '\0' && (value[0] >= '1' && value[0] <= '4')) { iso9660->opt.iso_level = value[0]-'0'; return (ARCHIVE_OK); } goto invalid_value; } break; case 'j': if (strcmp(key, "joliet") == 0) { if (value == NULL) iso9660->opt.joliet = OPT_JOLIET_DISABLE; else if (strcmp(value, "1") == 0) iso9660->opt.joliet = OPT_JOLIET_ENABLE; else if (strcmp(value, "long") == 0) iso9660->opt.joliet = OPT_JOLIET_LONGNAME; else goto invalid_value; return (ARCHIVE_OK); } break; case 'l': if (strcmp(key, "limit-depth") == 0) { iso9660->opt.limit_depth = value != NULL; return (ARCHIVE_OK); } if (strcmp(key, "limit-dirs") == 0) { iso9660->opt.limit_dirs = value != NULL; return (ARCHIVE_OK); } break; case 'p': if (strcmp(key, "pad") == 0) { iso9660->opt.pad = value != NULL; return (ARCHIVE_OK); } if (strcmp(key, "publisher") == 0) { r = get_str_opt(a, &(iso9660->publisher_identifier), PUBLISHER_IDENTIFIER_SIZE, key, value); iso9660->opt.publisher = r == ARCHIVE_OK; return (r); } break; case 'r': if (strcmp(key, "rockridge") == 0 || strcmp(key, "Rockridge") == 0) { if (value == NULL) iso9660->opt.rr = OPT_RR_DISABLED; else if (strcmp(value, "1") == 0) iso9660->opt.rr = OPT_RR_USEFUL; else if (strcmp(value, "strict") == 0) iso9660->opt.rr = OPT_RR_STRICT; else if (strcmp(value, "useful") == 0) iso9660->opt.rr = OPT_RR_USEFUL; else goto invalid_value; return (ARCHIVE_OK); } break; case 'v': if (strcmp(key, "volume-id") == 0) { r = get_str_opt(a, &(iso9660->volume_identifier), VOLUME_IDENTIFIER_SIZE, key, value); iso9660->opt.volume_id = r == ARCHIVE_OK; return (r); } break; case 'z': if (strcmp(key, "zisofs") == 0) { if (value == NULL) iso9660->opt.zisofs = OPT_ZISOFS_DISABLED; else { #ifdef HAVE_ZLIB_H iso9660->opt.zisofs = OPT_ZISOFS_DIRECT; #else archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "``zisofs'' " "is not supported on this platform."); return (ARCHIVE_FATAL); #endif } return (ARCHIVE_OK); } break; } invalid_value: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Invalid value for option ``%s''", key); return (ARCHIVE_FAILED); } static int iso9660_write_header(struct archive_write *a, struct archive_entry *entry) { struct iso9660 *iso9660; struct isofile *file; struct isoent *isoent; int r, ret = ARCHIVE_OK; iso9660 = a->format_data; iso9660->cur_file = NULL; iso9660->bytes_remaining = 0; iso9660->need_multi_extent = 0; if (archive_entry_filetype(entry) == AE_IFLNK && iso9660->opt.rr == OPT_RR_DISABLED) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Ignore symlink file."); iso9660->cur_file = NULL; return (ARCHIVE_WARN); } if (archive_entry_filetype(entry) == AE_IFREG && archive_entry_size(entry) >= MULTI_EXTENT_SIZE) { if (iso9660->opt.iso_level < 3) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Ignore over %lld bytes file. " "This file too large.", MULTI_EXTENT_SIZE); iso9660->cur_file = NULL; return (ARCHIVE_WARN); } iso9660->need_multi_extent = 1; } file = isofile_new(a, entry); if (file == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate data"); return (ARCHIVE_FATAL); } r = isofile_gen_utility_names(a, file); if (r < ARCHIVE_WARN) { isofile_free(file); return (r); } else if (r < ret) ret = r; /* * Ignore a path which looks like the top of directory name * since we have already made the root directory of an ISO image. */ if (archive_strlen(&(file->parentdir)) == 0 && archive_strlen(&(file->basename)) == 0) { isofile_free(file); return (r); } isofile_add_entry(iso9660, file); isoent = isoent_new(file); if (isoent == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate data"); return (ARCHIVE_FATAL); } if (isoent->file->dircnt > iso9660->dircnt_max) iso9660->dircnt_max = isoent->file->dircnt; /* Add the current file into tree */ r = isoent_tree(a, &isoent); if (r != ARCHIVE_OK) return (r); /* If there is the same file in tree and * the current file is older than the file in tree. * So we don't need the current file data anymore. */ if (isoent->file != file) return (ARCHIVE_OK); /* Non regular files contents are unneeded to be saved to * temporary files. */ if (archive_entry_filetype(file->entry) != AE_IFREG) return (ret); /* * Set the current file to cur_file to read its contents. */ iso9660->cur_file = file; if (archive_entry_nlink(file->entry) > 1) { r = isofile_register_hardlink(a, file); if (r != ARCHIVE_OK) return (ARCHIVE_FATAL); } /* * Prepare to save the contents of the file. */ if (iso9660->temp_fd < 0) { iso9660->temp_fd = __archive_mktemp(NULL); if (iso9660->temp_fd < 0) { archive_set_error(&a->archive, errno, "Couldn't create temporary file"); return (ARCHIVE_FATAL); } } /* Save an offset of current file in temporary file. */ file->content.offset_of_temp = wb_offset(a); file->cur_content = &(file->content); r = zisofs_init(a, file); if (r < ret) ret = r; iso9660->bytes_remaining = archive_entry_size(file->entry); return (ret); } static int write_to_temp(struct archive_write *a, const void *buff, size_t s) { struct iso9660 *iso9660 = a->format_data; ssize_t written; const unsigned char *b; b = (const unsigned char *)buff; while (s) { written = write(iso9660->temp_fd, b, s); if (written < 0) { archive_set_error(&a->archive, errno, "Can't write to temporary file"); return (ARCHIVE_FATAL); } s -= written; b += written; } return (ARCHIVE_OK); } static int wb_write_to_temp(struct archive_write *a, const void *buff, size_t s) { const char *xp = buff; size_t xs = s; /* * If a written data size is big enough to use system-call * and there is no waiting data, this calls write_to_temp() in * order to reduce a extra memory copy. */ if (wb_remaining(a) == wb_buffmax() && s > (1024 * 16)) { struct iso9660 *iso9660 = (struct iso9660 *)a->format_data; xs = s % LOGICAL_BLOCK_SIZE; iso9660->wbuff_offset += s - xs; if (write_to_temp(a, buff, s - xs) != ARCHIVE_OK) return (ARCHIVE_FATAL); if (xs == 0) return (ARCHIVE_OK); xp += s - xs; } while (xs) { size_t size = xs; if (size > wb_remaining(a)) size = wb_remaining(a); memcpy(wb_buffptr(a), xp, size); if (wb_consume(a, size) != ARCHIVE_OK) return (ARCHIVE_FATAL); xs -= size; xp += size; } return (ARCHIVE_OK); } static int wb_write_padding_to_temp(struct archive_write *a, int64_t csize) { size_t ns; int ret; ns = csize % LOGICAL_BLOCK_SIZE; if (ns != 0) ret = write_null(a, LOGICAL_BLOCK_SIZE - ns); else ret = ARCHIVE_OK; return (ret); } static ssize_t write_iso9660_data(struct archive_write *a, const void *buff, size_t s) { struct iso9660 *iso9660 = a->format_data; size_t ws; if (iso9660->temp_fd < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Couldn't create temporary file"); return (ARCHIVE_FATAL); } ws = s; if (iso9660->need_multi_extent && (iso9660->cur_file->cur_content->size + ws) >= (MULTI_EXTENT_SIZE - LOGICAL_BLOCK_SIZE)) { struct content *con; size_t ts; ts = MULTI_EXTENT_SIZE - LOGICAL_BLOCK_SIZE - iso9660->cur_file->cur_content->size; if (iso9660->zisofs.detect_magic) zisofs_detect_magic(a, buff, ts); if (iso9660->zisofs.making) { if (zisofs_write_to_temp(a, buff, ts) != ARCHIVE_OK) return (ARCHIVE_FATAL); } else { if (wb_write_to_temp(a, buff, ts) != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->cur_file->cur_content->size += ts; } /* Write padding. */ if (wb_write_padding_to_temp(a, iso9660->cur_file->cur_content->size) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Compute the logical block number. */ iso9660->cur_file->cur_content->blocks = (iso9660->cur_file->cur_content->size + LOGICAL_BLOCK_SIZE -1) >> LOGICAL_BLOCK_BITS; /* * Make next extent. */ ws -= ts; buff = (const void *)(((const unsigned char *)buff) + ts); /* Make a content for next extent. */ con = calloc(1, sizeof(*con)); if (con == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate content data"); return (ARCHIVE_FATAL); } con->offset_of_temp = wb_offset(a); iso9660->cur_file->cur_content->next = con; iso9660->cur_file->cur_content = con; #ifdef HAVE_ZLIB_H iso9660->zisofs.block_offset = 0; #endif } if (iso9660->zisofs.detect_magic) zisofs_detect_magic(a, buff, ws); if (iso9660->zisofs.making) { if (zisofs_write_to_temp(a, buff, ws) != ARCHIVE_OK) return (ARCHIVE_FATAL); } else { if (wb_write_to_temp(a, buff, ws) != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->cur_file->cur_content->size += ws; } return (s); } static ssize_t iso9660_write_data(struct archive_write *a, const void *buff, size_t s) { struct iso9660 *iso9660 = a->format_data; ssize_t r; if (iso9660->cur_file == NULL) return (0); if (archive_entry_filetype(iso9660->cur_file->entry) != AE_IFREG) return (0); if (s > iso9660->bytes_remaining) s = iso9660->bytes_remaining; if (s == 0) return (0); r = write_iso9660_data(a, buff, s); if (r > 0) iso9660->bytes_remaining -= r; return (r); } static int iso9660_finish_entry(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; if (iso9660->cur_file == NULL) return (ARCHIVE_OK); if (archive_entry_filetype(iso9660->cur_file->entry) != AE_IFREG) return (ARCHIVE_OK); if (iso9660->cur_file->content.size == 0) return (ARCHIVE_OK); /* If there are unwritten data, write null data instead. */ while (iso9660->bytes_remaining > 0) { size_t s; s = (iso9660->bytes_remaining > a->null_length)? a->null_length: (size_t)iso9660->bytes_remaining; if (write_iso9660_data(a, a->nulls, s) < 0) return (ARCHIVE_FATAL); iso9660->bytes_remaining -= s; } if (iso9660->zisofs.making && zisofs_finish_entry(a) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write padding. */ if (wb_write_padding_to_temp(a, iso9660->cur_file->cur_content->size) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Compute the logical block number. */ iso9660->cur_file->cur_content->blocks = (iso9660->cur_file->cur_content->size + LOGICAL_BLOCK_SIZE -1) >> LOGICAL_BLOCK_BITS; /* Add the current file to data file list. */ isofile_add_data_file(iso9660, iso9660->cur_file); return (ARCHIVE_OK); } static int iso9660_close(struct archive_write *a) { struct iso9660 *iso9660; int ret, blocks; iso9660 = a->format_data; /* * Write remaining data out to the temporary file. */ if (wb_remaining(a) > 0) { ret = wb_write_out(a); if (ret < 0) return (ret); } /* * Preparations... */ #ifdef DEBUG if (iso9660->birth_time == 0) #endif time(&(iso9660->birth_time)); /* * Prepare a bootable ISO image. */ if (iso9660->opt.boot) { /* Find out the boot file entry. */ ret = isoent_find_out_boot_file(a, iso9660->primary.rootent); if (ret < 0) return (ret); /* Reconvert the boot file from zisofs'ed form to * plain form. */ ret = zisofs_rewind_boot_file(a); if (ret < 0) return (ret); /* Write remaining data out to the temporary file. */ if (wb_remaining(a) > 0) { ret = wb_write_out(a); if (ret < 0) return (ret); } /* Create the boot catalog. */ ret = isoent_create_boot_catalog(a, iso9660->primary.rootent); if (ret < 0) return (ret); } /* * Prepare joliet extensions. */ if (iso9660->opt.joliet) { /* Make a new tree for joliet. */ ret = isoent_clone_tree(a, &(iso9660->joliet.rootent), iso9660->primary.rootent); if (ret < 0) return (ret); /* Make sure we have UTF-16BE convertors. * if there is no file entry, convertors are still * uninitilized. */ if (iso9660->sconv_to_utf16be == NULL) { iso9660->sconv_to_utf16be = archive_string_conversion_to_charset( &(a->archive), "UTF-16BE", 1); if (iso9660->sconv_to_utf16be == NULL) /* Couldn't allocate memory */ return (ARCHIVE_FATAL); iso9660->sconv_from_utf16be = archive_string_conversion_from_charset( &(a->archive), "UTF-16BE", 1); if (iso9660->sconv_from_utf16be == NULL) /* Couldn't allocate memory */ return (ARCHIVE_FATAL); } } /* * Make Path Tables. */ ret = isoent_make_path_table(a); if (ret < 0) return (ret); /* * Calculate a total volume size and setup all locations of * contents of an iso9660 image. */ blocks = SYSTEM_AREA_BLOCK + PRIMARY_VOLUME_DESCRIPTOR_BLOCK + VOLUME_DESCRIPTOR_SET_TERMINATOR_BLOCK + NON_ISO_FILE_SYSTEM_INFORMATION_BLOCK; if (iso9660->opt.boot) blocks += BOOT_RECORD_DESCRIPTOR_BLOCK; if (iso9660->opt.joliet) blocks += SUPPLEMENTARY_VOLUME_DESCRIPTOR_BLOCK; if (iso9660->opt.iso_level == 4) blocks += SUPPLEMENTARY_VOLUME_DESCRIPTOR_BLOCK; /* Setup the locations of Path Table. */ iso9660->primary.location_type_L_path_table = blocks; blocks += iso9660->primary.path_table_block; iso9660->primary.location_type_M_path_table = blocks; blocks += iso9660->primary.path_table_block; if (iso9660->opt.joliet) { iso9660->joliet.location_type_L_path_table = blocks; blocks += iso9660->joliet.path_table_block; iso9660->joliet.location_type_M_path_table = blocks; blocks += iso9660->joliet.path_table_block; } /* Setup the locations of directories. */ isoent_setup_directory_location(iso9660, blocks, &(iso9660->primary)); blocks += iso9660->primary.total_dir_block; if (iso9660->opt.joliet) { isoent_setup_directory_location(iso9660, blocks, &(iso9660->joliet)); blocks += iso9660->joliet.total_dir_block; } if (iso9660->opt.rr) { iso9660->location_rrip_er = blocks; blocks += RRIP_ER_BLOCK; } /* Setup the locations of all file contents. */ isoent_setup_file_location(iso9660, blocks); blocks += iso9660->total_file_block; if (iso9660->opt.boot && iso9660->opt.boot_info_table) { ret = setup_boot_information(a); if (ret < 0) return (ret); } /* Now we have a total volume size. */ iso9660->volume_space_size = blocks; if (iso9660->opt.pad) iso9660->volume_space_size += PADDING_BLOCK; iso9660->volume_sequence_number = 1; /* * Write an ISO 9660 image. */ /* Switc to start using wbuff as file buffer. */ iso9660->wbuff_remaining = wb_buffmax(); iso9660->wbuff_type = WB_TO_STREAM; iso9660->wbuff_offset = 0; iso9660->wbuff_written = 0; iso9660->wbuff_tail = 0; /* Write The System Area */ ret = write_null(a, SYSTEM_AREA_BLOCK * LOGICAL_BLOCK_SIZE); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Primary Volume Descriptor */ ret = write_VD(a, &(iso9660->primary)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); if (iso9660->opt.boot) { /* Write Boot Record Volume Descriptor */ ret = write_VD_boot_record(a); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } if (iso9660->opt.iso_level == 4) { /* Write Enhanced Volume Descriptor */ iso9660->primary.vdd_type = VDD_ENHANCED; ret = write_VD(a, &(iso9660->primary)); iso9660->primary.vdd_type = VDD_PRIMARY; if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } if (iso9660->opt.joliet) { ret = write_VD(a, &(iso9660->joliet)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } /* Write Volume Descriptor Set Terminator */ ret = write_VD_terminator(a); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Non-ISO File System Information */ ret = write_information_block(a); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Type L Path Table */ ret = write_path_table(a, 0, &(iso9660->primary)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Type M Path Table */ ret = write_path_table(a, 1, &(iso9660->primary)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); if (iso9660->opt.joliet) { /* Write Type L Path Table */ ret = write_path_table(a, 0, &(iso9660->joliet)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Type M Path Table */ ret = write_path_table(a, 1, &(iso9660->joliet)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } /* Write Directory Descriptors */ ret = write_directory_descriptors(a, &(iso9660->primary)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); if (iso9660->opt.joliet) { ret = write_directory_descriptors(a, &(iso9660->joliet)); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } if (iso9660->opt.rr) { /* Write Rockridge ER(Extensions Reference) */ ret = write_rr_ER(a); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } /* Write File Descriptors */ ret = write_file_descriptors(a); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Write Padding */ if (iso9660->opt.pad) { ret = write_null(a, PADDING_BLOCK * LOGICAL_BLOCK_SIZE); if (ret != ARCHIVE_OK) return (ARCHIVE_FATAL); } if (iso9660->directories_too_deep != NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "%s: Directories too deep.", archive_entry_pathname( iso9660->directories_too_deep->file->entry)); return (ARCHIVE_WARN); } /* Write remaining data out. */ ret = wb_write_out(a); return (ret); } static int iso9660_free(struct archive_write *a) { struct iso9660 *iso9660; int i, ret; iso9660 = a->format_data; /* Close the temporary file. */ if (iso9660->temp_fd >= 0) close(iso9660->temp_fd); /* Free some stuff for zisofs operations. */ ret = zisofs_free(a); /* Remove directory entries in tree which includes file entries. */ isoent_free_all(iso9660->primary.rootent); for (i = 0; i < iso9660->primary.max_depth; i++) free(iso9660->primary.pathtbl[i].sorted); free(iso9660->primary.pathtbl); if (iso9660->opt.joliet) { isoent_free_all(iso9660->joliet.rootent); for (i = 0; i < iso9660->joliet.max_depth; i++) free(iso9660->joliet.pathtbl[i].sorted); free(iso9660->joliet.pathtbl); } /* Remove isofile entries. */ isofile_free_all_entries(iso9660); isofile_free_hardlinks(iso9660); archive_string_free(&(iso9660->cur_dirstr)); archive_string_free(&(iso9660->volume_identifier)); archive_string_free(&(iso9660->publisher_identifier)); archive_string_free(&(iso9660->data_preparer_identifier)); archive_string_free(&(iso9660->application_identifier)); archive_string_free(&(iso9660->copyright_file_identifier)); archive_string_free(&(iso9660->abstract_file_identifier)); archive_string_free(&(iso9660->bibliographic_file_identifier)); archive_string_free(&(iso9660->el_torito.catalog_filename)); archive_string_free(&(iso9660->el_torito.boot_filename)); archive_string_free(&(iso9660->el_torito.id)); archive_string_free(&(iso9660->utf16be)); archive_string_free(&(iso9660->mbs)); free(iso9660); a->format_data = NULL; return (ret); } /* * Get the System Identifier */ static void get_system_identitier(char *system_id, size_t size) { #if defined(HAVE_SYS_UTSNAME_H) struct utsname u; uname(&u); strncpy(system_id, u.sysname, size-1); system_id[size-1] = '\0'; #elif defined(_WIN32) && !defined(__CYGWIN__) strncpy(system_id, "Windows", size-1); system_id[size-1] = '\0'; #else #error no way to get the system identifier on your platform. #endif } static void set_str(unsigned char *p, const char *s, size_t l, char f, const char *map) { unsigned char c; if (s == NULL) s = ""; while ((c = *s++) != 0 && l > 0) { if (c >= 0x80 || map[c] == 0) { /* illegal character */ if (c >= 'a' && c <= 'z') { /* convert c from a-z to A-Z */ c -= 0x20; } else c = 0x5f; } *p++ = c; l--; } /* If l isn't zero, fill p buffer by the character * which indicated by f. */ if (l > 0) memset(p , f, l); } static inline int joliet_allowed_char(unsigned char high, unsigned char low) { int utf16 = (high << 8) | low; if (utf16 <= 0x001F) return (0); switch (utf16) { case 0x002A: /* '*' */ case 0x002F: /* '/' */ case 0x003A: /* ':' */ case 0x003B: /* ';' */ case 0x003F: /* '?' */ case 0x005C: /* '\' */ return (0);/* Not allowed. */ } return (1); } static int set_str_utf16be(struct archive_write *a, unsigned char *p, const char *s, size_t l, uint16_t uf, enum vdc vdc) { size_t size, i; int onepad; if (s == NULL) s = ""; if (l & 0x01) { onepad = 1; l &= ~1; } else onepad = 0; if (vdc == VDC_UCS2) { struct iso9660 *iso9660 = a->format_data; if (archive_strncpy_in_locale(&iso9660->utf16be, s, strlen(s), iso9660->sconv_to_utf16be) != 0 && errno == ENOMEM) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for UTF-16BE"); return (ARCHIVE_FATAL); } size = iso9660->utf16be.length; if (size > l) size = l; memcpy(p, iso9660->utf16be.s, size); } else { const uint16_t *u16 = (const uint16_t *)s; size = 0; while (*u16++) size += 2; if (size > l) size = l; memcpy(p, s, size); } for (i = 0; i < size; i += 2, p += 2) { if (!joliet_allowed_char(p[0], p[1])) archive_be16enc(p, 0x005F);/* '_' */ } l -= size; while (l > 0) { archive_be16enc(p, uf); p += 2; l -= 2; } if (onepad) *p = 0; return (ARCHIVE_OK); } static const char a_characters_map[0x80] = { /* 0 1 2 3 4 5 6 7 8 9 A B C D E F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 00-0F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 10-1F */ 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20-2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30-3F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40-4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,/* 50-5F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 60-6F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 70-7F */ }; static const char a1_characters_map[0x80] = { /* 0 1 2 3 4 5 6 7 8 9 A B C D E F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 00-0F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 10-1F */ 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20-2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30-3F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40-4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,/* 50-5F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60-6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,/* 70-7F */ }; static const char d_characters_map[0x80] = { /* 0 1 2 3 4 5 6 7 8 9 A B C D E F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 00-0F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 10-1F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 20-2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,/* 30-3F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40-4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,/* 50-5F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 60-6F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 70-7F */ }; static const char d1_characters_map[0x80] = { /* 0 1 2 3 4 5 6 7 8 9 A B C D E F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 00-0F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 10-1F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 20-2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,/* 30-3F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40-4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,/* 50-5F */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60-6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,/* 70-7F */ }; static int set_str_a_characters_bp(struct archive_write *a, unsigned char *bp, int from, int to, const char *s, enum vdc vdc) { int r; switch (vdc) { case VDC_STD: set_str(bp+from, s, to - from + 1, 0x20, a_characters_map); r = ARCHIVE_OK; break; case VDC_LOWERCASE: set_str(bp+from, s, to - from + 1, 0x20, a1_characters_map); r = ARCHIVE_OK; break; case VDC_UCS2: case VDC_UCS2_DIRECT: r = set_str_utf16be(a, bp+from, s, to - from + 1, 0x0020, vdc); break; default: r = ARCHIVE_FATAL; } return (r); } static int set_str_d_characters_bp(struct archive_write *a, unsigned char *bp, int from, int to, const char *s, enum vdc vdc) { int r; switch (vdc) { case VDC_STD: set_str(bp+from, s, to - from + 1, 0x20, d_characters_map); r = ARCHIVE_OK; break; case VDC_LOWERCASE: set_str(bp+from, s, to - from + 1, 0x20, d1_characters_map); r = ARCHIVE_OK; break; case VDC_UCS2: case VDC_UCS2_DIRECT: r = set_str_utf16be(a, bp+from, s, to - from + 1, 0x0020, vdc); break; default: r = ARCHIVE_FATAL; } return (r); } static void set_VD_bp(unsigned char *bp, enum VD_type type, unsigned char ver) { /* Volume Descriptor Type */ bp[1] = (unsigned char)type; /* Standard Identifier */ memcpy(bp + 2, "CD001", 5); /* Volume Descriptor Version */ bp[7] = ver; } static inline void set_unused_field_bp(unsigned char *bp, int from, int to) { memset(bp + from, 0, to - from + 1); } /* * 8-bit unsigned numerical values. * ISO9660 Standard 7.1.1 */ static inline void set_num_711(unsigned char *p, unsigned char value) { *p = value; } /* * 8-bit signed numerical values. * ISO9660 Standard 7.1.2 */ static inline void set_num_712(unsigned char *p, char value) { *((char *)p) = value; } /* * Least significant byte first. * ISO9660 Standard 7.2.1 */ static inline void set_num_721(unsigned char *p, uint16_t value) { archive_le16enc(p, value); } /* * Most significant byte first. * ISO9660 Standard 7.2.2 */ static inline void set_num_722(unsigned char *p, uint16_t value) { archive_be16enc(p, value); } /* * Both-byte orders. * ISO9660 Standard 7.2.3 */ static void set_num_723(unsigned char *p, uint16_t value) { archive_le16enc(p, value); archive_be16enc(p+2, value); } /* * Least significant byte first. * ISO9660 Standard 7.3.1 */ static inline void set_num_731(unsigned char *p, uint32_t value) { archive_le32enc(p, value); } /* * Most significant byte first. * ISO9660 Standard 7.3.2 */ static inline void set_num_732(unsigned char *p, uint32_t value) { archive_be32enc(p, value); } /* * Both-byte orders. * ISO9660 Standard 7.3.3 */ static inline void set_num_733(unsigned char *p, uint32_t value) { archive_le32enc(p, value); archive_be32enc(p+4, value); } static void set_digit(unsigned char *p, size_t s, int value) { while (s--) { p[s] = '0' + (value % 10); value /= 10; } } #if defined(HAVE_STRUCT_TM_TM_GMTOFF) #define get_gmoffset(tm) ((tm)->tm_gmtoff) #elif defined(HAVE_STRUCT_TM___TM_GMTOFF) #define get_gmoffset(tm) ((tm)->__tm_gmtoff) #else static long get_gmoffset(struct tm *tm) { long offset; #if defined(HAVE__GET_TIMEZONE) _get_timezone(&offset); #elif defined(__CYGWIN__) || defined(__MINGW32__) || defined(__BORLANDC__) offset = _timezone; #else offset = timezone; #endif offset *= -1; if (tm->tm_isdst) offset += 3600; return (offset); } #endif static void get_tmfromtime(struct tm *tm, time_t *t) { #if HAVE_LOCALTIME_R tzset(); localtime_r(t, tm); #elif HAVE__LOCALTIME64_S _localtime64_s(tm, t); #else memcpy(tm, localtime(t), sizeof(*tm)); #endif } /* * Date and Time Format. * ISO9660 Standard 8.4.26.1 */ static void set_date_time(unsigned char *p, time_t t) { struct tm tm; get_tmfromtime(&tm, &t); set_digit(p, 4, tm.tm_year + 1900); set_digit(p+4, 2, tm.tm_mon + 1); set_digit(p+6, 2, tm.tm_mday); set_digit(p+8, 2, tm.tm_hour); set_digit(p+10, 2, tm.tm_min); set_digit(p+12, 2, tm.tm_sec); set_digit(p+14, 2, 0); set_num_712(p+16, get_gmoffset(&tm)/(60*15)); } static void set_date_time_null(unsigned char *p) { memset(p, '0', 16); p[16] = 0; } static void set_time_915(unsigned char *p, time_t t) { struct tm tm; get_tmfromtime(&tm, &t); set_num_711(p+0, tm.tm_year); set_num_711(p+1, tm.tm_mon+1); set_num_711(p+2, tm.tm_mday); set_num_711(p+3, tm.tm_hour); set_num_711(p+4, tm.tm_min); set_num_711(p+5, tm.tm_sec); set_num_712(p+6, get_gmoffset(&tm)/(60*15)); } /* * Write SUSP "CE" System Use Entry. */ static int set_SUSP_CE(unsigned char *p, int location, int offset, int size) { unsigned char *bp = p -1; /* Extend the System Use Area * "CE" Format: * len ver * +----+----+----+----+-----------+-----------+ * | 'C'| 'E'| 1C | 01 | LOCATION1 | LOCATION2 | * +----+----+----+----+-----------+-----------+ * 0 1 2 3 4 12 20 * +-----------+ * | LOCATION3 | * +-----------+ * 20 28 * LOCATION1 : Location of Continuation of System Use Area. * LOCATION2 : Offset to Start of Continuation. * LOCATION3 : Length of the Continuation. */ bp[1] = 'C'; bp[2] = 'E'; bp[3] = RR_CE_SIZE; /* length */ bp[4] = 1; /* version */ set_num_733(bp+5, location); set_num_733(bp+13, offset); set_num_733(bp+21, size); return (RR_CE_SIZE); } /* * The functions, which names are beginning with extra_, are used to * control extra records. * The maximum size of a Directory Record is 254. When a filename is * very long, all of RRIP data of a file won't stored to the Directory * Record and so remaining RRIP data store to an extra record instead. */ static unsigned char * extra_open_record(unsigned char *bp, int dr_len, struct isoent *isoent, struct ctl_extr_rec *ctl) { ctl->bp = bp; if (bp != NULL) bp += dr_len; ctl->use_extr = 0; ctl->isoent = isoent; ctl->ce_ptr = NULL; ctl->cur_len = ctl->dr_len = dr_len; ctl->limit = DR_LIMIT; return (bp); } static void extra_close_record(struct ctl_extr_rec *ctl, int ce_size) { int padding = 0; if (ce_size > 0) extra_tell_used_size(ctl, ce_size); /* Padding. */ if (ctl->cur_len & 0x01) { ctl->cur_len++; if (ctl->bp != NULL) ctl->bp[ctl->cur_len] = 0; padding = 1; } if (ctl->use_extr) { if (ctl->ce_ptr != NULL) set_SUSP_CE(ctl->ce_ptr, ctl->extr_loc, ctl->extr_off, ctl->cur_len - padding); } else ctl->dr_len = ctl->cur_len; } #define extra_space(ctl) ((ctl)->limit - (ctl)->cur_len) static unsigned char * extra_next_record(struct ctl_extr_rec *ctl, int length) { int cur_len = ctl->cur_len;/* save cur_len */ /* Close the current extra record or Directory Record. */ extra_close_record(ctl, RR_CE_SIZE); /* Get a next extra record. */ ctl->use_extr = 1; if (ctl->bp != NULL) { /* Storing data into an extra record. */ unsigned char *p; /* Save the pointer where a CE extension will be * stored to. */ ctl->ce_ptr = &ctl->bp[cur_len+1]; p = extra_get_record(ctl->isoent, &ctl->limit, &ctl->extr_off, &ctl->extr_loc); ctl->bp = p - 1;/* the base of bp offset is 1. */ } else /* Calculating the size of an extra record. */ (void)extra_get_record(ctl->isoent, &ctl->limit, NULL, NULL); ctl->cur_len = 0; /* Check if an extra record is almost full. * If so, get a next one. */ if (extra_space(ctl) < length) (void)extra_next_record(ctl, length); return (ctl->bp); } static inline struct extr_rec * extra_last_record(struct isoent *isoent) { if (isoent->extr_rec_list.first == NULL) return (NULL); return ((struct extr_rec *)(void *) ((char *)(isoent->extr_rec_list.last) - offsetof(struct extr_rec, next))); } static unsigned char * extra_get_record(struct isoent *isoent, int *space, int *off, int *loc) { struct extr_rec *rec; isoent = isoent->parent; if (off != NULL) { /* Storing data into an extra record. */ rec = isoent->extr_rec_list.current; if (DR_SAFETY > LOGICAL_BLOCK_SIZE - rec->offset) rec = rec->next; } else { /* Calculating the size of an extra record. */ rec = extra_last_record(isoent); if (rec == NULL || DR_SAFETY > LOGICAL_BLOCK_SIZE - rec->offset) { rec = malloc(sizeof(*rec)); if (rec == NULL) return (NULL); rec->location = 0; rec->offset = 0; /* Insert `rec` into the tail of isoent->extr_rec_list */ rec->next = NULL; *isoent->extr_rec_list.last = rec; isoent->extr_rec_list.last = &(rec->next); } } *space = LOGICAL_BLOCK_SIZE - rec->offset - DR_SAFETY; if (*space & 0x01) *space -= 1;/* Keep padding space. */ if (off != NULL) *off = rec->offset; if (loc != NULL) *loc = rec->location; isoent->extr_rec_list.current = rec; return (&rec->buf[rec->offset]); } static void extra_tell_used_size(struct ctl_extr_rec *ctl, int size) { struct isoent *isoent; struct extr_rec *rec; if (ctl->use_extr) { isoent = ctl->isoent->parent; rec = isoent->extr_rec_list.current; if (rec != NULL) rec->offset += size; } ctl->cur_len += size; } static int extra_setup_location(struct isoent *isoent, int location) { struct extr_rec *rec; int cnt; cnt = 0; rec = isoent->extr_rec_list.first; isoent->extr_rec_list.current = rec; while (rec) { cnt++; rec->location = location++; rec->offset = 0; rec = rec->next; } return (cnt); } /* * Create the RRIP entries. */ static int set_directory_record_rr(unsigned char *bp, int dr_len, struct isoent *isoent, struct iso9660 *iso9660, enum dir_rec_type t) { /* Flags(BP 5) of the Rockridge "RR" System Use Field */ unsigned char rr_flag; #define RR_USE_PX 0x01 #define RR_USE_PN 0x02 #define RR_USE_SL 0x04 #define RR_USE_NM 0x08 #define RR_USE_CL 0x10 #define RR_USE_PL 0x20 #define RR_USE_RE 0x40 #define RR_USE_TF 0x80 int length; struct ctl_extr_rec ctl; struct isoent *rr_parent, *pxent; struct isofile *file; bp = extra_open_record(bp, dr_len, isoent, &ctl); if (t == DIR_REC_PARENT) { rr_parent = isoent->rr_parent; pxent = isoent->parent; if (rr_parent != NULL) isoent = rr_parent; else isoent = isoent->parent; } else { rr_parent = NULL; pxent = isoent; } file = isoent->file; if (t != DIR_REC_NORMAL) { rr_flag = RR_USE_PX | RR_USE_TF; if (rr_parent != NULL) rr_flag |= RR_USE_PL; } else { rr_flag = RR_USE_PX | RR_USE_NM | RR_USE_TF; if (archive_entry_filetype(file->entry) == AE_IFLNK) rr_flag |= RR_USE_SL; if (isoent->rr_parent != NULL) rr_flag |= RR_USE_RE; if (isoent->rr_child != NULL) rr_flag |= RR_USE_CL; if (archive_entry_filetype(file->entry) == AE_IFCHR || archive_entry_filetype(file->entry) == AE_IFBLK) rr_flag |= RR_USE_PN; #ifdef COMPAT_MKISOFS /* * mkisofs 2.01.01a63 records "RE" extension to * the entry of "rr_moved" directory. * I don't understand this behavior. */ if (isoent->virtual && isoent->parent == iso9660->primary.rootent && strcmp(isoent->file->basename.s, "rr_moved") == 0) rr_flag |= RR_USE_RE; #endif } /* Write "SP" System Use Entry. */ if (t == DIR_REC_SELF && isoent == isoent->parent) { length = 7; if (bp != NULL) { bp[1] = 'S'; bp[2] = 'P'; bp[3] = length; bp[4] = 1; /* version */ bp[5] = 0xBE; /* Check Byte */ bp[6] = 0xEF; /* Check Byte */ bp[7] = 0; bp += length; } extra_tell_used_size(&ctl, length); } /* Write "RR" System Use Entry. */ length = 5; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'R'; bp[2] = 'R'; bp[3] = length; bp[4] = 1; /* version */ bp[5] = rr_flag; bp += length; } extra_tell_used_size(&ctl, length); /* Write "NM" System Use Entry. */ if (rr_flag & RR_USE_NM) { /* * "NM" Format: * e.g. a basename is 'foo' * len ver flg * +----+----+----+----+----+----+----+----+ * | 'N'| 'M'| 08 | 01 | 00 | 'f'| 'o'| 'o'| * +----+----+----+----+----+----+----+----+ * <----------------- len -----------------> */ size_t nmlen = file->basename.length; const char *nm = file->basename.s; size_t nmmax; if (extra_space(&ctl) < 6) bp = extra_next_record(&ctl, 6); if (bp != NULL) { bp[1] = 'N'; bp[2] = 'M'; bp[4] = 1; /* version */ } nmmax = extra_space(&ctl); if (nmmax > 0xff) nmmax = 0xff; while (nmlen + 5 > nmmax) { length = nmmax; if (bp != NULL) { bp[3] = length; bp[5] = 0x01;/* Alternate Name continues * in next "NM" field */ memcpy(bp+6, nm, length - 5); bp += length; } nmlen -= length - 5; nm += length - 5; extra_tell_used_size(&ctl, length); if (extra_space(&ctl) < 6) { bp = extra_next_record(&ctl, 6); nmmax = extra_space(&ctl); if (nmmax > 0xff) nmmax = 0xff; } if (bp != NULL) { bp[1] = 'N'; bp[2] = 'M'; bp[4] = 1; /* version */ } } length = 5 + nmlen; if (bp != NULL) { bp[3] = length; bp[5] = 0; memcpy(bp+6, nm, nmlen); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "PX" System Use Entry. */ if (rr_flag & RR_USE_PX) { /* * "PX" Format: * len ver * +----+----+----+----+-----------+-----------+ * | 'P'| 'X'| 2C | 01 | FILE MODE | LINKS | * +----+----+----+----+-----------+-----------+ * 0 1 2 3 4 12 20 * +-----------+-----------+------------------+ * | USER ID | GROUP ID |FILE SERIAL NUMBER| * +-----------+-----------+------------------+ * 20 28 36 44 */ length = 44; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { mode_t mode; uid_t uid; gid_t gid; mode = archive_entry_mode(file->entry); uid = archive_entry_uid(file->entry); gid = archive_entry_gid(file->entry); if (iso9660->opt.rr == OPT_RR_USEFUL) { /* * This action is simular mkisofs -r option * but our rockridge=useful option does not * set a zero to uid and gid. */ /* set all read bit ON */ mode |= 0444; #if !defined(_WIN32) && !defined(__CYGWIN__) if (mode & 0111) #endif /* set all exec bit ON */ mode |= 0111; /* clear all write bits. */ mode &= ~0222; /* clear setuid,setgid,sticky bits. */ mode &= ~07000; } bp[1] = 'P'; bp[2] = 'X'; bp[3] = length; bp[4] = 1; /* version */ /* file mode */ set_num_733(bp+5, mode); /* file links (stat.st_nlink) */ set_num_733(bp+13, archive_entry_nlink(file->entry)); set_num_733(bp+21, uid); set_num_733(bp+29, gid); /* File Serial Number */ if (pxent->dir) set_num_733(bp+37, pxent->dir_location); else if (file->hardlink_target != NULL) set_num_733(bp+37, file->hardlink_target->cur_content->location); else set_num_733(bp+37, file->cur_content->location); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "SL" System Use Entry. */ if (rr_flag & RR_USE_SL) { /* * "SL" Format: * e.g. a symbolic name is 'foo/bar' * len ver flg * +----+----+----+----+----+------------+ * | 'S'| 'L'| 0F | 01 | 00 | components | * +----+----+----+----+----+-----+------+ * 0 1 2 3 4 5 ...|... 15 * <----------------- len --------+------> * components : | * cflg clen | * +----+----+----+----+----+ | * | 00 | 03 | 'f'| 'o'| 'o'| <---+ * +----+----+----+----+----+ | * 5 6 7 8 9 10 | * cflg clen | * +----+----+----+----+----+ | * | 00 | 03 | 'b'| 'a'| 'r'| <---+ * +----+----+----+----+----+ * 10 11 12 13 14 15 * * - cflg : flag of componet * - clen : length of componet */ const char *sl; char sl_last; if (extra_space(&ctl) < 7) bp = extra_next_record(&ctl, 7); sl = file->symlink.s; sl_last = '\0'; if (bp != NULL) { bp[1] = 'S'; bp[2] = 'L'; bp[4] = 1; /* version */ } for (;;) { unsigned char *nc, *cf, *cl, cldmy = 0; int sllen, slmax; slmax = extra_space(&ctl); if (slmax > 0xff) slmax = 0xff; if (bp != NULL) nc = &bp[6]; else nc = NULL; cf = cl = NULL; sllen = 0; while (*sl && sllen + 11 < slmax) { if (sl_last == '\0' && sl[0] == '/') { /* * flg len * +----+----+ * | 08 | 00 | ROOT component. * +----+----+ ("/") * * Root component has to appear * at the first component only. */ if (nc != NULL) { cf = nc++; *cf = 0x08; /* ROOT */ *nc++ = 0; } sllen += 2; sl++; sl_last = '/'; cl = NULL; continue; } if (((sl_last == '\0' || sl_last == '/') && sl[0] == '.' && sl[1] == '.' && (sl[2] == '/' || sl[2] == '\0')) || (sl[0] == '/' && sl[1] == '.' && sl[2] == '.' && (sl[3] == '/' || sl[3] == '\0'))) { /* * flg len * +----+----+ * | 04 | 00 | PARENT component. * +----+----+ ("..") */ if (nc != NULL) { cf = nc++; *cf = 0x04; /* PARENT */ *nc++ = 0; } sllen += 2; if (sl[0] == '/') sl += 3;/* skip "/.." */ else sl += 2;/* skip ".." */ sl_last = '.'; cl = NULL; continue; } if (((sl_last == '\0' || sl_last == '/') && sl[0] == '.' && (sl[1] == '/' || sl[1] == '\0')) || (sl[0] == '/' && sl[1] == '.' && (sl[2] == '/' || sl[2] == '\0'))) { /* * flg len * +----+----+ * | 02 | 00 | CURREENT component. * +----+----+ (".") */ if (nc != NULL) { cf = nc++; *cf = 0x02; /* CURRENT */ *nc++ = 0; } sllen += 2; if (sl[0] == '/') sl += 2;/* skip "/." */ else sl ++; /* skip "." */ sl_last = '.'; cl = NULL; continue; } if (sl[0] == '/' || cl == NULL) { if (nc != NULL) { cf = nc++; *cf = 0; cl = nc++; *cl = 0; } else cl = &cldmy; sllen += 2; if (sl[0] == '/') { sl_last = *sl++; continue; } } sl_last = *sl++; if (nc != NULL) { *nc++ = sl_last; (*cl) ++; } sllen++; } if (*sl) { length = 5 + sllen; if (bp != NULL) { /* * Mark flg as CONTINUE component. */ *cf |= 0x01; /* * len ver flg * +----+----+----+----+----+- * | 'S'| 'L'| XX | 01 | 01 | * +----+----+----+----+----+- * ^ * continues in next "SL" */ bp[3] = length; bp[5] = 0x01;/* This Symbolic Link * continues in next * "SL" field */ bp += length; } extra_tell_used_size(&ctl, length); if (extra_space(&ctl) < 11) bp = extra_next_record(&ctl, 11); if (bp != NULL) { /* Next 'SL' */ bp[1] = 'S'; bp[2] = 'L'; bp[4] = 1; /* version */ } } else { length = 5 + sllen; if (bp != NULL) { bp[3] = length; bp[5] = 0; bp += length; } extra_tell_used_size(&ctl, length); break; } } } /* Write "TF" System Use Entry. */ if (rr_flag & RR_USE_TF) { /* * "TF" Format: * len ver * +----+----+----+----+-----+-------------+ * | 'T'| 'F'| XX | 01 |FLAGS| TIME STAMPS | * +----+----+----+----+-----+-------------+ * 0 1 2 3 4 5 XX * TIME STAMPS : ISO 9660 Standard 9.1.5. * If TF_LONG_FORM FLAGS is set, * use ISO9660 Standard 8.4.26.1. */ #define TF_CREATION 0x01 /* Creation time recorded */ #define TF_MODIFY 0x02 /* Modification time recorded */ #define TF_ACCESS 0x04 /* Last Access time recorded */ #define TF_ATTRIBUTES 0x08 /* Last Attribute Change time recorded */ #define TF_BACKUP 0x10 /* Last Backup time recorded */ #define TF_EXPIRATION 0x20 /* Expiration time recorded */ #define TF_EFFECTIVE 0x40 /* Effective time recorded */ #define TF_LONG_FORM 0x80 /* ISO 9660 17-byte time format used */ unsigned char tf_flags; length = 5; tf_flags = 0; #ifndef COMPAT_MKISOFS if (archive_entry_birthtime_is_set(file->entry) && archive_entry_birthtime(file->entry) <= archive_entry_mtime(file->entry)) { length += 7; tf_flags |= TF_CREATION; } #endif if (archive_entry_mtime_is_set(file->entry)) { length += 7; tf_flags |= TF_MODIFY; } if (archive_entry_atime_is_set(file->entry)) { length += 7; tf_flags |= TF_ACCESS; } if (archive_entry_ctime_is_set(file->entry)) { length += 7; tf_flags |= TF_ATTRIBUTES; } if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'T'; bp[2] = 'F'; bp[3] = length; bp[4] = 1; /* version */ bp[5] = tf_flags; bp += 5; /* Creation time */ if (tf_flags & TF_CREATION) { set_time_915(bp+1, archive_entry_birthtime(file->entry)); bp += 7; } /* Modification time */ if (tf_flags & TF_MODIFY) { set_time_915(bp+1, archive_entry_mtime(file->entry)); bp += 7; } /* Last Access time */ if (tf_flags & TF_ACCESS) { set_time_915(bp+1, archive_entry_atime(file->entry)); bp += 7; } /* Last Attribute Change time */ if (tf_flags & TF_ATTRIBUTES) { set_time_915(bp+1, archive_entry_ctime(file->entry)); bp += 7; } } extra_tell_used_size(&ctl, length); } /* Write "RE" System Use Entry. */ if (rr_flag & RR_USE_RE) { /* * "RE" Format: * len ver * +----+----+----+----+ * | 'R'| 'E'| 04 | 01 | * +----+----+----+----+ * 0 1 2 3 4 */ length = 4; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'R'; bp[2] = 'E'; bp[3] = length; bp[4] = 1; /* version */ bp += length; } extra_tell_used_size(&ctl, length); } /* Write "PL" System Use Entry. */ if (rr_flag & RR_USE_PL) { /* * "PL" Format: * len ver * +----+----+----+----+------------+ * | 'P'| 'L'| 0C | 01 | *LOCATION | * +----+----+----+----+------------+ * 0 1 2 3 4 12 * *LOCATION: location of parent directory */ length = 12; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'P'; bp[2] = 'L'; bp[3] = length; bp[4] = 1; /* version */ set_num_733(bp + 5, rr_parent->dir_location); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "CL" System Use Entry. */ if (rr_flag & RR_USE_CL) { /* * "CL" Format: * len ver * +----+----+----+----+------------+ * | 'C'| 'L'| 0C | 01 | *LOCATION | * +----+----+----+----+------------+ * 0 1 2 3 4 12 * *LOCATION: location of child directory */ length = 12; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'C'; bp[2] = 'L'; bp[3] = length; bp[4] = 1; /* version */ set_num_733(bp + 5, isoent->rr_child->dir_location); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "PN" System Use Entry. */ if (rr_flag & RR_USE_PN) { /* * "PN" Format: * len ver * +----+----+----+----+------------+------------+ * | 'P'| 'N'| 14 | 01 | dev_t high | dev_t low | * +----+----+----+----+------------+------------+ * 0 1 2 3 4 12 20 */ length = 20; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { uint64_t dev; bp[1] = 'P'; bp[2] = 'N'; bp[3] = length; bp[4] = 1; /* version */ dev = (uint64_t)archive_entry_rdev(file->entry); set_num_733(bp + 5, dev >> 32); set_num_733(bp + 13, dev & 0xFFFFFFFF); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "ZF" System Use Entry. */ if (file->zisofs.header_size) { /* * "ZF" Format: * len ver * +----+----+----+----+----+----+-------------+ * | 'Z'| 'F'| 10 | 01 | 'p'| 'z'| Header Size | * +----+----+----+----+----+----+-------------+ * 0 1 2 3 4 5 6 7 * +--------------------+-------------------+ * | Log2 of block Size | Uncompressed Size | * +--------------------+-------------------+ * 7 8 16 */ length = 16; if (extra_space(&ctl) < length) bp = extra_next_record(&ctl, length); if (bp != NULL) { bp[1] = 'Z'; bp[2] = 'F'; bp[3] = length; bp[4] = 1; /* version */ bp[5] = 'p'; bp[6] = 'z'; bp[7] = file->zisofs.header_size; bp[8] = file->zisofs.log2_bs; set_num_733(bp + 9, file->zisofs.uncompressed_size); bp += length; } extra_tell_used_size(&ctl, length); } /* Write "CE" System Use Entry. */ if (t == DIR_REC_SELF && isoent == isoent->parent) { length = RR_CE_SIZE; if (bp != NULL) set_SUSP_CE(bp+1, iso9660->location_rrip_er, 0, RRIP_ER_SIZE); extra_tell_used_size(&ctl, length); } extra_close_record(&ctl, 0); return (ctl.dr_len); } /* * Write data of a Directory Record or calculate writing bytes itself. * If parameter `p' is NULL, calculates the size of writing data, which * a Directory Record needs to write, then it saved and return * the calculated size. * Parameter `n' is a remaining size of buffer. when parameter `p' is * not NULL, check whether that `n' is not less than the saved size. * if that `n' is small, return zero. * * This format of the Directory Record is according to * ISO9660 Standard 9.1 */ static int set_directory_record(unsigned char *p, size_t n, struct isoent *isoent, struct iso9660 *iso9660, enum dir_rec_type t, enum vdd_type vdd_type) { unsigned char *bp; size_t dr_len; size_t fi_len; if (p != NULL) { /* * Check whether a write buffer size is less than the * saved size which is needed to write this Directory * Record. */ switch (t) { case DIR_REC_VD: dr_len = isoent->dr_len.vd; break; case DIR_REC_SELF: dr_len = isoent->dr_len.self; break; case DIR_REC_PARENT: dr_len = isoent->dr_len.parent; break; case DIR_REC_NORMAL: default: dr_len = isoent->dr_len.normal; break; } if (dr_len > n) return (0);/* Needs more buffer size. */ } if (t == DIR_REC_NORMAL && isoent->identifier != NULL) fi_len = isoent->id_len; else fi_len = 1; if (p != NULL) { struct isoent *xisoent; struct isofile *file; unsigned char flag; if (t == DIR_REC_PARENT) xisoent = isoent->parent; else xisoent = isoent; file = isoent->file; if (file->hardlink_target != NULL) file = file->hardlink_target; /* Make a file flag. */ if (xisoent->dir) flag = FILE_FLAG_DIRECTORY; else { if (file->cur_content->next != NULL) flag = FILE_FLAG_MULTI_EXTENT; else flag = 0; } bp = p -1; /* Extended Attribute Record Length */ set_num_711(bp+2, 0); /* Location of Extent */ if (xisoent->dir) set_num_733(bp+3, xisoent->dir_location); else set_num_733(bp+3, file->cur_content->location); /* Data Length */ if (xisoent->dir) set_num_733(bp+11, xisoent->dir_block * LOGICAL_BLOCK_SIZE); else set_num_733(bp+11, file->cur_content->size); /* Recording Date and Time */ /* NOTE: * If a file type is symbolic link, you are seeing this * field value is different from a value mkisofs makes. * libarchive uses lstat to get this one, but it * seems mkisofs uses stat to get. */ set_time_915(bp+19, archive_entry_mtime(xisoent->file->entry)); /* File Flags */ bp[26] = flag; /* File Unit Size */ set_num_711(bp+27, 0); /* Interleave Gap Size */ set_num_711(bp+28, 0); /* Volume Sequence Number */ set_num_723(bp+29, iso9660->volume_sequence_number); /* Length of File Identifier */ set_num_711(bp+33, fi_len); /* File Identifier */ switch (t) { case DIR_REC_VD: case DIR_REC_SELF: set_num_711(bp+34, 0); break; case DIR_REC_PARENT: set_num_711(bp+34, 1); break; case DIR_REC_NORMAL: if (isoent->identifier != NULL) memcpy(bp+34, isoent->identifier, fi_len); else set_num_711(bp+34, 0); break; } } else bp = NULL; dr_len = 33 + fi_len; /* Padding Field */ if (dr_len & 0x01) { dr_len ++; if (p != NULL) bp[dr_len] = 0; } /* Volume Descriptor does not record extension. */ if (t == DIR_REC_VD) { if (p != NULL) /* Length of Directory Record */ set_num_711(p, dr_len); else isoent->dr_len.vd = dr_len; return (dr_len); } /* Rockridge */ if (iso9660->opt.rr && vdd_type != VDD_JOLIET) dr_len = set_directory_record_rr(bp, dr_len, isoent, iso9660, t); if (p != NULL) /* Length of Directory Record */ set_num_711(p, dr_len); else { /* * Save the size which is needed to write this * Directory Record. */ switch (t) { case DIR_REC_VD: /* This case does not come, but compiler * complains that DIR_REC_VD not handled * in switch .... */ break; case DIR_REC_SELF: isoent->dr_len.self = dr_len; break; case DIR_REC_PARENT: isoent->dr_len.parent = dr_len; break; case DIR_REC_NORMAL: isoent->dr_len.normal = dr_len; break; } } return (dr_len); } /* * Calculate the size of a directory record. */ static inline int get_dir_rec_size(struct iso9660 *iso9660, struct isoent *isoent, enum dir_rec_type t, enum vdd_type vdd_type) { return (set_directory_record(NULL, SIZE_MAX, isoent, iso9660, t, vdd_type)); } /* * Manage to write ISO-image data with wbuff to reduce calling * __archive_write_output() for performance. */ static inline unsigned char * wb_buffptr(struct archive_write *a) { struct iso9660 *iso9660 = (struct iso9660 *)a->format_data; return (&(iso9660->wbuff[sizeof(iso9660->wbuff) - iso9660->wbuff_remaining])); } static int wb_write_out(struct archive_write *a) { struct iso9660 *iso9660 = (struct iso9660 *)a->format_data; size_t wsize, nw; int r; wsize = sizeof(iso9660->wbuff) - iso9660->wbuff_remaining; nw = wsize % LOGICAL_BLOCK_SIZE; if (iso9660->wbuff_type == WB_TO_STREAM) r = __archive_write_output(a, iso9660->wbuff, wsize - nw); else r = write_to_temp(a, iso9660->wbuff, wsize - nw); /* Increase the offset. */ iso9660->wbuff_offset += wsize - nw; if (iso9660->wbuff_offset > iso9660->wbuff_written) iso9660->wbuff_written = iso9660->wbuff_offset; iso9660->wbuff_remaining = sizeof(iso9660->wbuff); if (nw) { iso9660->wbuff_remaining -= nw; memmove(iso9660->wbuff, iso9660->wbuff + wsize - nw, nw); } return (r); } static int wb_consume(struct archive_write *a, size_t size) { struct iso9660 *iso9660 = (struct iso9660 *)a->format_data; if (size > iso9660->wbuff_remaining || iso9660->wbuff_remaining == 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Internal Programing error: iso9660:wb_consume()" " size=%jd, wbuff_remaining=%jd", (intmax_t)size, (intmax_t)iso9660->wbuff_remaining); return (ARCHIVE_FATAL); } iso9660->wbuff_remaining -= size; if (iso9660->wbuff_remaining < LOGICAL_BLOCK_SIZE) return (wb_write_out(a)); return (ARCHIVE_OK); } #ifdef HAVE_ZLIB_H static int wb_set_offset(struct archive_write *a, int64_t off) { struct iso9660 *iso9660 = (struct iso9660 *)a->format_data; int64_t used, ext_bytes; if (iso9660->wbuff_type != WB_TO_TEMP) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Internal Programing error: iso9660:wb_set_offset()"); return (ARCHIVE_FATAL); } used = sizeof(iso9660->wbuff) - iso9660->wbuff_remaining; if (iso9660->wbuff_offset + used > iso9660->wbuff_tail) iso9660->wbuff_tail = iso9660->wbuff_offset + used; if (iso9660->wbuff_offset < iso9660->wbuff_written) { if (used > 0 && write_to_temp(a, iso9660->wbuff, used) != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->wbuff_offset = iso9660->wbuff_written; lseek(iso9660->temp_fd, iso9660->wbuff_offset, SEEK_SET); iso9660->wbuff_remaining = sizeof(iso9660->wbuff); used = 0; } if (off < iso9660->wbuff_offset) { /* * Write out waiting data. */ if (used > 0) { if (wb_write_out(a) != ARCHIVE_OK) return (ARCHIVE_FATAL); } lseek(iso9660->temp_fd, off, SEEK_SET); iso9660->wbuff_offset = off; iso9660->wbuff_remaining = sizeof(iso9660->wbuff); } else if (off <= iso9660->wbuff_tail) { iso9660->wbuff_remaining = sizeof(iso9660->wbuff) - (off - iso9660->wbuff_offset); } else { ext_bytes = off - iso9660->wbuff_tail; iso9660->wbuff_remaining = sizeof(iso9660->wbuff) - (iso9660->wbuff_tail - iso9660->wbuff_offset); - while (ext_bytes >= iso9660->wbuff_remaining) { + while (ext_bytes >= (int64_t)iso9660->wbuff_remaining) { if (write_null(a, (size_t)iso9660->wbuff_remaining) != ARCHIVE_OK) return (ARCHIVE_FATAL); ext_bytes -= iso9660->wbuff_remaining; } if (ext_bytes > 0) { if (write_null(a, (size_t)ext_bytes) != ARCHIVE_OK) return (ARCHIVE_FATAL); } } return (ARCHIVE_OK); } #endif /* HAVE_ZLIB_H */ static int write_null(struct archive_write *a, size_t size) { size_t remaining; unsigned char *p, *old; int r; remaining = wb_remaining(a); p = wb_buffptr(a); if (size <= remaining) { memset(p, 0, size); return (wb_consume(a, size)); } memset(p, 0, remaining); r = wb_consume(a, remaining); if (r != ARCHIVE_OK) return (r); size -= remaining; old = p; p = wb_buffptr(a); memset(p, 0, old - p); remaining = wb_remaining(a); while (size) { size_t wsize = size; if (wsize > remaining) wsize = remaining; r = wb_consume(a, wsize); if (r != ARCHIVE_OK) return (r); size -= wsize; } return (ARCHIVE_OK); } /* * Write Volume Descriptor Set Terminator */ static int write_VD_terminator(struct archive_write *a) { unsigned char *bp; bp = wb_buffptr(a) -1; set_VD_bp(bp, VDT_TERMINATOR, 1); set_unused_field_bp(bp, 8, LOGICAL_BLOCK_SIZE); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } static int set_file_identifier(unsigned char *bp, int from, int to, enum vdc vdc, struct archive_write *a, struct vdd *vdd, struct archive_string *id, const char *label, int leading_under, enum char_type char_type) { char identifier[256]; struct isoent *isoent; const char *ids; size_t len; int r; if (id->length > 0 && leading_under && id->s[0] != '_') { if (char_type == A_CHAR) r = set_str_a_characters_bp(a, bp, from, to, id->s, vdc); else r = set_str_d_characters_bp(a, bp, from, to, id->s, vdc); } else if (id->length > 0) { ids = id->s; if (leading_under) ids++; isoent = isoent_find_entry(vdd->rootent, ids); if (isoent == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Not Found %s `%s'.", label, ids); return (ARCHIVE_FATAL); } len = isoent->ext_off + isoent->ext_len; if (vdd->vdd_type == VDD_JOLIET) { if (len > sizeof(identifier)-2) len = sizeof(identifier)-2; } else { if (len > sizeof(identifier)-1) len = sizeof(identifier)-1; } memcpy(identifier, isoent->identifier, len); identifier[len] = '\0'; if (vdd->vdd_type == VDD_JOLIET) { identifier[len+1] = 0; vdc = VDC_UCS2_DIRECT; } if (char_type == A_CHAR) r = set_str_a_characters_bp(a, bp, from, to, identifier, vdc); else r = set_str_d_characters_bp(a, bp, from, to, identifier, vdc); } else { if (char_type == A_CHAR) r = set_str_a_characters_bp(a, bp, from, to, NULL, vdc); else r = set_str_d_characters_bp(a, bp, from, to, NULL, vdc); } return (r); } /* * Write Primary/Supplementary Volume Descriptor */ static int write_VD(struct archive_write *a, struct vdd *vdd) { struct iso9660 *iso9660; unsigned char *bp; uint16_t volume_set_size = 1; char identifier[256]; enum VD_type vdt; enum vdc vdc; unsigned char vd_ver, fst_ver; int r; iso9660 = a->format_data; switch (vdd->vdd_type) { case VDD_JOLIET: vdt = VDT_SUPPLEMENTARY; vd_ver = fst_ver = 1; vdc = VDC_UCS2; break; case VDD_ENHANCED: vdt = VDT_SUPPLEMENTARY; vd_ver = fst_ver = 2; vdc = VDC_LOWERCASE; break; case VDD_PRIMARY: default: vdt = VDT_PRIMARY; vd_ver = fst_ver = 1; #ifdef COMPAT_MKISOFS vdc = VDC_LOWERCASE; #else vdc = VDC_STD; #endif break; } bp = wb_buffptr(a) -1; /* Volume Descriptor Type */ set_VD_bp(bp, vdt, vd_ver); /* Unused Field */ set_unused_field_bp(bp, 8, 8); /* System Identifier */ get_system_identitier(identifier, sizeof(identifier)); r = set_str_a_characters_bp(a, bp, 9, 40, identifier, vdc); if (r != ARCHIVE_OK) return (r); /* Volume Identifier */ r = set_str_d_characters_bp(a, bp, 41, 72, iso9660->volume_identifier.s, vdc); if (r != ARCHIVE_OK) return (r); /* Unused Field */ set_unused_field_bp(bp, 73, 80); /* Volume Space Size */ set_num_733(bp+81, iso9660->volume_space_size); if (vdd->vdd_type == VDD_JOLIET) { /* Escape Sequences */ bp[89] = 0x25;/* UCS-2 Level 3 */ bp[90] = 0x2F; bp[91] = 0x45; memset(bp + 92, 0, 120 - 92 + 1); } else { /* Unused Field */ set_unused_field_bp(bp, 89, 120); } /* Volume Set Size */ set_num_723(bp+121, volume_set_size); /* Volume Sequence Number */ set_num_723(bp+125, iso9660->volume_sequence_number); /* Logical Block Size */ set_num_723(bp+129, LOGICAL_BLOCK_SIZE); /* Path Table Size */ set_num_733(bp+133, vdd->path_table_size); /* Location of Occurrence of Type L Path Table */ set_num_731(bp+141, vdd->location_type_L_path_table); /* Location of Optional Occurrence of Type L Path Table */ set_num_731(bp+145, 0); /* Location of Occurrence of Type M Path Table */ set_num_732(bp+149, vdd->location_type_M_path_table); /* Location of Optional Occurrence of Type M Path Table */ set_num_732(bp+153, 0); /* Directory Record for Root Directory(BP 157 to 190) */ set_directory_record(bp+157, 190-157+1, vdd->rootent, iso9660, DIR_REC_VD, vdd->vdd_type); /* Volume Set Identifier */ r = set_str_d_characters_bp(a, bp, 191, 318, "", vdc); if (r != ARCHIVE_OK) return (r); /* Publisher Identifier */ r = set_file_identifier(bp, 319, 446, vdc, a, vdd, &(iso9660->publisher_identifier), "Publisher File", 1, A_CHAR); if (r != ARCHIVE_OK) return (r); /* Data Preparer Identifier */ r = set_file_identifier(bp, 447, 574, vdc, a, vdd, &(iso9660->data_preparer_identifier), "Data Preparer File", 1, A_CHAR); if (r != ARCHIVE_OK) return (r); /* Application Identifier */ r = set_file_identifier(bp, 575, 702, vdc, a, vdd, &(iso9660->application_identifier), "Application File", 1, A_CHAR); if (r != ARCHIVE_OK) return (r); /* Copyright File Identifier */ r = set_file_identifier(bp, 703, 739, vdc, a, vdd, &(iso9660->copyright_file_identifier), "Copyright File", 0, D_CHAR); if (r != ARCHIVE_OK) return (r); /* Abstract File Identifier */ r = set_file_identifier(bp, 740, 776, vdc, a, vdd, &(iso9660->abstract_file_identifier), "Abstract File", 0, D_CHAR); if (r != ARCHIVE_OK) return (r); /* Bibliongraphic File Identifier */ r = set_file_identifier(bp, 777, 813, vdc, a, vdd, &(iso9660->bibliographic_file_identifier), "Bibliongraphic File", 0, D_CHAR); if (r != ARCHIVE_OK) return (r); /* Volume Creation Date and Time */ set_date_time(bp+814, iso9660->birth_time); /* Volume Modification Date and Time */ set_date_time(bp+831, iso9660->birth_time); /* Volume Expiration Date and Time(obsolete) */ set_date_time_null(bp+848); /* Volume Effective Date and Time */ set_date_time(bp+865, iso9660->birth_time); /* File Structure Version */ bp[882] = fst_ver; /* Reserved */ bp[883] = 0; /* Application Use */ memset(bp + 884, 0x20, 1395 - 884 + 1); /* Reserved */ set_unused_field_bp(bp, 1396, LOGICAL_BLOCK_SIZE); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } /* * Write Boot Record Volume Descriptor */ static int write_VD_boot_record(struct archive_write *a) { struct iso9660 *iso9660; unsigned char *bp; iso9660 = a->format_data; bp = wb_buffptr(a) -1; /* Volume Descriptor Type */ set_VD_bp(bp, VDT_BOOT_RECORD, 1); /* Boot System Identifier */ memcpy(bp+8, "EL TORITO SPECIFICATION", 23); set_unused_field_bp(bp, 8+23, 39); /* Unused */ set_unused_field_bp(bp, 40, 71); /* Absolute pointer to first sector of Boot Catalog */ set_num_731(bp+72, iso9660->el_torito.catalog->file->content.location); /* Unused */ set_unused_field_bp(bp, 76, LOGICAL_BLOCK_SIZE); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } enum keytype { KEY_FLG, KEY_STR, KEY_INT, KEY_HEX, }; static void set_option_info(struct archive_string *info, int *opt, const char *key, enum keytype type, ...) { va_list ap; char prefix; const char *s; int d; prefix = (*opt==0)? ' ':','; va_start(ap, type); switch (type) { case KEY_FLG: d = va_arg(ap, int); archive_string_sprintf(info, "%c%s%s", prefix, (d == 0)?"!":"", key); break; case KEY_STR: s = va_arg(ap, const char *); archive_string_sprintf(info, "%c%s=%s", prefix, key, s); break; case KEY_INT: d = va_arg(ap, int); archive_string_sprintf(info, "%c%s=%d", prefix, key, d); break; case KEY_HEX: d = va_arg(ap, int); archive_string_sprintf(info, "%c%s=%x", prefix, key, d); break; } va_end(ap); *opt = 1; } /* * Make Non-ISO File System Information */ static int write_information_block(struct archive_write *a) { struct iso9660 *iso9660; char buf[128]; const char *v; int opt, r; struct archive_string info; size_t info_size = LOGICAL_BLOCK_SIZE * NON_ISO_FILE_SYSTEM_INFORMATION_BLOCK; iso9660 = (struct iso9660 *)a->format_data; if (info_size > wb_remaining(a)) { r = wb_write_out(a); if (r != ARCHIVE_OK) return (r); } archive_string_init(&info); if (archive_string_ensure(&info, info_size) == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } memset(info.s, 0, info_size); opt = 0; #if defined(HAVE__CTIME64_S) _ctime64_s(buf, sizeof(buf), &(iso9660->birth_time)); #elif defined(HAVE_CTIME_R) ctime_r(&(iso9660->birth_time), buf); #else strncpy(buf, ctime(&(iso9660->birth_time)), sizeof(buf)-1); buf[sizeof(buf)-1] = '\0'; #endif archive_string_sprintf(&info, "INFO %s%s", buf, archive_version_string()); if (iso9660->opt.abstract_file != OPT_ABSTRACT_FILE_DEFAULT) set_option_info(&info, &opt, "abstract-file", KEY_STR, iso9660->abstract_file_identifier.s); if (iso9660->opt.application_id != OPT_APPLICATION_ID_DEFAULT) set_option_info(&info, &opt, "application-id", KEY_STR, iso9660->application_identifier.s); if (iso9660->opt.allow_vernum != OPT_ALLOW_VERNUM_DEFAULT) set_option_info(&info, &opt, "allow-vernum", KEY_FLG, iso9660->opt.allow_vernum); if (iso9660->opt.biblio_file != OPT_BIBLIO_FILE_DEFAULT) set_option_info(&info, &opt, "biblio-file", KEY_STR, iso9660->bibliographic_file_identifier.s); if (iso9660->opt.boot != OPT_BOOT_DEFAULT) set_option_info(&info, &opt, "boot", KEY_STR, iso9660->el_torito.boot_filename.s); if (iso9660->opt.boot_catalog != OPT_BOOT_CATALOG_DEFAULT) set_option_info(&info, &opt, "boot-catalog", KEY_STR, iso9660->el_torito.catalog_filename.s); if (iso9660->opt.boot_info_table != OPT_BOOT_INFO_TABLE_DEFAULT) set_option_info(&info, &opt, "boot-info-table", KEY_FLG, iso9660->opt.boot_info_table); if (iso9660->opt.boot_load_seg != OPT_BOOT_LOAD_SEG_DEFAULT) set_option_info(&info, &opt, "boot-load-seg", KEY_HEX, iso9660->el_torito.boot_load_seg); if (iso9660->opt.boot_load_size != OPT_BOOT_LOAD_SIZE_DEFAULT) set_option_info(&info, &opt, "boot-load-size", KEY_INT, iso9660->el_torito.boot_load_size); if (iso9660->opt.boot_type != OPT_BOOT_TYPE_DEFAULT) { v = "no-emulation"; if (iso9660->opt.boot_type == OPT_BOOT_TYPE_FD) v = "fd"; if (iso9660->opt.boot_type == OPT_BOOT_TYPE_HARD_DISK) v = "hard-disk"; set_option_info(&info, &opt, "boot-type", KEY_STR, v); } #ifdef HAVE_ZLIB_H if (iso9660->opt.compression_level != OPT_COMPRESSION_LEVEL_DEFAULT) set_option_info(&info, &opt, "compression-level", KEY_INT, iso9660->zisofs.compression_level); #endif if (iso9660->opt.copyright_file != OPT_COPYRIGHT_FILE_DEFAULT) set_option_info(&info, &opt, "copyright-file", KEY_STR, iso9660->copyright_file_identifier.s); if (iso9660->opt.iso_level != OPT_ISO_LEVEL_DEFAULT) set_option_info(&info, &opt, "iso-level", KEY_INT, iso9660->opt.iso_level); if (iso9660->opt.joliet != OPT_JOLIET_DEFAULT) { if (iso9660->opt.joliet == OPT_JOLIET_LONGNAME) set_option_info(&info, &opt, "joliet", KEY_STR, "long"); else set_option_info(&info, &opt, "joliet", KEY_FLG, iso9660->opt.joliet); } if (iso9660->opt.limit_depth != OPT_LIMIT_DEPTH_DEFAULT) set_option_info(&info, &opt, "limit-depth", KEY_FLG, iso9660->opt.limit_depth); if (iso9660->opt.limit_dirs != OPT_LIMIT_DIRS_DEFAULT) set_option_info(&info, &opt, "limit-dirs", KEY_FLG, iso9660->opt.limit_dirs); if (iso9660->opt.pad != OPT_PAD_DEFAULT) set_option_info(&info, &opt, "pad", KEY_FLG, iso9660->opt.pad); if (iso9660->opt.publisher != OPT_PUBLISHER_DEFAULT) set_option_info(&info, &opt, "publisher", KEY_STR, iso9660->publisher_identifier.s); if (iso9660->opt.rr != OPT_RR_DEFAULT) { if (iso9660->opt.rr == OPT_RR_DISABLED) set_option_info(&info, &opt, "rockridge", KEY_FLG, iso9660->opt.rr); else if (iso9660->opt.rr == OPT_RR_STRICT) set_option_info(&info, &opt, "rockridge", KEY_STR, "strict"); else if (iso9660->opt.rr == OPT_RR_USEFUL) set_option_info(&info, &opt, "rockridge", KEY_STR, "useful"); } if (iso9660->opt.volume_id != OPT_VOLUME_ID_DEFAULT) set_option_info(&info, &opt, "volume-id", KEY_STR, iso9660->volume_identifier.s); if (iso9660->opt.zisofs != OPT_ZISOFS_DEFAULT) set_option_info(&info, &opt, "zisofs", KEY_FLG, iso9660->opt.zisofs); memcpy(wb_buffptr(a), info.s, info_size); archive_string_free(&info); return (wb_consume(a, info_size)); } static int write_rr_ER(struct archive_write *a) { unsigned char *p; p = wb_buffptr(a); memset(p, 0, LOGICAL_BLOCK_SIZE); p[0] = 'E'; p[1] = 'R'; p[3] = 0x01; p[2] = RRIP_ER_SIZE; p[4] = RRIP_ER_ID_SIZE; p[5] = RRIP_ER_DSC_SIZE; p[6] = RRIP_ER_SRC_SIZE; p[7] = 0x01; memcpy(&p[8], rrip_identifier, p[4]); memcpy(&p[8+p[4]], rrip_descriptor, p[5]); memcpy(&p[8+p[4]+p[5]], rrip_source, p[6]); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } static void calculate_path_table_size(struct vdd *vdd) { int depth, size; struct path_table *pt; pt = vdd->pathtbl; size = 0; for (depth = 0; depth < vdd->max_depth; depth++) { struct isoent **ptbl; int i, cnt; if ((cnt = pt[depth].cnt) == 0) break; ptbl = pt[depth].sorted; for (i = 0; i < cnt; i++) { int len; if (ptbl[i]->identifier == NULL) len = 1; /* root directory */ else len = ptbl[i]->id_len; if (len & 0x01) len++; /* Padding Field */ size += 8 + len; } } vdd->path_table_size = size; vdd->path_table_block = ((size + PATH_TABLE_BLOCK_SIZE -1) / PATH_TABLE_BLOCK_SIZE) * (PATH_TABLE_BLOCK_SIZE / LOGICAL_BLOCK_SIZE); } static int _write_path_table(struct archive_write *a, int type_m, int depth, struct vdd *vdd) { unsigned char *bp, *wb; struct isoent **ptbl; size_t wbremaining; int i, r, wsize; if (vdd->pathtbl[depth].cnt == 0) return (0); wsize = 0; wb = wb_buffptr(a); wbremaining = wb_remaining(a); bp = wb - 1; ptbl = vdd->pathtbl[depth].sorted; for (i = 0; i < vdd->pathtbl[depth].cnt; i++) { struct isoent *np; size_t len; np = ptbl[i]; if (np->identifier == NULL) len = 1; /* root directory */ else len = np->id_len; if (wbremaining - ((bp+1) - wb) < (len + 1 + 8)) { r = wb_consume(a, (bp+1) - wb); if (r < 0) return (r); wb = wb_buffptr(a); wbremaining = wb_remaining(a); bp = wb -1; } /* Length of Directory Identifier */ set_num_711(bp+1, len); /* Extended Attribute Record Length */ set_num_711(bp+2, 0); /* Location of Extent */ if (type_m) set_num_732(bp+3, np->dir_location); else set_num_731(bp+3, np->dir_location); /* Parent Directory Number */ if (type_m) set_num_722(bp+7, np->parent->dir_number); else set_num_721(bp+7, np->parent->dir_number); /* Directory Identifier */ if (np->identifier == NULL) bp[9] = 0; else memcpy(&bp[9], np->identifier, len); if (len & 0x01) { /* Padding Field */ bp[9+len] = 0; len++; } wsize += 8 + len; bp += 8 + len; } if ((bp + 1) > wb) { r = wb_consume(a, (bp+1)-wb); if (r < 0) return (r); } return (wsize); } static int write_path_table(struct archive_write *a, int type_m, struct vdd *vdd) { int depth, r; size_t path_table_size; r = ARCHIVE_OK; path_table_size = 0; for (depth = 0; depth < vdd->max_depth; depth++) { r = _write_path_table(a, type_m, depth, vdd); if (r < 0) return (r); path_table_size += r; } /* Write padding data. */ path_table_size = path_table_size % PATH_TABLE_BLOCK_SIZE; if (path_table_size > 0) r = write_null(a, PATH_TABLE_BLOCK_SIZE - path_table_size); return (r); } static int calculate_directory_descriptors(struct iso9660 *iso9660, struct vdd *vdd, struct isoent *isoent, int depth) { struct isoent **enttbl; int bs, block, i; block = 1; bs = get_dir_rec_size(iso9660, isoent, DIR_REC_SELF, vdd->vdd_type); bs += get_dir_rec_size(iso9660, isoent, DIR_REC_PARENT, vdd->vdd_type); if (isoent->children.cnt <= 0 || (vdd->vdd_type != VDD_JOLIET && !iso9660->opt.rr && depth + 1 >= vdd->max_depth)) return (block); enttbl = isoent->children_sorted; for (i = 0; i < isoent->children.cnt; i++) { struct isoent *np = enttbl[i]; struct isofile *file; file = np->file; if (file->hardlink_target != NULL) file = file->hardlink_target; file->cur_content = &(file->content); do { int dr_l; dr_l = get_dir_rec_size(iso9660, np, DIR_REC_NORMAL, vdd->vdd_type); if ((bs + dr_l) > LOGICAL_BLOCK_SIZE) { block ++; bs = dr_l; } else bs += dr_l; file->cur_content = file->cur_content->next; } while (file->cur_content != NULL); } return (block); } static int _write_directory_descriptors(struct archive_write *a, struct vdd *vdd, struct isoent *isoent, int depth) { struct iso9660 *iso9660 = a->format_data; struct isoent **enttbl; unsigned char *p, *wb; int i, r; int dr_l; p = wb = wb_buffptr(a); #define WD_REMAINING (LOGICAL_BLOCK_SIZE - (p - wb)) p += set_directory_record(p, WD_REMAINING, isoent, iso9660, DIR_REC_SELF, vdd->vdd_type); p += set_directory_record(p, WD_REMAINING, isoent, iso9660, DIR_REC_PARENT, vdd->vdd_type); if (isoent->children.cnt <= 0 || (vdd->vdd_type != VDD_JOLIET && !iso9660->opt.rr && depth + 1 >= vdd->max_depth)) { memset(p, 0, WD_REMAINING); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } enttbl = isoent->children_sorted; for (i = 0; i < isoent->children.cnt; i++) { struct isoent *np = enttbl[i]; struct isofile *file = np->file; if (file->hardlink_target != NULL) file = file->hardlink_target; file->cur_content = &(file->content); do { dr_l = set_directory_record(p, WD_REMAINING, np, iso9660, DIR_REC_NORMAL, vdd->vdd_type); if (dr_l == 0) { memset(p, 0, WD_REMAINING); r = wb_consume(a, LOGICAL_BLOCK_SIZE); if (r < 0) return (r); p = wb = wb_buffptr(a); dr_l = set_directory_record(p, WD_REMAINING, np, iso9660, DIR_REC_NORMAL, vdd->vdd_type); } p += dr_l; file->cur_content = file->cur_content->next; } while (file->cur_content != NULL); } memset(p, 0, WD_REMAINING); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } static int write_directory_descriptors(struct archive_write *a, struct vdd *vdd) { struct isoent *np; int depth, r; depth = 0; np = vdd->rootent; do { struct extr_rec *extr; r = _write_directory_descriptors(a, vdd, np, depth); if (r < 0) return (r); if (vdd->vdd_type != VDD_JOLIET) { /* * This extract record is used by SUSP,RRIP. * Not for joliet. */ for (extr = np->extr_rec_list.first; extr != NULL; extr = extr->next) { unsigned char *wb; wb = wb_buffptr(a); memcpy(wb, extr->buf, extr->offset); memset(wb + extr->offset, 0, LOGICAL_BLOCK_SIZE - extr->offset); r = wb_consume(a, LOGICAL_BLOCK_SIZE); if (r < 0) return (r); } } if (np->subdirs.first != NULL && depth + 1 < vdd->max_depth) { /* Enter to sub directories. */ np = np->subdirs.first; depth++; continue; } while (np != np->parent) { if (np->drnext == NULL) { /* Return to the parent directory. */ np = np->parent; depth--; } else { np = np->drnext; break; } } } while (np != np->parent); return (ARCHIVE_OK); } /* * Read file contents from the temporary file, and write it. */ static int write_file_contents(struct archive_write *a, int64_t offset, int64_t size) { struct iso9660 *iso9660 = a->format_data; int r; lseek(iso9660->temp_fd, offset, SEEK_SET); while (size) { size_t rsize; ssize_t rs; unsigned char *wb; wb = wb_buffptr(a); rsize = wb_remaining(a); if (rsize > (size_t)size) rsize = (size_t)size; rs = read(iso9660->temp_fd, wb, rsize); if (rs <= 0) { archive_set_error(&a->archive, errno, "Can't read temporary file(%jd)", (intmax_t)rs); return (ARCHIVE_FATAL); } size -= rs; r = wb_consume(a, rs); if (r < 0) return (r); } return (ARCHIVE_OK); } static int write_file_descriptors(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; struct isofile *file; int64_t blocks, offset; int r; blocks = 0; offset = 0; /* Make the boot catalog contents, and write it. */ if (iso9660->el_torito.catalog != NULL) { r = make_boot_catalog(a); if (r < 0) return (r); } /* Write the boot file contents. */ if (iso9660->el_torito.boot != NULL) { file = iso9660->el_torito.boot->file; blocks = file->content.blocks; offset = file->content.offset_of_temp; if (offset != 0) { r = write_file_contents(a, offset, blocks << LOGICAL_BLOCK_BITS); if (r < 0) return (r); blocks = 0; offset = 0; } } /* Write out all file contents. */ for (file = iso9660->data_file_list.first; file != NULL; file = file->datanext) { if (!file->write_content) continue; if ((offset + (blocks << LOGICAL_BLOCK_BITS)) < file->content.offset_of_temp) { if (blocks > 0) { r = write_file_contents(a, offset, blocks << LOGICAL_BLOCK_BITS); if (r < 0) return (r); } blocks = 0; offset = file->content.offset_of_temp; } file->cur_content = &(file->content); do { blocks += file->cur_content->blocks; /* Next fragument */ file->cur_content = file->cur_content->next; } while (file->cur_content != NULL); } /* Flush out remaining blocks. */ if (blocks > 0) { r = write_file_contents(a, offset, blocks << LOGICAL_BLOCK_BITS); if (r < 0) return (r); } return (ARCHIVE_OK); } static void isofile_init_entry_list(struct iso9660 *iso9660) { iso9660->all_file_list.first = NULL; iso9660->all_file_list.last = &(iso9660->all_file_list.first); } static void isofile_add_entry(struct iso9660 *iso9660, struct isofile *file) { file->allnext = NULL; *iso9660->all_file_list.last = file; iso9660->all_file_list.last = &(file->allnext); } static void isofile_free_all_entries(struct iso9660 *iso9660) { struct isofile *file, *file_next; file = iso9660->all_file_list.first; while (file != NULL) { file_next = file->allnext; isofile_free(file); file = file_next; } } static void isofile_init_entry_data_file_list(struct iso9660 *iso9660) { iso9660->data_file_list.first = NULL; iso9660->data_file_list.last = &(iso9660->data_file_list.first); } static void isofile_add_data_file(struct iso9660 *iso9660, struct isofile *file) { file->datanext = NULL; *iso9660->data_file_list.last = file; iso9660->data_file_list.last = &(file->datanext); } static struct isofile * isofile_new(struct archive_write *a, struct archive_entry *entry) { struct isofile *file; file = calloc(1, sizeof(*file)); if (file == NULL) return (NULL); if (entry != NULL) file->entry = archive_entry_clone(entry); else file->entry = archive_entry_new2(&a->archive); if (file->entry == NULL) { free(file); return (NULL); } archive_string_init(&(file->parentdir)); archive_string_init(&(file->basename)); archive_string_init(&(file->basename_utf16)); archive_string_init(&(file->symlink)); file->cur_content = &(file->content); return (file); } static void isofile_free(struct isofile *file) { struct content *con, *tmp; con = file->content.next; while (con != NULL) { tmp = con; con = con->next; free(tmp); } archive_entry_free(file->entry); archive_string_free(&(file->parentdir)); archive_string_free(&(file->basename)); archive_string_free(&(file->basename_utf16)); archive_string_free(&(file->symlink)); free(file); } #if defined(_WIN32) || defined(__CYGWIN__) static int cleanup_backslash_1(char *p) { int mb, dos; mb = dos = 0; while (*p) { if (*(unsigned char *)p > 127) mb = 1; if (*p == '\\') { /* If we have not met any multi-byte characters, * we can replace '\' with '/'. */ if (!mb) *p = '/'; dos = 1; } p++; } if (!mb || !dos) return (0); return (-1); } static void cleanup_backslash_2(wchar_t *p) { /* Convert a path-separator from '\' to '/' */ while (*p != L'\0') { if (*p == L'\\') *p = L'/'; p++; } } #endif /* * Generate a parent directory name and a base name from a pathname. */ static int isofile_gen_utility_names(struct archive_write *a, struct isofile *file) { struct iso9660 *iso9660; const char *pathname; char *p, *dirname, *slash; size_t len; int ret = ARCHIVE_OK; iso9660 = a->format_data; archive_string_empty(&(file->parentdir)); archive_string_empty(&(file->basename)); archive_string_empty(&(file->basename_utf16)); archive_string_empty(&(file->symlink)); pathname = archive_entry_pathname(file->entry); if (pathname == NULL || pathname[0] == '\0') {/* virtual root */ file->dircnt = 0; return (ret); } /* * Make a UTF-16BE basename if Joliet extension enabled. */ if (iso9660->opt.joliet) { const char *u16, *ulast; size_t u16len, ulen_last; if (iso9660->sconv_to_utf16be == NULL) { iso9660->sconv_to_utf16be = archive_string_conversion_to_charset( &(a->archive), "UTF-16BE", 1); if (iso9660->sconv_to_utf16be == NULL) /* Couldn't allocate memory */ return (ARCHIVE_FATAL); iso9660->sconv_from_utf16be = archive_string_conversion_from_charset( &(a->archive), "UTF-16BE", 1); if (iso9660->sconv_from_utf16be == NULL) /* Couldn't allocate memory */ return (ARCHIVE_FATAL); } /* * Converte a filename to UTF-16BE. */ if (0 > archive_entry_pathname_l(file->entry, &u16, &u16len, iso9660->sconv_to_utf16be)) { if (errno == ENOMEM) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for UTF-16BE"); return (ARCHIVE_FATAL); } archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "A filename cannot be converted to UTF-16BE;" "You should disable making Joliet extension"); ret = ARCHIVE_WARN; } /* * Make sure a path separator is not in the last; * Remove trailing '/'. */ while (u16len >= 2) { #if defined(_WIN32) || defined(__CYGWIN__) if (u16[u16len-2] == 0 && (u16[u16len-1] == '/' || u16[u16len-1] == '\\')) #else if (u16[u16len-2] == 0 && u16[u16len-1] == '/') #endif { u16len -= 2; } else break; } /* * Find a basename in UTF-16BE. */ ulast = u16; u16len >>= 1; ulen_last = u16len; while (u16len > 0) { #if defined(_WIN32) || defined(__CYGWIN__) if (u16[0] == 0 && (u16[1] == '/' || u16[1] == '\\')) #else if (u16[0] == 0 && u16[1] == '/') #endif { ulast = u16 + 2; ulen_last = u16len -1; } u16 += 2; u16len --; } ulen_last <<= 1; if (archive_string_ensure(&(file->basename_utf16), ulen_last) == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory for UTF-16BE"); return (ARCHIVE_FATAL); } /* * Set UTF-16BE basename. */ memcpy(file->basename_utf16.s, ulast, ulen_last); file->basename_utf16.length = ulen_last; } archive_strcpy(&(file->parentdir), pathname); #if defined(_WIN32) || defined(__CYGWIN__) /* * Convert a path-separator from '\' to '/' */ if (cleanup_backslash_1(file->parentdir.s) != 0) { const wchar_t *wp = archive_entry_pathname_w(file->entry); struct archive_wstring ws; if (wp != NULL) { archive_string_init(&ws); archive_wstrcpy(&ws, wp); cleanup_backslash_2(ws.s); archive_string_empty(&(file->parentdir)); archive_string_append_from_wcs(&(file->parentdir), ws.s, ws.length); archive_wstring_free(&ws); } } #endif len = file->parentdir.length; p = dirname = file->parentdir.s; /* * Remove leading '/', '../' and './' elements */ while (*p) { if (p[0] == '/') { p++; len--; } else if (p[0] != '.') break; else if (p[1] == '.' && p[2] == '/') { p += 3; len -= 3; } else if (p[1] == '/' || (p[1] == '.' && p[2] == '\0')) { p += 2; len -= 2; } else if (p[1] == '\0') { p++; len--; } else break; } if (p != dirname) { memmove(dirname, p, len+1); p = dirname; } /* * Remove "/","/." and "/.." elements from tail. */ while (len > 0) { size_t ll = len; if (len > 0 && p[len-1] == '/') { p[len-1] = '\0'; len--; } if (len > 1 && p[len-2] == '/' && p[len-1] == '.') { p[len-2] = '\0'; len -= 2; } if (len > 2 && p[len-3] == '/' && p[len-2] == '.' && p[len-1] == '.') { p[len-3] = '\0'; len -= 3; } if (ll == len) break; } while (*p) { if (p[0] == '/') { if (p[1] == '/') /* Convert '//' --> '/' */ strcpy(p, p+1); else if (p[1] == '.' && p[2] == '/') /* Convert '/./' --> '/' */ strcpy(p, p+2); else if (p[1] == '.' && p[2] == '.' && p[3] == '/') { /* Convert 'dir/dir1/../dir2/' * --> 'dir/dir2/' */ char *rp = p -1; while (rp >= dirname) { if (*rp == '/') break; --rp; } if (rp > dirname) { strcpy(rp, p+3); p = rp; } else { strcpy(dirname, p+4); p = dirname; } } else p++; } else p++; } p = dirname; len = strlen(p); if (archive_entry_filetype(file->entry) == AE_IFLNK) { /* Convert symlink name too. */ pathname = archive_entry_symlink(file->entry); archive_strcpy(&(file->symlink), pathname); #if defined(_WIN32) || defined(__CYGWIN__) /* * Convert a path-separator from '\' to '/' */ if (archive_strlen(&(file->symlink)) > 0 && cleanup_backslash_1(file->symlink.s) != 0) { const wchar_t *wp = archive_entry_symlink_w(file->entry); struct archive_wstring ws; if (wp != NULL) { archive_string_init(&ws); archive_wstrcpy(&ws, wp); cleanup_backslash_2(ws.s); archive_string_empty(&(file->symlink)); archive_string_append_from_wcs( &(file->symlink), ws.s, ws.length); archive_wstring_free(&ws); } } #endif } /* * - Count up directory elements. * - Find out the position which points the last position of * path separator('/'). */ slash = NULL; file->dircnt = 0; for (; *p != '\0'; p++) if (*p == '/') { slash = p; file->dircnt++; } if (slash == NULL) { /* The pathname doesn't have a parent directory. */ file->parentdir.length = len; archive_string_copy(&(file->basename), &(file->parentdir)); archive_string_empty(&(file->parentdir)); *file->parentdir.s = '\0'; return (ret); } /* Make a basename from dirname and slash */ *slash = '\0'; file->parentdir.length = slash - dirname; archive_strcpy(&(file->basename), slash + 1); if (archive_entry_filetype(file->entry) == AE_IFDIR) file->dircnt ++; return (ret); } /* * Register a entry to get a hardlink target. */ static int isofile_register_hardlink(struct archive_write *a, struct isofile *file) { struct iso9660 *iso9660 = a->format_data; struct hardlink *hl; const char *pathname; archive_entry_set_nlink(file->entry, 1); pathname = archive_entry_hardlink(file->entry); if (pathname == NULL) { /* This `file` is a hardlink target. */ hl = malloc(sizeof(*hl)); if (hl == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } hl->nlink = 1; /* A hardlink target must be the first position. */ file->hlnext = NULL; hl->file_list.first = file; hl->file_list.last = &(file->hlnext); __archive_rb_tree_insert_node(&(iso9660->hardlink_rbtree), (struct archive_rb_node *)hl); } else { hl = (struct hardlink *)__archive_rb_tree_find_node( &(iso9660->hardlink_rbtree), pathname); if (hl != NULL) { /* Insert `file` entry into the tail. */ file->hlnext = NULL; *hl->file_list.last = file; hl->file_list.last = &(file->hlnext); hl->nlink++; } archive_entry_unset_size(file->entry); } return (ARCHIVE_OK); } /* * Hardlinked files have to have the same location of extent. * We have to find out hardlink target entries for the entries * which have a hardlink target name. */ static void isofile_connect_hardlink_files(struct iso9660 *iso9660) { struct archive_rb_node *n; struct hardlink *hl; struct isofile *target, *nf; ARCHIVE_RB_TREE_FOREACH(n, &(iso9660->hardlink_rbtree)) { hl = (struct hardlink *)n; /* The first entry must be a hardlink target. */ target = hl->file_list.first; archive_entry_set_nlink(target->entry, hl->nlink); /* Set a hardlink target to reference entries. */ for (nf = target->hlnext; nf != NULL; nf = nf->hlnext) { nf->hardlink_target = target; archive_entry_set_nlink(nf->entry, hl->nlink); } } } static int isofile_hd_cmp_node(const struct archive_rb_node *n1, const struct archive_rb_node *n2) { const struct hardlink *h1 = (const struct hardlink *)n1; const struct hardlink *h2 = (const struct hardlink *)n2; return (strcmp(archive_entry_pathname(h1->file_list.first->entry), archive_entry_pathname(h2->file_list.first->entry))); } static int isofile_hd_cmp_key(const struct archive_rb_node *n, const void *key) { const struct hardlink *h = (const struct hardlink *)n; return (strcmp(archive_entry_pathname(h->file_list.first->entry), (const char *)key)); } static void isofile_init_hardlinks(struct iso9660 *iso9660) { static const struct archive_rb_tree_ops rb_ops = { isofile_hd_cmp_node, isofile_hd_cmp_key, }; __archive_rb_tree_init(&(iso9660->hardlink_rbtree), &rb_ops); } static void isofile_free_hardlinks(struct iso9660 *iso9660) { struct archive_rb_node *n, *next; for (n = ARCHIVE_RB_TREE_MIN(&(iso9660->hardlink_rbtree)); n;) { next = __archive_rb_tree_iterate(&(iso9660->hardlink_rbtree), n, ARCHIVE_RB_DIR_RIGHT); free(n); n = next; } } static struct isoent * isoent_new(struct isofile *file) { struct isoent *isoent; static const struct archive_rb_tree_ops rb_ops = { isoent_cmp_node, isoent_cmp_key, }; isoent = calloc(1, sizeof(*isoent)); if (isoent == NULL) return (NULL); isoent->file = file; isoent->children.first = NULL; isoent->children.last = &(isoent->children.first); __archive_rb_tree_init(&(isoent->rbtree), &rb_ops); isoent->subdirs.first = NULL; isoent->subdirs.last = &(isoent->subdirs.first); isoent->extr_rec_list.first = NULL; isoent->extr_rec_list.last = &(isoent->extr_rec_list.first); isoent->extr_rec_list.current = NULL; if (archive_entry_filetype(file->entry) == AE_IFDIR) isoent->dir = 1; return (isoent); } static inline struct isoent * isoent_clone(struct isoent *src) { return (isoent_new(src->file)); } static void _isoent_free(struct isoent *isoent) { struct extr_rec *er, *er_next; free(isoent->children_sorted); free(isoent->identifier); er = isoent->extr_rec_list.first; while (er != NULL) { er_next = er->next; free(er); er = er_next; } free(isoent); } static void isoent_free_all(struct isoent *isoent) { struct isoent *np, *np_temp; if (isoent == NULL) return; np = isoent; for (;;) { if (np->dir) { if (np->children.first != NULL) { /* Enter to sub directories. */ np = np->children.first; continue; } } for (;;) { np_temp = np; if (np->chnext == NULL) { /* Return to the parent directory. */ np = np->parent; _isoent_free(np_temp); if (np == np_temp) return; } else { np = np->chnext; _isoent_free(np_temp); break; } } } } static struct isoent * isoent_create_virtual_dir(struct archive_write *a, struct iso9660 *iso9660, const char *pathname) { struct isofile *file; struct isoent *isoent; file = isofile_new(a, NULL); if (file == NULL) return (NULL); archive_entry_set_pathname(file->entry, pathname); archive_entry_unset_mtime(file->entry); archive_entry_unset_atime(file->entry); archive_entry_unset_ctime(file->entry); archive_entry_set_uid(file->entry, getuid()); archive_entry_set_gid(file->entry, getgid()); archive_entry_set_mode(file->entry, 0555 | AE_IFDIR); archive_entry_set_nlink(file->entry, 2); if (isofile_gen_utility_names(a, file) < ARCHIVE_WARN) { isofile_free(file); return (NULL); } isofile_add_entry(iso9660, file); isoent = isoent_new(file); if (isoent == NULL) return (NULL); isoent->dir = 1; isoent->virtual = 1; return (isoent); } static int isoent_cmp_node(const struct archive_rb_node *n1, const struct archive_rb_node *n2) { const struct isoent *e1 = (const struct isoent *)n1; const struct isoent *e2 = (const struct isoent *)n2; return (strcmp(e1->file->basename.s, e2->file->basename.s)); } static int isoent_cmp_key(const struct archive_rb_node *n, const void *key) { const struct isoent *e = (const struct isoent *)n; return (strcmp(e->file->basename.s, (const char *)key)); } static int isoent_add_child_head(struct isoent *parent, struct isoent *child) { if (!__archive_rb_tree_insert_node( &(parent->rbtree), (struct archive_rb_node *)child)) return (0); if ((child->chnext = parent->children.first) == NULL) parent->children.last = &(child->chnext); parent->children.first = child; parent->children.cnt++; child->parent = parent; /* Add a child to a sub-directory chain */ if (child->dir) { if ((child->drnext = parent->subdirs.first) == NULL) parent->subdirs.last = &(child->drnext); parent->subdirs.first = child; parent->subdirs.cnt++; child->parent = parent; } else child->drnext = NULL; return (1); } static int isoent_add_child_tail(struct isoent *parent, struct isoent *child) { if (!__archive_rb_tree_insert_node( &(parent->rbtree), (struct archive_rb_node *)child)) return (0); child->chnext = NULL; *parent->children.last = child; parent->children.last = &(child->chnext); parent->children.cnt++; child->parent = parent; /* Add a child to a sub-directory chain */ child->drnext = NULL; if (child->dir) { *parent->subdirs.last = child; parent->subdirs.last = &(child->drnext); parent->subdirs.cnt++; child->parent = parent; } return (1); } static void isoent_remove_child(struct isoent *parent, struct isoent *child) { struct isoent *ent; /* Remove a child entry from children chain. */ ent = parent->children.first; while (ent->chnext != child) ent = ent->chnext; if ((ent->chnext = ent->chnext->chnext) == NULL) parent->children.last = &(ent->chnext); parent->children.cnt--; if (child->dir) { /* Remove a child entry from sub-directory chain. */ ent = parent->subdirs.first; while (ent->drnext != child) ent = ent->drnext; if ((ent->drnext = ent->drnext->drnext) == NULL) parent->subdirs.last = &(ent->drnext); parent->subdirs.cnt--; } __archive_rb_tree_remove_node(&(parent->rbtree), (struct archive_rb_node *)child); } static int isoent_clone_tree(struct archive_write *a, struct isoent **nroot, struct isoent *root) { struct isoent *np, *xroot, *newent; np = root; xroot = NULL; do { newent = isoent_clone(np); if (newent == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } if (xroot == NULL) { *nroot = xroot = newent; newent->parent = xroot; } else isoent_add_child_tail(xroot, newent); if (np->dir && np->children.first != NULL) { /* Enter to sub directories. */ np = np->children.first; xroot = newent; continue; } while (np != np->parent) { if (np->chnext == NULL) { /* Return to the parent directory. */ np = np->parent; xroot = xroot->parent; } else { np = np->chnext; break; } } } while (np != np->parent); return (ARCHIVE_OK); } /* * Setup directory locations. */ static void isoent_setup_directory_location(struct iso9660 *iso9660, int location, struct vdd *vdd) { struct isoent *np; int depth; vdd->total_dir_block = 0; depth = 0; np = vdd->rootent; do { int block; np->dir_block = calculate_directory_descriptors( iso9660, vdd, np, depth); vdd->total_dir_block += np->dir_block; np->dir_location = location; location += np->dir_block; block = extra_setup_location(np, location); vdd->total_dir_block += block; location += block; if (np->subdirs.first != NULL && depth + 1 < vdd->max_depth) { /* Enter to sub directories. */ np = np->subdirs.first; depth++; continue; } while (np != np->parent) { if (np->drnext == NULL) { /* Return to the parent directory. */ np = np->parent; depth--; } else { np = np->drnext; break; } } } while (np != np->parent); } static void _isoent_file_location(struct iso9660 *iso9660, struct isoent *isoent, int *symlocation) { struct isoent **children; int n; if (isoent->children.cnt == 0) return; children = isoent->children_sorted; for (n = 0; n < isoent->children.cnt; n++) { struct isoent *np; struct isofile *file; np = children[n]; if (np->dir) continue; if (np == iso9660->el_torito.boot) continue; file = np->file; if (file->boot || file->hardlink_target != NULL) continue; if (archive_entry_filetype(file->entry) == AE_IFLNK || file->content.size == 0) { /* * Do not point a valid location. * Make sure entry is not hardlink file. */ file->content.location = (*symlocation)--; continue; } file->write_content = 1; } } /* * Setup file locations. */ static void isoent_setup_file_location(struct iso9660 *iso9660, int location) { struct isoent *isoent; struct isoent *np; struct isofile *file; size_t size; int block; int depth; int joliet; int symlocation; int total_block; iso9660->total_file_block = 0; if ((isoent = iso9660->el_torito.catalog) != NULL) { isoent->file->content.location = location; block = (archive_entry_size(isoent->file->entry) + LOGICAL_BLOCK_SIZE -1) >> LOGICAL_BLOCK_BITS; location += block; iso9660->total_file_block += block; } if ((isoent = iso9660->el_torito.boot) != NULL) { isoent->file->content.location = location; size = fd_boot_image_size(iso9660->el_torito.media_type); if (size == 0) size = archive_entry_size(isoent->file->entry); block = (size + LOGICAL_BLOCK_SIZE -1) >> LOGICAL_BLOCK_BITS; location += block; iso9660->total_file_block += block; isoent->file->content.blocks = block; } depth = 0; symlocation = -16; if (!iso9660->opt.rr && iso9660->opt.joliet) { joliet = 1; np = iso9660->joliet.rootent; } else { joliet = 0; np = iso9660->primary.rootent; } do { _isoent_file_location(iso9660, np, &symlocation); if (np->subdirs.first != NULL && (joliet || ((iso9660->opt.rr == OPT_RR_DISABLED && depth + 2 < iso9660->primary.max_depth) || (iso9660->opt.rr && depth + 1 < iso9660->primary.max_depth)))) { /* Enter to sub directories. */ np = np->subdirs.first; depth++; continue; } while (np != np->parent) { if (np->drnext == NULL) { /* Return to the parent directory. */ np = np->parent; depth--; } else { np = np->drnext; break; } } } while (np != np->parent); total_block = 0; for (file = iso9660->data_file_list.first; file != NULL; file = file->datanext) { if (!file->write_content) continue; file->cur_content = &(file->content); do { file->cur_content->location = location; location += file->cur_content->blocks; total_block += file->cur_content->blocks; /* Next fragument */ file->cur_content = file->cur_content->next; } while (file->cur_content != NULL); } iso9660->total_file_block += total_block; } static int get_path_component(char *name, int n, const char *fn) { char *p; int l; p = strchr(fn, '/'); if (p == NULL) { if ((l = strlen(fn)) == 0) return (0); } else l = p - fn; if (l > n -1) return (-1); memcpy(name, fn, l); name[l] = '\0'; return (l); } /* * Add a new entry into the tree. */ static int isoent_tree(struct archive_write *a, struct isoent **isoentpp) { #if defined(_WIN32) && !defined(__CYGWIN__) char name[_MAX_FNAME];/* Included null terminator size. */ #elif defined(NAME_MAX) && NAME_MAX >= 255 char name[NAME_MAX+1]; #else char name[256]; #endif struct iso9660 *iso9660 = a->format_data; struct isoent *dent, *isoent, *np; struct isofile *f1, *f2; const char *fn, *p; int l; isoent = *isoentpp; dent = iso9660->primary.rootent; if (isoent->file->parentdir.length > 0) fn = p = isoent->file->parentdir.s; else fn = p = ""; /* * If the path of the parent directory of `isoent' entry is * the same as the path of `cur_dirent', add isoent to * `cur_dirent'. */ if (archive_strlen(&(iso9660->cur_dirstr)) == archive_strlen(&(isoent->file->parentdir)) && strcmp(iso9660->cur_dirstr.s, fn) == 0) { if (!isoent_add_child_tail(iso9660->cur_dirent, isoent)) { np = (struct isoent *)__archive_rb_tree_find_node( &(iso9660->cur_dirent->rbtree), isoent->file->basename.s); goto same_entry; } return (ARCHIVE_OK); } for (;;) { l = get_path_component(name, sizeof(name), fn); if (l == 0) { np = NULL; break; } if (l < 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "A name buffer is too small"); _isoent_free(isoent); return (ARCHIVE_FATAL); } np = isoent_find_child(dent, name); if (np == NULL || fn[0] == '\0') break; /* Find next subdirectory. */ if (!np->dir) { /* NOT Directory! */ archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "`%s' is not directory, we cannot insert `%s' ", archive_entry_pathname(np->file->entry), archive_entry_pathname(isoent->file->entry)); _isoent_free(isoent); *isoentpp = NULL; return (ARCHIVE_FAILED); } fn += l; if (fn[0] == '/') fn++; dent = np; } if (np == NULL) { /* * Create virtual parent directories. */ while (fn[0] != '\0') { struct isoent *vp; struct archive_string as; archive_string_init(&as); archive_strncat(&as, p, fn - p + l); if (as.s[as.length-1] == '/') { as.s[as.length-1] = '\0'; as.length--; } vp = isoent_create_virtual_dir(a, iso9660, as.s); if (vp == NULL) { archive_string_free(&as); archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); _isoent_free(isoent); *isoentpp = NULL; return (ARCHIVE_FATAL); } archive_string_free(&as); if (vp->file->dircnt > iso9660->dircnt_max) iso9660->dircnt_max = vp->file->dircnt; isoent_add_child_tail(dent, vp); np = vp; fn += l; if (fn[0] == '/') fn++; l = get_path_component(name, sizeof(name), fn); if (l < 0) { archive_string_free(&as); archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "A name buffer is too small"); _isoent_free(isoent); *isoentpp = NULL; return (ARCHIVE_FATAL); } dent = np; } /* Found out the parent directory where isoent can be * inserted. */ iso9660->cur_dirent = dent; archive_string_empty(&(iso9660->cur_dirstr)); archive_string_ensure(&(iso9660->cur_dirstr), archive_strlen(&(dent->file->parentdir)) + archive_strlen(&(dent->file->basename)) + 2); if (archive_strlen(&(dent->file->parentdir)) + archive_strlen(&(dent->file->basename)) == 0) iso9660->cur_dirstr.s[0] = 0; else { if (archive_strlen(&(dent->file->parentdir)) > 0) { archive_string_copy(&(iso9660->cur_dirstr), &(dent->file->parentdir)); archive_strappend_char(&(iso9660->cur_dirstr), '/'); } archive_string_concat(&(iso9660->cur_dirstr), &(dent->file->basename)); } if (!isoent_add_child_tail(dent, isoent)) { np = (struct isoent *)__archive_rb_tree_find_node( &(dent->rbtree), isoent->file->basename.s); goto same_entry; } return (ARCHIVE_OK); } same_entry: /* * We have already has the entry the filename of which is * the same. */ f1 = np->file; f2 = isoent->file; /* If the file type of entries is different, * we cannot handle it. */ if (archive_entry_filetype(f1->entry) != archive_entry_filetype(f2->entry)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Found duplicate entries `%s' and its file type is " "different", archive_entry_pathname(f1->entry)); _isoent_free(isoent); *isoentpp = NULL; return (ARCHIVE_FAILED); } /* Swap file entries. */ np->file = f2; isoent->file = f1; np->virtual = 0; _isoent_free(isoent); *isoentpp = np; return (ARCHIVE_OK); } /* * Find a entry from `isoent' */ static struct isoent * isoent_find_child(struct isoent *isoent, const char *child_name) { struct isoent *np; np = (struct isoent *)__archive_rb_tree_find_node( &(isoent->rbtree), child_name); return (np); } /* * Find a entry full-path of which is specified by `fn' parameter, * in the tree. */ static struct isoent * isoent_find_entry(struct isoent *rootent, const char *fn) { #if defined(_WIN32) && !defined(__CYGWIN__) char name[_MAX_FNAME];/* Included null terminator size. */ #elif defined(NAME_MAX) && NAME_MAX >= 255 char name[NAME_MAX+1]; #else char name[256]; #endif struct isoent *isoent, *np; int l; isoent = rootent; np = NULL; for (;;) { l = get_path_component(name, sizeof(name), fn); if (l == 0) break; fn += l; if (fn[0] == '/') fn++; np = isoent_find_child(isoent, name); if (np == NULL) break; if (fn[0] == '\0') break;/* We found out the entry */ /* Try sub directory. */ isoent = np; np = NULL; if (!isoent->dir) break;/* Not directory */ } return (np); } /* * Following idr_* functions are used for resolving duplicated filenames * and unreceivable filenames to generate ISO9660/Joliet Identifiers. */ static void idr_relaxed_filenames(char *map) { int i; for (i = 0x21; i <= 0x2F; i++) map[i] = 1; for (i = 0x3A; i <= 0x41; i++) map[i] = 1; for (i = 0x5B; i <= 0x5E; i++) map[i] = 1; map[0x60] = 1; for (i = 0x7B; i <= 0x7E; i++) map[i] = 1; } static void idr_init(struct iso9660 *iso9660, struct vdd *vdd, struct idr *idr) { idr->idrent_pool = NULL; idr->pool_size = 0; if (vdd->vdd_type != VDD_JOLIET) { if (iso9660->opt.iso_level <= 3) { memcpy(idr->char_map, d_characters_map, sizeof(idr->char_map)); } else { memcpy(idr->char_map, d1_characters_map, sizeof(idr->char_map)); idr_relaxed_filenames(idr->char_map); } } } static void idr_cleanup(struct idr *idr) { free(idr->idrent_pool); } static int idr_ensure_poolsize(struct archive_write *a, struct idr *idr, int cnt) { if (idr->pool_size < cnt) { const int bk = (1 << 7) - 1; int psize; psize = (cnt + bk) & ~bk; idr->idrent_pool = realloc(idr->idrent_pool, sizeof(struct idrent) * psize); if (idr->idrent_pool == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } idr->pool_size = psize; } return (ARCHIVE_OK); } static int idr_start(struct archive_write *a, struct idr *idr, int cnt, int ffmax, int num_size, int null_size, const struct archive_rb_tree_ops *rbt_ops) { int r; (void)ffmax; /* UNUSED */ r = idr_ensure_poolsize(a, idr, cnt); if (r != ARCHIVE_OK) return (r); __archive_rb_tree_init(&(idr->rbtree), rbt_ops); idr->wait_list.first = NULL; idr->wait_list.last = &(idr->wait_list.first); idr->pool_idx = 0; idr->num_size = num_size; idr->null_size = null_size; return (ARCHIVE_OK); } static void idr_register(struct idr *idr, struct isoent *isoent, int weight, int noff) { struct idrent *idrent, *n; idrent = &(idr->idrent_pool[idr->pool_idx++]); idrent->wnext = idrent->avail = NULL; idrent->isoent = isoent; idrent->weight = weight; idrent->noff = noff; idrent->rename_num = 0; if (!__archive_rb_tree_insert_node(&(idr->rbtree), &(idrent->rbnode))) { n = (struct idrent *)__archive_rb_tree_find_node( &(idr->rbtree), idrent->isoent); if (n != NULL) { /* this `idrent' needs to rename. */ idrent->avail = n; *idr->wait_list.last = idrent; idr->wait_list.last = &(idrent->wnext); } } } static void idr_extend_identifier(struct idrent *wnp, int numsize, int nullsize) { unsigned char *p; int wnp_ext_off; wnp_ext_off = wnp->isoent->ext_off; if (wnp->noff + numsize != wnp_ext_off) { p = (unsigned char *)wnp->isoent->identifier; /* Extend the filename; foo.c --> foo___.c */ memmove(p + wnp->noff + numsize, p + wnp_ext_off, wnp->isoent->ext_len + nullsize); wnp->isoent->ext_off = wnp_ext_off = wnp->noff + numsize; wnp->isoent->id_len = wnp_ext_off + wnp->isoent->ext_len; } } static void idr_resolve(struct idr *idr, void (*fsetnum)(unsigned char *p, int num)) { struct idrent *n; unsigned char *p; for (n = idr->wait_list.first; n != NULL; n = n->wnext) { idr_extend_identifier(n, idr->num_size, idr->null_size); p = (unsigned char *)n->isoent->identifier + n->noff; do { fsetnum(p, n->avail->rename_num++); } while (!__archive_rb_tree_insert_node( &(idr->rbtree), &(n->rbnode))); } } static void idr_set_num(unsigned char *p, int num) { static const char xdig[] = { '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z' }; num %= sizeof(xdig) * sizeof(xdig) * sizeof(xdig); p[0] = xdig[(num / (sizeof(xdig) * sizeof(xdig)))]; num %= sizeof(xdig) * sizeof(xdig); p[1] = xdig[ (num / sizeof(xdig))]; num %= sizeof(xdig); p[2] = xdig[num]; } static void idr_set_num_beutf16(unsigned char *p, int num) { static const uint16_t xdig[] = { 0x0030, 0x0031, 0x0032, 0x0033, 0x0034, 0x0035, 0x0036, 0x0037, 0x0038, 0x0039, 0x0041, 0x0042, 0x0043, 0x0044, 0x0045, 0x0046, 0x0047, 0x0048, 0x0049, 0x004A, 0x004B, 0x004C, 0x004D, 0x004E, 0x004F, 0x0050, 0x0051, 0x0052, 0x0053, 0x0054, 0x0055, 0x0056, 0x0057, 0x0058, 0x0059, 0x005A }; #define XDIG_CNT (sizeof(xdig)/sizeof(xdig[0])) num %= XDIG_CNT * XDIG_CNT * XDIG_CNT; archive_be16enc(p, xdig[(num / (XDIG_CNT * XDIG_CNT))]); num %= XDIG_CNT * XDIG_CNT; archive_be16enc(p+2, xdig[ (num / XDIG_CNT)]); num %= XDIG_CNT; archive_be16enc(p+4, xdig[num]); } /* * Generate ISO9660 Identifier. */ static int isoent_gen_iso9660_identifier(struct archive_write *a, struct isoent *isoent, struct idr *idr) { struct iso9660 *iso9660; struct isoent *np; char *p; int l, r; const char *char_map; char allow_ldots, allow_multidot, allow_period, allow_vernum; int fnmax, ffmax, dnmax; static const struct archive_rb_tree_ops rb_ops = { isoent_cmp_node_iso9660, isoent_cmp_key_iso9660 }; if (isoent->children.cnt == 0) return (0); iso9660 = a->format_data; char_map = idr->char_map; if (iso9660->opt.iso_level <= 3) { allow_ldots = 0; allow_multidot = 0; allow_period = 1; allow_vernum = iso9660->opt.allow_vernum; if (iso9660->opt.iso_level == 1) { fnmax = 8; ffmax = 12;/* fnmax + '.' + 3 */ dnmax = 8; } else { fnmax = 30; ffmax = 31; dnmax = 31; } } else { allow_ldots = allow_multidot = 1; allow_period = allow_vernum = 0; if (iso9660->opt.rr) /* * MDR : The maximum size of Directory Record(254). * DRL : A Directory Record Length(33). * CE : A size of SUSP CE System Use Entry(28). * MDR - DRL - CE = 254 - 33 - 28 = 193. */ fnmax = ffmax = dnmax = 193; else /* * XA : CD-ROM XA System Use Extension * Information(14). * MDR - DRL - XA = 254 - 33 -14 = 207. */ fnmax = ffmax = dnmax = 207; } r = idr_start(a, idr, isoent->children.cnt, ffmax, 3, 1, &rb_ops); if (r < 0) return (r); for (np = isoent->children.first; np != NULL; np = np->chnext) { char *dot, *xdot; int ext_off, noff, weight; l = np->file->basename.length; p = malloc(l+31+2+1); if (p == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } memcpy(p, np->file->basename.s, l); p[l] = '\0'; np->identifier = p; dot = xdot = NULL; if (!allow_ldots) { /* * If there is a '.' character at the first byte, * it has to be replaced by '_' character. */ if (*p == '.') *p++ = '_'; } for (;*p; p++) { if (*p & 0x80) { *p = '_'; continue; } if (char_map[(unsigned char)*p]) { /* if iso-level is '4', a character '.' is * allowed by char_map. */ if (*p == '.') { xdot = dot; dot = p; } continue; } if (*p >= 'a' && *p <= 'z') { *p -= 'a' - 'A'; continue; } if (*p == '.') { xdot = dot; dot = p; if (allow_multidot) continue; } *p = '_'; } p = np->identifier; weight = -1; if (dot == NULL) { int nammax; if (np->dir) nammax = dnmax; else nammax = fnmax; if (l > nammax) { p[nammax] = '\0'; weight = nammax; ext_off = nammax; } else ext_off = l; } else { *dot = '.'; ext_off = dot - p; if (iso9660->opt.iso_level == 1) { if (dot - p <= 8) { if (strlen(dot) > 4) { /* A length of a file extension * must be less than 4 */ dot[4] = '\0'; weight = 0; } } else { p[8] = dot[0]; p[9] = dot[1]; p[10] = dot[2]; p[11] = dot[3]; p[12] = '\0'; weight = 8; ext_off = 8; } } else if (np->dir) { if (l > dnmax) { p[dnmax] = '\0'; weight = dnmax; if (ext_off > dnmax) ext_off = dnmax; } } else if (l > ffmax) { int extlen = strlen(dot); int xdoff; if (xdot != NULL) xdoff = xdot - p; else xdoff = 0; if (extlen > 1 && xdoff < fnmax-1) { int off; if (extlen > ffmax) extlen = ffmax; off = ffmax - extlen; if (off == 0) { /* A dot('.') character * does't place to the first * byte of identifier. */ off ++; extlen --; } memmove(p+off, dot, extlen); p[ffmax] = '\0'; ext_off = off; weight = off; #ifdef COMPAT_MKISOFS } else if (xdoff >= fnmax-1) { /* Simulate a bug(?) of mkisofs. */ p[fnmax-1] = '\0'; ext_off = fnmax-1; weight = fnmax-1; #endif } else { p[fnmax] = '\0'; ext_off = fnmax; weight = fnmax; } } } /* Save an offset of a file name extension to sort files. */ np->ext_off = ext_off; np->ext_len = strlen(&p[ext_off]); np->id_len = l = ext_off + np->ext_len; /* Make an offset of the number which is used to be set * hexadecimal number to avoid duplicate identififier. */ if (iso9660->opt.iso_level == 1) { if (ext_off >= 5) noff = 5; else noff = ext_off; } else { if (l == ffmax) noff = ext_off - 3; else if (l == ffmax-1) noff = ext_off - 2; else if (l == ffmax-2) noff = ext_off - 1; else noff = ext_off; } /* Register entry to the identifier resolver. */ idr_register(idr, np, weight, noff); } /* Resolve duplicate identifier. */ idr_resolve(idr, idr_set_num); /* Add a period and a version number to identifiers. */ for (np = isoent->children.first; np != NULL; np = np->chnext) { if (!np->dir && np->rr_child == NULL) { p = np->identifier + np->ext_off + np->ext_len; if (np->ext_len == 0 && allow_period) { *p++ = '.'; np->ext_len = 1; } if (np->ext_len == 1 && !allow_period) { *--p = '\0'; np->ext_len = 0; } np->id_len = np->ext_off + np->ext_len; if (allow_vernum) { *p++ = ';'; *p++ = '1'; np->id_len += 2; } *p = '\0'; } else np->id_len = np->ext_off + np->ext_len; np->mb_len = np->id_len; } return (ARCHIVE_OK); } /* * Generate Joliet Identifier. */ static int isoent_gen_joliet_identifier(struct archive_write *a, struct isoent *isoent, struct idr *idr) { struct iso9660 *iso9660; struct isoent *np; unsigned char *p; size_t l; int r; int ffmax, parent_len; static const struct archive_rb_tree_ops rb_ops = { isoent_cmp_node_joliet, isoent_cmp_key_joliet }; if (isoent->children.cnt == 0) return (0); iso9660 = a->format_data; if (iso9660->opt.joliet == OPT_JOLIET_LONGNAME) ffmax = 206; else ffmax = 128; r = idr_start(a, idr, isoent->children.cnt, ffmax, 6, 2, &rb_ops); if (r < 0) return (r); parent_len = 1; for (np = isoent; np->parent != np; np = np->parent) parent_len += np->mb_len + 1; for (np = isoent->children.first; np != NULL; np = np->chnext) { unsigned char *dot; int ext_off, noff, weight; size_t lt; if ((int)(l = np->file->basename_utf16.length) > ffmax) l = ffmax; p = malloc((l+1)*2); if (p == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } memcpy(p, np->file->basename_utf16.s, l); p[l] = 0; p[l+1] = 0; np->identifier = (char *)p; lt = l; dot = p + l; weight = 0; while (lt > 0) { if (!joliet_allowed_char(p[0], p[1])) archive_be16enc(p, 0x005F); /* '_' */ else if (p[0] == 0 && p[1] == 0x2E) /* '.' */ dot = p; p += 2; lt -= 2; } ext_off = dot - (unsigned char *)np->identifier; np->ext_off = ext_off; np->ext_len = l - ext_off; np->id_len = l; /* * Get a length of MBS of a full-pathname. */ if ((int)np->file->basename_utf16.length > ffmax) { archive_strncpy_in_locale(&iso9660->mbs, (const char *)np->identifier, l, iso9660->sconv_from_utf16be); np->mb_len = iso9660->mbs.length; if (np->mb_len != (int)np->file->basename.length) weight = np->mb_len; } else np->mb_len = np->file->basename.length; /* If a length of full-pathname is longer than 240 bytes, * it violates Joliet extensions regulation. */ if (parent_len + np->mb_len > 240) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "The regulation of Joliet extensions;" " A length of a full-pathname of `%s' is " "longer than 240 bytes, (p=%d, b=%d)", archive_entry_pathname(np->file->entry), (int)parent_len, (int)np->mb_len); return (ARCHIVE_FATAL); } /* Make an offset of the number which is used to be set * hexadecimal number to avoid duplicate identifier. */ if ((int)l == ffmax) noff = ext_off - 6; else if ((int)l == ffmax-2) noff = ext_off - 4; else if ((int)l == ffmax-4) noff = ext_off - 2; else noff = ext_off; /* Register entry to the identifier resolver. */ idr_register(idr, np, weight, noff); } /* Resolve duplicate identifier with Joliet Volume. */ idr_resolve(idr, idr_set_num_beutf16); return (ARCHIVE_OK); } /* * This comparing rule is according to ISO9660 Standard 9.3 */ static int isoent_cmp_iso9660_identifier(const struct isoent *p1, const struct isoent *p2) { const char *s1, *s2; int cmp; int l; s1 = p1->identifier; s2 = p2->identifier; /* Compare File Name */ l = p1->ext_off; if (l > p2->ext_off) l = p2->ext_off; cmp = memcmp(s1, s2, l); if (cmp != 0) return (cmp); if (p1->ext_off < p2->ext_off) { s2 += l; l = p2->ext_off - p1->ext_off; while (l--) if (0x20 != *s2++) return (0x20 - *(const unsigned char *)(s2 - 1)); } else if (p1->ext_off > p2->ext_off) { s1 += l; l = p1->ext_off - p2->ext_off; while (l--) if (0x20 != *s1++) return (*(const unsigned char *)(s1 - 1) - 0x20); } /* Compare File Name Extension */ if (p1->ext_len == 0 && p2->ext_len == 0) return (0); if (p1->ext_len == 1 && p2->ext_len == 1) return (0); if (p1->ext_len <= 1) return (-1); if (p2->ext_len <= 1) return (1); l = p1->ext_len; if (l > p2->ext_len) l = p2->ext_len; s1 = p1->identifier + p1->ext_off; s2 = p2->identifier + p2->ext_off; if (l > 1) { cmp = memcmp(s1, s2, l); if (cmp != 0) return (cmp); } if (p1->ext_len < p2->ext_len) { s2 += l; l = p2->ext_len - p1->ext_len; while (l--) if (0x20 != *s2++) return (0x20 - *(const unsigned char *)(s2 - 1)); } else if (p1->ext_len < p2->ext_len) { s1 += l; l = p1->ext_len - p2->ext_len; while (l--) if (0x20 != *s1++) return (*(const unsigned char *)(s1 - 1) - 0x20); } /* Compare File Version Number */ /* No operation. The File Version Number is always one. */ return (cmp); } static int isoent_cmp_node_iso9660(const struct archive_rb_node *n1, const struct archive_rb_node *n2) { const struct idrent *e1 = (const struct idrent *)n1; const struct idrent *e2 = (const struct idrent *)n2; return (isoent_cmp_iso9660_identifier(e2->isoent, e1->isoent)); } static int isoent_cmp_key_iso9660(const struct archive_rb_node *node, const void *key) { const struct isoent *isoent = (const struct isoent *)key; const struct idrent *idrent = (const struct idrent *)node; return (isoent_cmp_iso9660_identifier(isoent, idrent->isoent)); } static int isoent_cmp_joliet_identifier(const struct isoent *p1, const struct isoent *p2) { const unsigned char *s1, *s2; int cmp; int l; s1 = (const unsigned char *)p1->identifier; s2 = (const unsigned char *)p2->identifier; /* Compare File Name */ l = p1->ext_off; if (l > p2->ext_off) l = p2->ext_off; cmp = memcmp(s1, s2, l); if (cmp != 0) return (cmp); if (p1->ext_off < p2->ext_off) { s2 += l; l = p2->ext_off - p1->ext_off; while (l--) if (0 != *s2++) return (- *(const unsigned char *)(s2 - 1)); } else if (p1->ext_off > p2->ext_off) { s1 += l; l = p1->ext_off - p2->ext_off; while (l--) if (0 != *s1++) return (*(const unsigned char *)(s1 - 1)); } /* Compare File Name Extension */ if (p1->ext_len == 0 && p2->ext_len == 0) return (0); if (p1->ext_len == 2 && p2->ext_len == 2) return (0); if (p1->ext_len <= 2) return (-1); if (p2->ext_len <= 2) return (1); l = p1->ext_len; if (l > p2->ext_len) l = p2->ext_len; s1 = (unsigned char *)(p1->identifier + p1->ext_off); s2 = (unsigned char *)(p2->identifier + p2->ext_off); if (l > 1) { cmp = memcmp(s1, s2, l); if (cmp != 0) return (cmp); } if (p1->ext_len < p2->ext_len) { s2 += l; l = p2->ext_len - p1->ext_len; while (l--) if (0 != *s2++) return (- *(const unsigned char *)(s2 - 1)); } else if (p1->ext_len < p2->ext_len) { s1 += l; l = p1->ext_len - p2->ext_len; while (l--) if (0 != *s1++) return (*(const unsigned char *)(s1 - 1)); } /* Compare File Version Number */ /* No operation. The File Version Number is always one. */ return (cmp); } static int isoent_cmp_node_joliet(const struct archive_rb_node *n1, const struct archive_rb_node *n2) { const struct idrent *e1 = (const struct idrent *)n1; const struct idrent *e2 = (const struct idrent *)n2; return (isoent_cmp_joliet_identifier(e2->isoent, e1->isoent)); } static int isoent_cmp_key_joliet(const struct archive_rb_node *node, const void *key) { const struct isoent *isoent = (const struct isoent *)key; const struct idrent *idrent = (const struct idrent *)node; return (isoent_cmp_joliet_identifier(isoent, idrent->isoent)); } static int isoent_make_sorted_files(struct archive_write *a, struct isoent *isoent, struct idr *idr) { struct archive_rb_node *rn; struct isoent **children; children = malloc(isoent->children.cnt * sizeof(struct isoent *)); if (children == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } isoent->children_sorted = children; ARCHIVE_RB_TREE_FOREACH(rn, &(idr->rbtree)) { struct idrent *idrent = (struct idrent *)rn; *children ++ = idrent->isoent; } return (ARCHIVE_OK); } /* * - Generate ISO9660 and Joliet identifiers from basenames. * - Sort files by each directory. */ static int isoent_traverse_tree(struct archive_write *a, struct vdd* vdd) { struct iso9660 *iso9660 = a->format_data; struct isoent *np; struct idr idr; int depth; int r; - int (*genid)(struct archive_write *a, struct isoent *isoent, - struct idr *idr); + int (*genid)(struct archive_write *, struct isoent *, struct idr *); idr_init(iso9660, vdd, &idr); np = vdd->rootent; depth = 0; if (vdd->vdd_type == VDD_JOLIET) genid = isoent_gen_joliet_identifier; else genid = isoent_gen_iso9660_identifier; do { if (np->virtual && !archive_entry_mtime_is_set(np->file->entry)) { /* Set properly times to virtual directory */ archive_entry_set_mtime(np->file->entry, iso9660->birth_time, 0); archive_entry_set_atime(np->file->entry, iso9660->birth_time, 0); archive_entry_set_ctime(np->file->entry, iso9660->birth_time, 0); } if (np->children.first != NULL) { if (vdd->vdd_type != VDD_JOLIET && !iso9660->opt.rr && depth + 1 >= vdd->max_depth) { if (np->children.cnt > 0) iso9660->directories_too_deep = np; } else { /* Generate Identifier */ r = genid(a, np, &idr); if (r < 0) goto exit_traverse_tree; r = isoent_make_sorted_files(a, np, &idr); if (r < 0) goto exit_traverse_tree; if (np->subdirs.first != NULL && depth + 1 < vdd->max_depth) { /* Enter to sub directories. */ np = np->subdirs.first; depth++; continue; } } } while (np != np->parent) { if (np->drnext == NULL) { /* Return to the parent directory. */ np = np->parent; depth--; } else { np = np->drnext; break; } } } while (np != np->parent); r = ARCHIVE_OK; exit_traverse_tree: idr_cleanup(&idr); return (r); } /* * Collect directory entries into path_table by a directory depth. */ static int isoent_collect_dirs(struct vdd *vdd, struct isoent *rootent, int depth) { struct isoent *np; if (rootent == NULL) rootent = vdd->rootent; np = rootent; do { /* Register current directory to pathtable. */ path_table_add_entry(&(vdd->pathtbl[depth]), np); if (np->subdirs.first != NULL && depth + 1 < vdd->max_depth) { /* Enter to sub directories. */ np = np->subdirs.first; depth++; continue; } while (np != rootent) { if (np->drnext == NULL) { /* Return to the parent directory. */ np = np->parent; depth--; } else { np = np->drnext; break; } } } while (np != rootent); return (ARCHIVE_OK); } /* * The entry whose number of levels in a directory hierarchy is * large than eight relocate to rr_move directory. */ static int isoent_rr_move_dir(struct archive_write *a, struct isoent **rr_moved, struct isoent *isoent, struct isoent **newent) { struct iso9660 *iso9660 = a->format_data; struct isoent *rrmoved, *mvent, *np; if ((rrmoved = *rr_moved) == NULL) { struct isoent *rootent = iso9660->primary.rootent; /* There isn't rr_move entry. * Create rr_move entry and insert it into the root entry. */ rrmoved = isoent_create_virtual_dir(a, iso9660, "rr_moved"); if (rrmoved == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } /* Add "rr_moved" entry to the root entry. */ isoent_add_child_head(rootent, rrmoved); archive_entry_set_nlink(rootent->file->entry, archive_entry_nlink(rootent->file->entry) + 1); /* Register "rr_moved" entry to second level pathtable. */ path_table_add_entry(&(iso9660->primary.pathtbl[1]), rrmoved); /* Save rr_moved. */ *rr_moved = rrmoved; } /* * Make a clone of isoent which is going to be relocated * to rr_moved. */ mvent = isoent_clone(isoent); if (mvent == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } /* linking.. and use for creating "CL", "PL" and "RE" */ mvent->rr_parent = isoent->parent; isoent->rr_child = mvent; /* * Move subdirectories from the isoent to mvent */ if (isoent->children.first != NULL) { *mvent->children.last = isoent->children.first; mvent->children.last = isoent->children.last; } for (np = mvent->children.first; np != NULL; np = np->chnext) np->parent = mvent; mvent->children.cnt = isoent->children.cnt; isoent->children.cnt = 0; isoent->children.first = NULL; isoent->children.last = &isoent->children.first; if (isoent->subdirs.first != NULL) { *mvent->subdirs.last = isoent->subdirs.first; mvent->subdirs.last = isoent->subdirs.last; } mvent->subdirs.cnt = isoent->subdirs.cnt; isoent->subdirs.cnt = 0; isoent->subdirs.first = NULL; isoent->subdirs.last = &isoent->subdirs.first; /* * The mvent becomes a child of the rr_moved entry. */ isoent_add_child_tail(rrmoved, mvent); archive_entry_set_nlink(rrmoved->file->entry, archive_entry_nlink(rrmoved->file->entry) + 1); /* * This entry which relocated to the rr_moved directory * has to set the flag as a file. * See also RRIP 4.1.5.1 Description of the "CL" System Use Entry. */ isoent->dir = 0; *newent = mvent; return (ARCHIVE_OK); } static int isoent_rr_move(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; struct path_table *pt; struct isoent *rootent, *rr_moved; struct isoent *np, *last; int r; pt = &(iso9660->primary.pathtbl[MAX_DEPTH-1]); /* Theare aren't level 8 directories reaching a deepr level. */ if (pt->cnt == 0) return (ARCHIVE_OK); rootent = iso9660->primary.rootent; /* If "rr_moved" directory is already existing, * we have to use it. */ rr_moved = isoent_find_child(rootent, "rr_moved"); if (rr_moved != NULL && rr_moved != rootent->children.first) { /* * It's necessary that rr_move is the first entry * of the root. */ /* Remove "rr_moved" entry from children chain. */ isoent_remove_child(rootent, rr_moved); /* Add "rr_moved" entry into the head of children chain. */ isoent_add_child_head(rootent, rr_moved); } /* * Check level 8 path_table. * If find out sub directory entries, that entries move to rr_move. */ np = pt->first; while (np != NULL) { last = path_table_last_entry(pt); for (; np != NULL; np = np->ptnext) { struct isoent *mvent; struct isoent *newent; if (!np->dir) continue; for (mvent = np->subdirs.first; mvent != NULL; mvent = mvent->drnext) { r = isoent_rr_move_dir(a, &rr_moved, mvent, &newent); if (r < 0) return (r); isoent_collect_dirs(&(iso9660->primary), newent, 2); } } /* If new entries are added to level 8 path_talbe, * its sub directory entries move to rr_move too. */ np = last->ptnext; } return (ARCHIVE_OK); } /* * This comparing rule is according to ISO9660 Standard 6.9.1 */ static int _compare_path_table(const void *v1, const void *v2) { const struct isoent *p1, *p2; const char *s1, *s2; int cmp, l; p1 = *((const struct isoent **)(uintptr_t)v1); p2 = *((const struct isoent **)(uintptr_t)v2); /* Compare parent directory number */ cmp = p1->parent->dir_number - p2->parent->dir_number; if (cmp != 0) return (cmp); /* Compare indetifier */ s1 = p1->identifier; s2 = p2->identifier; l = p1->ext_off; if (l > p2->ext_off) l = p2->ext_off; cmp = strncmp(s1, s2, l); if (cmp != 0) return (cmp); if (p1->ext_off < p2->ext_off) { s2 += l; l = p2->ext_off - p1->ext_off; while (l--) if (0x20 != *s2++) return (0x20 - *(const unsigned char *)(s2 - 1)); } else if (p1->ext_off > p2->ext_off) { s1 += l; l = p1->ext_off - p2->ext_off; while (l--) if (0x20 != *s1++) return (*(const unsigned char *)(s1 - 1) - 0x20); } return (0); } static int _compare_path_table_joliet(const void *v1, const void *v2) { const struct isoent *p1, *p2; const unsigned char *s1, *s2; int cmp, l; p1 = *((const struct isoent **)(uintptr_t)v1); p2 = *((const struct isoent **)(uintptr_t)v2); /* Compare parent directory number */ cmp = p1->parent->dir_number - p2->parent->dir_number; if (cmp != 0) return (cmp); /* Compare indetifier */ s1 = (const unsigned char *)p1->identifier; s2 = (const unsigned char *)p2->identifier; l = p1->ext_off; if (l > p2->ext_off) l = p2->ext_off; cmp = memcmp(s1, s2, l); if (cmp != 0) return (cmp); if (p1->ext_off < p2->ext_off) { s2 += l; l = p2->ext_off - p1->ext_off; while (l--) if (0 != *s2++) return (- *(const unsigned char *)(s2 - 1)); } else if (p1->ext_off > p2->ext_off) { s1 += l; l = p1->ext_off - p2->ext_off; while (l--) if (0 != *s1++) return (*(const unsigned char *)(s1 - 1)); } return (0); } static inline void path_table_add_entry(struct path_table *pathtbl, struct isoent *ent) { ent->ptnext = NULL; *pathtbl->last = ent; pathtbl->last = &(ent->ptnext); pathtbl->cnt ++; } static inline struct isoent * path_table_last_entry(struct path_table *pathtbl) { if (pathtbl->first == NULL) return (NULL); return (((struct isoent *)(void *) ((char *)(pathtbl->last) - offsetof(struct isoent, ptnext)))); } /* * Sort directory entries in path_table * and assign directory number to each entries. */ static int isoent_make_path_table_2(struct archive_write *a, struct vdd *vdd, int depth, int *dir_number) { struct isoent *np; struct isoent **enttbl; struct path_table *pt; int i; pt = &vdd->pathtbl[depth]; if (pt->cnt == 0) { pt->sorted = NULL; return (ARCHIVE_OK); } enttbl = malloc(pt->cnt * sizeof(struct isoent *)); if (enttbl == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } pt->sorted = enttbl; for (np = pt->first; np != NULL; np = np->ptnext) *enttbl ++ = np; enttbl = pt->sorted; switch (vdd->vdd_type) { case VDD_PRIMARY: case VDD_ENHANCED: qsort(enttbl, pt->cnt, sizeof(struct isoent *), _compare_path_table); break; case VDD_JOLIET: qsort(enttbl, pt->cnt, sizeof(struct isoent *), _compare_path_table_joliet); break; } for (i = 0; i < pt->cnt; i++) enttbl[i]->dir_number = (*dir_number)++; return (ARCHIVE_OK); } static int isoent_alloc_path_table(struct archive_write *a, struct vdd *vdd, int max_depth) { int i; vdd->max_depth = max_depth; vdd->pathtbl = malloc(sizeof(*vdd->pathtbl) * vdd->max_depth); if (vdd->pathtbl == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } for (i = 0; i < vdd->max_depth; i++) { vdd->pathtbl[i].first = NULL; vdd->pathtbl[i].last = &(vdd->pathtbl[i].first); vdd->pathtbl[i].sorted = NULL; vdd->pathtbl[i].cnt = 0; } return (ARCHIVE_OK); } /* * Make Path Tables */ static int isoent_make_path_table(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; int depth, r; int dir_number; /* * Init Path Table. */ if (iso9660->dircnt_max >= MAX_DEPTH && (!iso9660->opt.limit_depth || iso9660->opt.iso_level == 4)) r = isoent_alloc_path_table(a, &(iso9660->primary), iso9660->dircnt_max + 1); else /* The number of levels in the hierarchy cannot exceed * eight. */ r = isoent_alloc_path_table(a, &(iso9660->primary), MAX_DEPTH); if (r < 0) return (r); if (iso9660->opt.joliet) { r = isoent_alloc_path_table(a, &(iso9660->joliet), iso9660->dircnt_max + 1); if (r < 0) return (r); } /* Step 0. * - Collect directories for primary and joliet. */ isoent_collect_dirs(&(iso9660->primary), NULL, 0); if (iso9660->opt.joliet) isoent_collect_dirs(&(iso9660->joliet), NULL, 0); /* * Rockridge; move deeper depth directories to rr_moved. */ if (iso9660->opt.rr) { r = isoent_rr_move(a); if (r < 0) return (r); } /* Update nlink. */ isofile_connect_hardlink_files(iso9660); /* Step 1. * - Renew a value of the depth of that directories. * - Resolve hardlinks. * - Convert pathnames to ISO9660 name or UCS2(joliet). * - Sort files by each directory. */ r = isoent_traverse_tree(a, &(iso9660->primary)); if (r < 0) return (r); if (iso9660->opt.joliet) { r = isoent_traverse_tree(a, &(iso9660->joliet)); if (r < 0) return (r); } /* Step 2. * - Sort directories. * - Assign all directory number. */ dir_number = 1; for (depth = 0; depth < iso9660->primary.max_depth; depth++) { r = isoent_make_path_table_2(a, &(iso9660->primary), depth, &dir_number); if (r < 0) return (r); } if (iso9660->opt.joliet) { dir_number = 1; for (depth = 0; depth < iso9660->joliet.max_depth; depth++) { r = isoent_make_path_table_2(a, &(iso9660->joliet), depth, &dir_number); if (r < 0) return (r); } } if (iso9660->opt.limit_dirs && dir_number > 0xffff) { /* * Maximum number of directories is 65535(0xffff) * doe to size(16bit) of Parent Directory Number of * the Path Table. * See also ISO9660 Standard 9.4. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Too many directories(%d) over 65535.", dir_number); return (ARCHIVE_FATAL); } /* Get the size of the Path Table. */ calculate_path_table_size(&(iso9660->primary)); if (iso9660->opt.joliet) calculate_path_table_size(&(iso9660->joliet)); return (ARCHIVE_OK); } static int isoent_find_out_boot_file(struct archive_write *a, struct isoent *rootent) { struct iso9660 *iso9660 = a->format_data; /* Find a isoent of the boot file. */ iso9660->el_torito.boot = isoent_find_entry(rootent, iso9660->el_torito.boot_filename.s); if (iso9660->el_torito.boot == NULL) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't find the boot image file ``%s''", iso9660->el_torito.boot_filename.s); return (ARCHIVE_FATAL); } iso9660->el_torito.boot->file->boot = BOOT_IMAGE; return (ARCHIVE_OK); } static int isoent_create_boot_catalog(struct archive_write *a, struct isoent *rootent) { struct iso9660 *iso9660 = a->format_data; struct isofile *file; struct isoent *isoent; struct archive_entry *entry; (void)rootent; /* UNUSED */ /* * Create the entry which is the "boot.catalog" file. */ file = isofile_new(a, NULL); if (file == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } archive_entry_set_pathname(file->entry, iso9660->el_torito.catalog_filename.s); archive_entry_set_size(file->entry, LOGICAL_BLOCK_SIZE); archive_entry_set_mtime(file->entry, iso9660->birth_time, 0); archive_entry_set_atime(file->entry, iso9660->birth_time, 0); archive_entry_set_ctime(file->entry, iso9660->birth_time, 0); archive_entry_set_uid(file->entry, getuid()); archive_entry_set_gid(file->entry, getgid()); archive_entry_set_mode(file->entry, AE_IFREG | 0444); archive_entry_set_nlink(file->entry, 1); if (isofile_gen_utility_names(a, file) < ARCHIVE_WARN) { isofile_free(file); return (ARCHIVE_FATAL); } file->boot = BOOT_CATALOG; file->content.size = LOGICAL_BLOCK_SIZE; isofile_add_entry(iso9660, file); isoent = isoent_new(file); if (isoent == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } isoent->virtual = 1; /* Add the "boot.catalog" entry into tree */ if (isoent_tree(a, &isoent) != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->el_torito.catalog = isoent; /* * Get a boot medai type. */ switch (iso9660->opt.boot_type) { default: case OPT_BOOT_TYPE_AUTO: /* Try detecting a media type of the boot image. */ entry = iso9660->el_torito.boot->file->entry; if (archive_entry_size(entry) == FD_1_2M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_1_2M_DISKETTE; else if (archive_entry_size(entry) == FD_1_44M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_1_44M_DISKETTE; else if (archive_entry_size(entry) == FD_2_88M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_2_88M_DISKETTE; else /* We cannot decide whether the boot image is * hard-disk. */ iso9660->el_torito.media_type = BOOT_MEDIA_NO_EMULATION; break; case OPT_BOOT_TYPE_NO_EMU: iso9660->el_torito.media_type = BOOT_MEDIA_NO_EMULATION; break; case OPT_BOOT_TYPE_HARD_DISK: iso9660->el_torito.media_type = BOOT_MEDIA_HARD_DISK; break; case OPT_BOOT_TYPE_FD: entry = iso9660->el_torito.boot->file->entry; if (archive_entry_size(entry) <= FD_1_2M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_1_2M_DISKETTE; else if (archive_entry_size(entry) <= FD_1_44M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_1_44M_DISKETTE; else if (archive_entry_size(entry) <= FD_2_88M_SIZE) iso9660->el_torito.media_type = BOOT_MEDIA_2_88M_DISKETTE; else { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Boot image file(``%s'') size is too big " "for fd type.", iso9660->el_torito.boot_filename.s); return (ARCHIVE_FATAL); } break; } /* * Get a system type. * TODO: `El Torito' specification says "A copy of byte 5 from the * Partition Table found in the boot image". */ iso9660->el_torito.system_type = 0; /* * Get an ID. */ if (iso9660->opt.publisher) archive_string_copy(&(iso9660->el_torito.id), &(iso9660->publisher_identifier)); return (ARCHIVE_OK); } /* * If a media type is floppy, return its image size. * otherwise return 0. */ static size_t fd_boot_image_size(int media_type) { switch (media_type) { case BOOT_MEDIA_1_2M_DISKETTE: return (FD_1_2M_SIZE); case BOOT_MEDIA_1_44M_DISKETTE: return (FD_1_44M_SIZE); case BOOT_MEDIA_2_88M_DISKETTE: return (FD_2_88M_SIZE); default: return (0); } } /* * Make a boot catalog image data. */ static int make_boot_catalog(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; unsigned char *block; unsigned char *p; uint16_t sum, *wp; block = wb_buffptr(a); memset(block, 0, LOGICAL_BLOCK_SIZE); p = block; /* * Validation Entry */ /* Header ID */ p[0] = 1; /* Platform ID */ p[1] = iso9660->el_torito.platform_id; /* Reserved */ p[2] = p[3] = 0; /* ID */ if (archive_strlen(&(iso9660->el_torito.id)) > 0) strncpy((char *)p+4, iso9660->el_torito.id.s, 23); p[27] = 0; /* Checksum */ p[28] = p[29] = 0; /* Key */ p[30] = 0x55; p[31] = 0xAA; sum = 0; wp = (uint16_t *)block; while (wp < (uint16_t *)&block[32]) sum += archive_le16dec(wp++); set_num_721(&block[28], (~sum) + 1); /* * Initial/Default Entry */ p = &block[32]; /* Boot Indicator */ p[0] = 0x88; /* Boot media type */ p[1] = iso9660->el_torito.media_type; /* Load Segment */ if (iso9660->el_torito.media_type == BOOT_MEDIA_NO_EMULATION) set_num_721(&p[2], iso9660->el_torito.boot_load_seg); else set_num_721(&p[2], 0); /* System Type */ p[4] = iso9660->el_torito.system_type; /* Unused */ p[5] = 0; /* Sector Count */ if (iso9660->el_torito.media_type == BOOT_MEDIA_NO_EMULATION) set_num_721(&p[6], iso9660->el_torito.boot_load_size); else set_num_721(&p[6], 1); /* Load RBA */ set_num_731(&p[8], iso9660->el_torito.boot->file->content.location); /* Unused */ memset(&p[12], 0, 20); return (wb_consume(a, LOGICAL_BLOCK_SIZE)); } static int setup_boot_information(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; struct isoent *np; int64_t size; uint32_t sum; unsigned char buff[4096]; np = iso9660->el_torito.boot; lseek(iso9660->temp_fd, np->file->content.offset_of_temp + 64, SEEK_SET); size = archive_entry_size(np->file->entry) - 64; if (size <= 0) { archive_set_error(&a->archive, errno, "Boot file(%jd) is too small", (intmax_t)size + 64); return (ARCHIVE_FATAL); } sum = 0; while (size > 0) { size_t rsize; ssize_t i, rs; - if (size > sizeof(buff)) + if (size > (int64_t)sizeof(buff)) rsize = sizeof(buff); else rsize = (size_t)size; rs = read(iso9660->temp_fd, buff, rsize); if (rs <= 0) { archive_set_error(&a->archive, errno, "Can't read temporary file(%jd)", (intmax_t)rs); return (ARCHIVE_FATAL); } for (i = 0; i < rs; i += 4) sum += archive_le32dec(buff + i); size -= rs; } /* Set the location of Primary Volume Descriptor. */ set_num_731(buff, SYSTEM_AREA_BLOCK); /* Set the location of the boot file. */ set_num_731(buff+4, np->file->content.location); /* Set the size of the boot file. */ size = fd_boot_image_size(iso9660->el_torito.media_type); if (size == 0) size = archive_entry_size(np->file->entry); set_num_731(buff+8, (uint32_t)size); /* Set the sum of the boot file. */ set_num_731(buff+12, sum); /* Clear reserved bytes. */ memset(buff+16, 0, 40); /* Overwrite the boot file. */ lseek(iso9660->temp_fd, np->file->content.offset_of_temp + 8, SEEK_SET); return (write_to_temp(a, buff, 56)); } #ifdef HAVE_ZLIB_H static int zisofs_init_zstream(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; int r; iso9660->zisofs.stream.next_in = NULL; iso9660->zisofs.stream.avail_in = 0; iso9660->zisofs.stream.total_in = 0; iso9660->zisofs.stream.total_out = 0; if (iso9660->zisofs.stream_valid) r = deflateReset(&(iso9660->zisofs.stream)); else { r = deflateInit(&(iso9660->zisofs.stream), iso9660->zisofs.compression_level); iso9660->zisofs.stream_valid = 1; } switch (r) { case Z_OK: break; default: case Z_STREAM_ERROR: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Internal error initializing " "compression library: invalid setup parameter"); return (ARCHIVE_FATAL); case Z_MEM_ERROR: archive_set_error(&a->archive, ENOMEM, "Internal error initializing " "compression library"); return (ARCHIVE_FATAL); case Z_VERSION_ERROR: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Internal error initializing " "compression library: invalid library version"); return (ARCHIVE_FATAL); } return (ARCHIVE_OK); } #endif /* HAVE_ZLIB_H */ static int zisofs_init(struct archive_write *a, struct isofile *file) { struct iso9660 *iso9660 = a->format_data; #ifdef HAVE_ZLIB_H uint64_t tsize; size_t ceil, bpsize; int r; #endif iso9660->zisofs.detect_magic = 0; iso9660->zisofs.making = 0; if (!iso9660->opt.rr || !iso9660->opt.zisofs) return (ARCHIVE_OK); if (archive_entry_size(file->entry) >= 24 && archive_entry_size(file->entry) < MULTI_EXTENT_SIZE) { /* Acceptable file size for zisofs. */ iso9660->zisofs.detect_magic = 1; iso9660->zisofs.magic_cnt = 0; } if (!iso9660->zisofs.detect_magic) return (ARCHIVE_OK); #ifdef HAVE_ZLIB_H /* The number of Logical Blocks which uncompressed data * will use in iso-image file is the same as the number of * Logical Blocks which zisofs(compressed) data will use * in ISO-image file. It won't reduce iso-image file size. */ if (archive_entry_size(file->entry) <= LOGICAL_BLOCK_SIZE) return (ARCHIVE_OK); /* Initialize compression library */ r = zisofs_init_zstream(a); if (r != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Mark file->zisofs to create RRIP 'ZF' Use Entry. */ file->zisofs.header_size = ZF_HEADER_SIZE >> 2; file->zisofs.log2_bs = ZF_LOG2_BS; file->zisofs.uncompressed_size = archive_entry_size(file->entry); /* Calculate a size of Block Pointers of zisofs. */ ceil = (file->zisofs.uncompressed_size + ZF_BLOCK_SIZE -1) >> file->zisofs.log2_bs; iso9660->zisofs.block_pointers_cnt = ceil + 1; iso9660->zisofs.block_pointers_idx = 0; /* Ensure a buffer size used for Block Pointers */ bpsize = iso9660->zisofs.block_pointers_cnt * sizeof(iso9660->zisofs.block_pointers[0]); if (iso9660->zisofs.block_pointers_allocated < bpsize) { free(iso9660->zisofs.block_pointers); iso9660->zisofs.block_pointers = malloc(bpsize); if (iso9660->zisofs.block_pointers == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate data"); return (ARCHIVE_FATAL); } iso9660->zisofs.block_pointers_allocated = bpsize; } /* * Skip zisofs header and Block Pointers, which we will write * after all compressed data of a file written to the temporary * file. */ tsize = ZF_HEADER_SIZE + bpsize; if (write_null(a, tsize) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* * Initialize some variables to make zisofs. */ archive_le32enc(&(iso9660->zisofs.block_pointers[0]), tsize); iso9660->zisofs.remaining = file->zisofs.uncompressed_size; iso9660->zisofs.making = 1; iso9660->zisofs.allzero = 1; iso9660->zisofs.block_offset = tsize; iso9660->zisofs.total_size = tsize; iso9660->cur_file->cur_content->size = tsize; #endif return (ARCHIVE_OK); } static void zisofs_detect_magic(struct archive_write *a, const void *buff, size_t s) { struct iso9660 *iso9660 = a->format_data; struct isofile *file = iso9660->cur_file; const unsigned char *p, *endp; const unsigned char *magic_buff; uint32_t uncompressed_size; unsigned char header_size; unsigned char log2_bs; size_t ceil, doff; uint32_t bst, bed; int magic_max; int64_t entry_size; entry_size = archive_entry_size(file->entry); - if (sizeof(iso9660->zisofs.magic_buffer) > entry_size) + if ((int64_t)sizeof(iso9660->zisofs.magic_buffer) > entry_size) magic_max = entry_size; else magic_max = sizeof(iso9660->zisofs.magic_buffer); if (iso9660->zisofs.magic_cnt == 0 && s >= (size_t)magic_max) /* It's unnecessary we copy buffer. */ magic_buff = buff; else { if (iso9660->zisofs.magic_cnt < magic_max) { size_t l; l = sizeof(iso9660->zisofs.magic_buffer) - iso9660->zisofs.magic_cnt; if (l > s) l = s; memcpy(iso9660->zisofs.magic_buffer + iso9660->zisofs.magic_cnt, buff, l); iso9660->zisofs.magic_cnt += l; if (iso9660->zisofs.magic_cnt < magic_max) return; } magic_buff = iso9660->zisofs.magic_buffer; } iso9660->zisofs.detect_magic = 0; p = magic_buff; /* Check the magic code of zisofs. */ if (memcmp(p, zisofs_magic, sizeof(zisofs_magic)) != 0) /* This is not zisofs file which made by mkzftree. */ return; p += sizeof(zisofs_magic); /* Read a zisofs header. */ uncompressed_size = archive_le32dec(p); header_size = p[4]; log2_bs = p[5]; if (uncompressed_size < 24 || header_size != 4 || log2_bs > 30 || log2_bs < 7) return;/* Invalid or not supported header. */ /* Calculate a size of Block Pointers of zisofs. */ ceil = (uncompressed_size + (ARCHIVE_LITERAL_LL(1) << log2_bs) -1) >> log2_bs; doff = (ceil + 1) * 4 + 16; - if (entry_size < doff) + if (entry_size < (int64_t)doff) return;/* Invalid data. */ /* Check every Block Pointer has valid value. */ p = magic_buff + 16; endp = magic_buff + magic_max; while (ceil && p + 8 <= endp) { bst = archive_le32dec(p); if (bst != doff) return;/* Invalid data. */ p += 4; bed = archive_le32dec(p); if (bed < bst || bed > entry_size) return;/* Invalid data. */ doff += bed - bst; ceil--; } file->zisofs.uncompressed_size = uncompressed_size; file->zisofs.header_size = header_size; file->zisofs.log2_bs = log2_bs; /* Disable making a zisofs image. */ iso9660->zisofs.making = 0; } #ifdef HAVE_ZLIB_H /* * Compress data and write it to a temporary file. */ static int zisofs_write_to_temp(struct archive_write *a, const void *buff, size_t s) { struct iso9660 *iso9660 = a->format_data; struct isofile *file = iso9660->cur_file; const unsigned char *b; z_stream *zstrm; size_t avail, csize; int flush, r; zstrm = &(iso9660->zisofs.stream); zstrm->next_out = wb_buffptr(a); zstrm->avail_out = wb_remaining(a); b = (const unsigned char *)buff; do { avail = ZF_BLOCK_SIZE - zstrm->total_in; if (s < avail) { avail = s; flush = Z_NO_FLUSH; } else flush = Z_FINISH; iso9660->zisofs.remaining -= avail; if (iso9660->zisofs.remaining <= 0) flush = Z_FINISH; zstrm->next_in = (Bytef *)(uintptr_t)(const void *)b; zstrm->avail_in = avail; /* * Check if current data block are all zero. */ if (iso9660->zisofs.allzero) { const unsigned char *nonzero = b; const unsigned char *nonzeroend = b + avail; while (nonzero < nonzeroend) if (*nonzero++) { iso9660->zisofs.allzero = 0; break; } } b += avail; s -= avail; /* * If current data block are all zero, we do not use * compressed data. */ if (flush == Z_FINISH && iso9660->zisofs.allzero && avail + zstrm->total_in == ZF_BLOCK_SIZE) { if (iso9660->zisofs.block_offset != file->cur_content->size) { int64_t diff; r = wb_set_offset(a, file->cur_content->offset_of_temp + iso9660->zisofs.block_offset); if (r != ARCHIVE_OK) return (r); diff = file->cur_content->size - iso9660->zisofs.block_offset; file->cur_content->size -= diff; iso9660->zisofs.total_size -= diff; } zstrm->avail_in = 0; } /* * Compress file data. */ while (zstrm->avail_in > 0) { csize = zstrm->total_out; r = deflate(zstrm, flush); switch (r) { case Z_OK: case Z_STREAM_END: csize = zstrm->total_out - csize; if (wb_consume(a, csize) != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->zisofs.total_size += csize; iso9660->cur_file->cur_content->size += csize; zstrm->next_out = wb_buffptr(a); zstrm->avail_out = wb_remaining(a); break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Compression failed:" " deflate() call returned status %d", r); return (ARCHIVE_FATAL); } } if (flush == Z_FINISH) { /* * Save the information of one zisofs block. */ iso9660->zisofs.block_pointers_idx ++; archive_le32enc(&(iso9660->zisofs.block_pointers[ iso9660->zisofs.block_pointers_idx]), iso9660->zisofs.total_size); r = zisofs_init_zstream(a); if (r != ARCHIVE_OK) return (ARCHIVE_FATAL); iso9660->zisofs.allzero = 1; iso9660->zisofs.block_offset = file->cur_content->size; } } while (s); return (ARCHIVE_OK); } static int zisofs_finish_entry(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; struct isofile *file = iso9660->cur_file; unsigned char buff[16]; size_t s; int64_t tail; /* Direct temp file stream to zisofs temp file stream. */ archive_entry_set_size(file->entry, iso9660->zisofs.total_size); /* * Save a file pointer which points the end of current zisofs data. */ tail = wb_offset(a); /* * Make a header. * * +-----------------+----------------+-----------------+ * | Header 16 bytes | Block Pointers | Compressed data | * +-----------------+----------------+-----------------+ * 0 16 +X * Block Pointers : * 4 * (((Uncompressed file size + block_size -1) / block_size) + 1) * * Write zisofs header. * Magic number * +----+----+----+----+----+----+----+----+ * | 37 | E4 | 53 | 96 | C9 | DB | D6 | 07 | * +----+----+----+----+----+----+----+----+ * 0 1 2 3 4 5 6 7 8 * * +------------------------+------------------+ * | Uncompressed file size | header_size >> 2 | * +------------------------+------------------+ * 8 12 13 * * +-----------------+----------------+ * | log2 block_size | Reserved(0000) | * +-----------------+----------------+ * 13 14 16 */ memcpy(buff, zisofs_magic, 8); set_num_731(buff+8, file->zisofs.uncompressed_size); buff[12] = file->zisofs.header_size; buff[13] = file->zisofs.log2_bs; buff[14] = buff[15] = 0;/* Reserved */ /* Move to the right position to write the header. */ wb_set_offset(a, file->content.offset_of_temp); /* Write the header. */ if (wb_write_to_temp(a, buff, 16) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* * Write zisofs Block Pointers. */ s = iso9660->zisofs.block_pointers_cnt * sizeof(iso9660->zisofs.block_pointers[0]); if (wb_write_to_temp(a, iso9660->zisofs.block_pointers, s) != ARCHIVE_OK) return (ARCHIVE_FATAL); /* Set a file pointer back to the end of the temporary file. */ wb_set_offset(a, tail); return (ARCHIVE_OK); } static int zisofs_free(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; int ret = ARCHIVE_OK; free(iso9660->zisofs.block_pointers); if (iso9660->zisofs.stream_valid && deflateEnd(&(iso9660->zisofs.stream)) != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to clean up compressor"); ret = ARCHIVE_FATAL; } iso9660->zisofs.block_pointers = NULL; iso9660->zisofs.stream_valid = 0; return (ret); } struct zisofs_extract { int pz_log2_bs; /* Log2 of block size */ uint64_t pz_uncompressed_size; size_t uncompressed_buffer_size; int initialized:1; int header_passed:1; uint32_t pz_offset; unsigned char *block_pointers; size_t block_pointers_size; size_t block_pointers_avail; size_t block_off; uint32_t block_avail; z_stream stream; int stream_valid; }; static ssize_t zisofs_extract_init(struct archive_write *a, struct zisofs_extract *zisofs, const unsigned char *p, size_t bytes) { size_t avail = bytes; size_t ceil, xsize; /* Allocate block pointers buffer. */ ceil = (zisofs->pz_uncompressed_size + (1LL << zisofs->pz_log2_bs) - 1) >> zisofs->pz_log2_bs; xsize = (ceil + 1) * 4; if (zisofs->block_pointers == NULL) { size_t alloc = ((xsize >> 10) + 1) << 10; zisofs->block_pointers = malloc(alloc); if (zisofs->block_pointers == NULL) { archive_set_error(&a->archive, ENOMEM, "No memory for zisofs decompression"); return (ARCHIVE_FATAL); } } zisofs->block_pointers_size = xsize; /* Allocate uncompressed data buffer. */ zisofs->uncompressed_buffer_size = 1UL << zisofs->pz_log2_bs; /* * Read the file header, and check the magic code of zisofs. */ if (!zisofs->header_passed) { int err = 0; if (avail < 16) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Illegal zisofs file body"); return (ARCHIVE_FATAL); } if (memcmp(p, zisofs_magic, sizeof(zisofs_magic)) != 0) err = 1; else if (archive_le32dec(p + 8) != zisofs->pz_uncompressed_size) err = 1; else if (p[12] != 4 || p[13] != zisofs->pz_log2_bs) err = 1; if (err) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Illegal zisofs file body"); return (ARCHIVE_FATAL); } avail -= 16; p += 16; zisofs->header_passed = 1; } /* * Read block pointers. */ if (zisofs->header_passed && zisofs->block_pointers_avail < zisofs->block_pointers_size) { xsize = zisofs->block_pointers_size - zisofs->block_pointers_avail; if (avail < xsize) xsize = avail; memcpy(zisofs->block_pointers + zisofs->block_pointers_avail, p, xsize); zisofs->block_pointers_avail += xsize; avail -= xsize; if (zisofs->block_pointers_avail == zisofs->block_pointers_size) { /* We've got all block pointers and initialize * related variables. */ zisofs->block_off = 0; zisofs->block_avail = 0; /* Complete a initialization */ zisofs->initialized = 1; } } return ((ssize_t)avail); } static ssize_t zisofs_extract(struct archive_write *a, struct zisofs_extract *zisofs, const unsigned char *p, size_t bytes) { size_t avail; int r; if (!zisofs->initialized) { ssize_t rs = zisofs_extract_init(a, zisofs, p, bytes); if (rs < 0) return (rs); if (!zisofs->initialized) { /* We need more data. */ zisofs->pz_offset += bytes; return (bytes); } avail = rs; p += bytes - avail; } else avail = bytes; /* * Get block offsets from block pointers. */ if (zisofs->block_avail == 0) { uint32_t bst, bed; if (zisofs->block_off + 4 >= zisofs->block_pointers_size) { /* There isn't a pair of offsets. */ archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Illegal zisofs block pointers"); return (ARCHIVE_FATAL); } bst = archive_le32dec( zisofs->block_pointers + zisofs->block_off); if (bst != zisofs->pz_offset + (bytes - avail)) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Illegal zisofs block pointers(cannot seek)"); return (ARCHIVE_FATAL); } bed = archive_le32dec( zisofs->block_pointers + zisofs->block_off + 4); if (bed < bst) { archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, "Illegal zisofs block pointers"); return (ARCHIVE_FATAL); } zisofs->block_avail = bed - bst; zisofs->block_off += 4; /* Initialize compression library for new block. */ if (zisofs->stream_valid) r = inflateReset(&zisofs->stream); else r = inflateInit(&zisofs->stream); if (r != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Can't initialize zisofs decompression."); return (ARCHIVE_FATAL); } zisofs->stream_valid = 1; zisofs->stream.total_in = 0; zisofs->stream.total_out = 0; } /* * Make uncompressed data. */ if (zisofs->block_avail == 0) { /* * It's basically 32K bytes NUL data. */ unsigned char *wb; size_t size, wsize; size = zisofs->uncompressed_buffer_size; while (size) { wb = wb_buffptr(a); if (size > wb_remaining(a)) wsize = wb_remaining(a); else wsize = size; memset(wb, 0, wsize); r = wb_consume(a, wsize); if (r < 0) return (r); size -= wsize; } } else { zisofs->stream.next_in = (Bytef *)(uintptr_t)(const void *)p; if (avail > zisofs->block_avail) zisofs->stream.avail_in = zisofs->block_avail; else zisofs->stream.avail_in = avail; zisofs->stream.next_out = wb_buffptr(a); zisofs->stream.avail_out = wb_remaining(a); r = inflate(&zisofs->stream, 0); switch (r) { case Z_OK: /* Decompressor made some progress.*/ case Z_STREAM_END: /* Found end of stream. */ break; default: archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "zisofs decompression failed (%d)", r); return (ARCHIVE_FATAL); } avail -= zisofs->stream.next_in - p; zisofs->block_avail -= zisofs->stream.next_in - p; r = wb_consume(a, wb_remaining(a) - zisofs->stream.avail_out); if (r < 0) return (r); } zisofs->pz_offset += bytes; return (bytes - avail); } static int zisofs_rewind_boot_file(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; struct isofile *file; unsigned char *rbuff; ssize_t r; size_t remaining, rbuff_size; struct zisofs_extract zext; int64_t read_offset, write_offset, new_offset; int fd, ret = ARCHIVE_OK; file = iso9660->el_torito.boot->file; /* * There is nothing to do if this boot file does not have * zisofs header. */ if (file->zisofs.header_size == 0) return (ARCHIVE_OK); /* * Uncompress the zisofs'ed file contents. */ memset(&zext, 0, sizeof(zext)); zext.pz_uncompressed_size = file->zisofs.uncompressed_size; zext.pz_log2_bs = file->zisofs.log2_bs; fd = iso9660->temp_fd; new_offset = wb_offset(a); read_offset = file->content.offset_of_temp; remaining = file->content.size; if (remaining > 1024 * 32) rbuff_size = 1024 * 32; else rbuff_size = remaining; rbuff = malloc(rbuff_size); if (rbuff == NULL) { archive_set_error(&a->archive, ENOMEM, "Can't allocate memory"); return (ARCHIVE_FATAL); } while (remaining) { size_t rsize; ssize_t rs; /* Get the current file pointer. */ write_offset = lseek(fd, 0, SEEK_CUR); /* Change the file pointer to read. */ lseek(fd, read_offset, SEEK_SET); rsize = rbuff_size; if (rsize > remaining) rsize = remaining; rs = read(iso9660->temp_fd, rbuff, rsize); if (rs <= 0) { archive_set_error(&a->archive, errno, "Can't read temporary file(%jd)", (intmax_t)rs); ret = ARCHIVE_FATAL; break; } remaining -= rs; read_offset += rs; /* Put the file pointer back to write. */ lseek(fd, write_offset, SEEK_SET); r = zisofs_extract(a, &zext, rbuff, rs); if (r < 0) { ret = (int)r; break; } } if (ret == ARCHIVE_OK) { /* * Change the boot file content from zisofs'ed data * to plain data. */ file->content.offset_of_temp = new_offset; file->content.size = file->zisofs.uncompressed_size; archive_entry_set_size(file->entry, file->content.size); /* Set to be no zisofs. */ file->zisofs.header_size = 0; file->zisofs.log2_bs = 0; file->zisofs.uncompressed_size = 0; r = wb_write_padding_to_temp(a, file->content.size); if (r < 0) ret = ARCHIVE_FATAL; } /* * Free the resource we used in this function only. */ free(rbuff); free(zext.block_pointers); if (zext.stream_valid && inflateEnd(&(zext.stream)) != Z_OK) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Failed to clean up compressor"); ret = ARCHIVE_FATAL; } return (ret); } #else static int zisofs_write_to_temp(struct archive_write *a, const void *buff, size_t s) { (void)buff; /* UNUSED */ (void)s; /* UNUSED */ archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "Programing error"); return (ARCHIVE_FATAL); } static int zisofs_rewind_boot_file(struct archive_write *a) { struct iso9660 *iso9660 = a->format_data; if (iso9660->el_torito.boot->file->zisofs.header_size != 0) { archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, "We cannot extract the zisofs imaged boot file;" " this may not boot in being zisofs imaged"); return (ARCHIVE_FAILED); } return (ARCHIVE_OK); } static int zisofs_finish_entry(struct archive_write *a) { (void)a; /* UNUSED */ return (ARCHIVE_OK); } static int zisofs_free(struct archive_write *a) { (void)a; /* UNUSED */ return (ARCHIVE_OK); } #endif /* HAVE_ZLIB_H */ Index: vendor/libarchive/dist/libarchive/test/main.c =================================================================== --- vendor/libarchive/dist/libarchive/test/main.c (revision 231296) +++ vendor/libarchive/dist/libarchive/test/main.c (revision 231297) @@ -1,2610 +1,2610 @@ /* * Copyright (c) 2003-2009 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" #ifdef HAVE_SYS_TIME_H #include #endif #include #ifdef HAVE_ICONV_H #include #endif #include #include #ifdef HAVE_SIGNAL_H #include #endif #include #include /* * This same file is used pretty much verbatim for all test harnesses. * * The next few lines are the only differences. * TODO: Move this into a separate configuration header, have all test * suites share one copy of this file. */ __FBSDID("$FreeBSD: head/lib/libarchive/test/main.c 201247 2009-12-30 05:59:21Z kientzle $"); #define KNOWNREF "test_compat_gtar_1.tar.uu" #define ENVBASE "LIBARCHIVE" /* Prefix for environment variables. */ #undef PROGRAM /* Testing a library, not a program. */ #define LIBRARY "libarchive" #define EXTRA_DUMP(x) archive_error_string((struct archive *)(x)) #define EXTRA_ERRNO(x) archive_errno((struct archive *)(x)) #define EXTRA_VERSION archive_version_string() /* * * Windows support routines * * Note: Configuration is a tricky issue. Using HAVE_* feature macros * in the test harness is dangerous because they cover up * configuration errors. The classic example of this is omitting a * configure check. If libarchive and libarchive_test both look for * the same feature macro, such errors are hard to detect. Platform * macros (e.g., _WIN32 or __GNUC__) are a little better, but can * easily lead to very messy code. It's best to limit yourself * to only the most generic programming techniques in the test harness * and thus avoid conditionals altogether. Where that's not possible, * try to minimize conditionals by grouping platform-specific tests in * one place (e.g., test_acl_freebsd) or by adding new assert() * functions (e.g., assertMakeHardlink()) to cover up platform * differences. Platform-specific coding in libarchive_test is often * a symptom that some capability is missing from libarchive itself. */ #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #ifndef F_OK #define F_OK (0) #endif #ifndef S_ISDIR #define S_ISDIR(m) ((m) & _S_IFDIR) #endif #ifndef S_ISREG #define S_ISREG(m) ((m) & _S_IFREG) #endif #if !defined(__BORLANDC__) #define access _access #undef chdir #define chdir _chdir #endif #ifndef fileno #define fileno _fileno #endif /*#define fstat _fstat64*/ #if !defined(__BORLANDC__) #define getcwd _getcwd #endif #define lstat stat /*#define lstat _stat64*/ /*#define stat _stat64*/ #define rmdir _rmdir #if !defined(__BORLANDC__) #define strdup _strdup #define umask _umask #endif #define int64_t __int64 #endif #if defined(HAVE__CrtSetReportMode) # include #endif #if defined(_WIN32) && !defined(__CYGWIN__) void *GetFunctionKernel32(const char *name) { static HINSTANCE lib; static int set; if (!set) { set = 1; lib = LoadLibrary("kernel32.dll"); } if (lib == NULL) { fprintf(stderr, "Can't load kernel32.dll?!\n"); exit(1); } return (void *)GetProcAddress(lib, name); } static int my_CreateSymbolicLinkA(const char *linkname, const char *target, int flags) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, DWORD); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateSymbolicLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, flags); } static int my_CreateHardLinkA(const char *linkname, const char *target) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, LPSECURITY_ATTRIBUTES); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateHardLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, NULL); } int my_GetFileInformationByName(const char *path, BY_HANDLE_FILE_INFORMATION *bhfi) { HANDLE h; int r; memset(bhfi, 0, sizeof(*bhfi)); h = CreateFile(path, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) return (0); r = GetFileInformationByHandle(h, bhfi); CloseHandle(h); return (r); } #endif #if defined(HAVE__CrtSetReportMode) static void invalid_parameter_handler(const wchar_t * expression, const wchar_t * function, const wchar_t * file, unsigned int line, uintptr_t pReserved) { /* nop */ } #endif /* * * OPTIONS FLAGS * */ /* Enable core dump on failure. */ static int dump_on_failure = 0; /* Default is to remove temp dirs and log data for successful tests. */ static int keep_temp_files = 0; /* Default is to run the specified tests once and report errors. */ static int until_failure = 0; /* Default is to just report pass/fail for each test. */ static int verbosity = 0; #define VERBOSITY_SUMMARY_ONLY -1 /* -q */ #define VERBOSITY_PASSFAIL 0 /* Default */ #define VERBOSITY_LIGHT_REPORT 1 /* -v */ #define VERBOSITY_FULL 2 /* -vv */ /* A few places generate even more output for verbosity > VERBOSITY_FULL, * mostly for debugging the test harness itself. */ /* Cumulative count of assertion failures. */ static int failures = 0; /* Cumulative count of reported skips. */ static int skips = 0; /* Cumulative count of assertions checked. */ static int assertions = 0; /* Directory where uuencoded reference files can be found. */ static const char *refdir; /* * Report log information selectively to console and/or disk log. */ static int log_console = 0; static FILE *logfile; static void vlogprintf(const char *fmt, va_list ap) { #ifdef va_copy va_list lfap; va_copy(lfap, ap); #endif if (log_console) vfprintf(stdout, fmt, ap); if (logfile != NULL) #ifdef va_copy vfprintf(logfile, fmt, lfap); va_end(lfap); #else vfprintf(logfile, fmt, ap); #endif } static void logprintf(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vlogprintf(fmt, ap); va_end(ap); } /* Set up a message to display only if next assertion fails. */ static char msgbuff[4096]; static const char *msg, *nextmsg; void failure(const char *fmt, ...) { va_list ap; if (fmt == NULL) { nextmsg = NULL; } else { va_start(ap, fmt); vsprintf(msgbuff, fmt, ap); va_end(ap); nextmsg = msgbuff; } } /* * Copy arguments into file-local variables. * This was added to permit vararg assert() functions without needing * variadic wrapper macros. Turns out that the vararg capability is almost * never used, so almost all of the vararg assertions can be simplified * by removing the vararg capability and reworking the wrapper macro to * pass __FILE__, __LINE__ directly into the function instead of using * this hook. I suspect this machinery is used so rarely that we * would be better off just removing it entirely. That would simplify * the code here noticeably. */ static const char *skipping_filename; static int skipping_line; void skipping_setup(const char *filename, int line) { skipping_filename = filename; skipping_line = line; } /* Called at the beginning of each assert() function. */ static void assertion_count(const char *file, int line) { (void)file; /* UNUSED */ (void)line; /* UNUSED */ ++assertions; /* Proper handling of "failure()" message. */ msg = nextmsg; nextmsg = NULL; /* Uncomment to print file:line after every assertion. * Verbose, but occasionally useful in tracking down crashes. */ /* printf("Checked %s:%d\n", file, line); */ } /* * For each test source file, we remember how many times each * assertion was reported. Cleared before each new test, * used by test_summarize(). */ static struct line { int count; int skip; } failed_lines[10000]; const char *failed_filename; /* Count this failure, setup up log destination and handle initial report. */ static void failure_start(const char *filename, int line, const char *fmt, ...) { va_list ap; /* Record another failure for this line. */ ++failures; failed_filename = filename; failed_lines[line].count++; /* Determine whether to log header to console. */ switch (verbosity) { case VERBOSITY_LIGHT_REPORT: log_console = (failed_lines[line].count < 2); break; default: log_console = (verbosity >= VERBOSITY_FULL); } /* Log file:line header for this failure */ va_start(ap, fmt); #if _MSC_VER logprintf("%s(%d): ", filename, line); #else logprintf("%s:%d: ", filename, line); #endif vlogprintf(fmt, ap); va_end(ap); logprintf("\n"); if (msg != NULL && msg[0] != '\0') { logprintf(" Description: %s\n", msg); msg = NULL; } /* Determine whether to log details to console. */ if (verbosity == VERBOSITY_LIGHT_REPORT) log_console = 0; } /* Complete reporting of failed tests. */ /* * The 'extra' hook here is used by libarchive to include libarchive * error messages with assertion failures. It could also be used * to add strerror() output, for example. Just define the EXTRA_DUMP() * macro appropriately. */ static void failure_finish(void *extra) { (void)extra; /* UNUSED (maybe) */ #ifdef EXTRA_DUMP if (extra != NULL) { logprintf(" errno: %d\n", EXTRA_ERRNO(extra)); logprintf(" detail: %s\n", EXTRA_DUMP(extra)); } #endif if (dump_on_failure) { fprintf(stderr, " *** forcing core dump so failure can be debugged ***\n"); abort(); exit(1); } } /* Inform user that we're skipping some checks. */ void test_skipping(const char *fmt, ...) { char buff[1024]; va_list ap; va_start(ap, fmt); vsprintf(buff, fmt, ap); va_end(ap); /* Use failure() message if set. */ msg = nextmsg; nextmsg = NULL; /* failure_start() isn't quite right, but is awfully convenient. */ failure_start(skipping_filename, skipping_line, "SKIPPING: %s", buff); --failures; /* Undo failures++ in failure_start() */ /* Don't failure_finish() here. */ /* Mark as skip, so doesn't count as failed test. */ failed_lines[skipping_line].skip = 1; ++skips; } /* * * ASSERTIONS * */ /* Generic assert() just displays the failed condition. */ int assertion_assert(const char *file, int line, int value, const char *condition, void *extra) { assertion_count(file, line); if (!value) { failure_start(file, line, "Assertion failed: %s", condition); failure_finish(extra); } return (value); } /* chdir() and report any errors */ int assertion_chdir(const char *file, int line, const char *pathname) { assertion_count(file, line); if (chdir(pathname) == 0) return (1); failure_start(file, line, "chdir(\"%s\")", pathname); failure_finish(NULL); return (0); } /* Verify two integers are equal. */ int assertion_equal_int(const char *file, int line, long long v1, const char *e1, long long v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e1, v1, v1, v1); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e2, v2, v2, v2); failure_finish(extra); return (0); } /* * Utility to convert a single UTF-8 sequence. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch; int cnt; uint32_t wc; *pwc = 0; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if (n < (size_t)cnt) return (-1); /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) return (-1);/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if (n < 4) return (-1); if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); if ((s[3] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) return (-1);/* Overlong sequence. */ break; default: return (-1); } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > 0x10FFFF) return (-1); /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); } static void strdump(const char *e, const char *p, int ewidth, int utf8) { const char *q = p; logprintf(" %*s = ", ewidth, e); if (p == NULL) { logprintf("NULL\n"); return; } logprintf("\""); while (*p != '\0') { unsigned int c = 0xff & *p++; switch (c) { case '\a': printf("\a"); break; case '\b': printf("\b"); break; case '\n': printf("\n"); break; case '\r': printf("\r"); break; default: if (c >= 32 && c < 127) logprintf("%c", c); else logprintf("\\x%02X", c); } } logprintf("\""); logprintf(" (length %d)", q == NULL ? -1 : (int)strlen(q)); /* * If the current string is UTF-8, dump its code points. */ if (utf8) { size_t len; uint32_t uc; int n; int cnt = 0; p = q; len = strlen(p); logprintf(" ["); while ((n = _utf8_to_unicode(&uc, p, len)) > 0) { if (p != q) logprintf(" "); logprintf("%04X", uc); p += n; len -= n; cnt++; } logprintf("]"); logprintf(" (count %d", cnt); if (n < 0) { logprintf(",unknown %d bytes", len); } logprintf(")"); } logprintf("\n"); } /* Verify two strings are equal, dump them if not. */ int assertion_equal_string(const char *file, int line, const char *v1, const char *e1, const char *v2, const char *e2, void *extra, int utf8) { int l1, l2; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && strcmp(v1, v2) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); l1 = strlen(e1); l2 = strlen(e2); if (l1 < l2) l1 = l2; strdump(e1, v1, l1, utf8); strdump(e2, v2, l1, utf8); failure_finish(extra); return (0); } static void wcsdump(const char *e, const wchar_t *w) { logprintf(" %s = ", e); if (w == NULL) { logprintf("(null)"); return; } logprintf("\""); while (*w != L'\0') { unsigned int c = *w++; if (c >= 32 && c < 127) logprintf("%c", c); else if (c < 256) logprintf("\\x%02X", c); else if (c < 0x10000) logprintf("\\u%04X", c); else logprintf("\\U%08X", c); } logprintf("\"\n"); } #ifndef HAVE_WCSCMP static int wcscmp(const wchar_t *s1, const wchar_t *s2) { while (*s1 == *s2++) { if (*s1++ == L'\0') return 0; } if (*s1 > *--s2) return 1; else return -1; } #endif /* Verify that two wide strings are equal, dump them if not. */ int assertion_equal_wstring(const char *file, int line, const wchar_t *v1, const char *e1, const wchar_t *v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); if (v1 != NULL && v2 != NULL && wcscmp(v1, v2) == 0) return (1); failure_start(file, line, "%s != %s", e1, e2); wcsdump(e1, v1); wcsdump(e2, v2); failure_finish(extra); return (0); } /* * Pretty standard hexdump routine. As a bonus, if ref != NULL, then * any bytes in p that differ from ref will be highlighted with '_' * before and after the hex value. */ static void hexdump(const char *p, const char *ref, size_t l, size_t offset) { size_t i, j; char sep; if (p == NULL) { logprintf("(null)\n"); return; } for(i=0; i < l; i+=16) { logprintf("%04x", (unsigned)(i + offset)); sep = ' '; for (j = 0; j < 16 && i + j < l; j++) { if (ref != NULL && p[i + j] != ref[i + j]) sep = '_'; logprintf("%c%02x", sep, 0xff & (int)p[i+j]); if (ref != NULL && p[i + j] == ref[i + j]) sep = ' '; } for (; j < 16; j++) { logprintf("%c ", sep); sep = ' '; } logprintf("%c", sep); for (j=0; j < 16 && i + j < l; j++) { int c = p[i + j]; if (c >= ' ' && c <= 126) logprintf("%c", c); else logprintf("."); } logprintf("\n"); } } /* Verify that two blocks of memory are the same, display the first * block of differences if they're not. */ int assertion_equal_mem(const char *file, int line, const void *_v1, const char *e1, const void *_v2, const char *e2, size_t l, const char *ld, void *extra) { const char *v1 = (const char *)_v1; const char *v2 = (const char *)_v2; size_t offset; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && memcmp(v1, v2, l) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" size %s = %d\n", ld, (int)l); /* Dump 48 bytes (3 lines) so that the first difference is * in the second line. */ offset = 0; while (l > 64 && memcmp(v1, v2, 32) == 0) { /* Two lines agree, so step forward one line. */ v1 += 16; v2 += 16; l -= 16; offset += 16; } logprintf(" Dump of %s\n", e1); hexdump(v1, v2, l < 128 ? l : 128, offset); logprintf(" Dump of %s\n", e2); hexdump(v2, v1, l < 128 ? l : 128, offset); logprintf("\n"); failure_finish(extra); return (0); } /* Verify that the named file exists and is empty. */ int assertion_empty_file(const char *filename, int line, const char *f1) { char buff[1024]; struct stat st; ssize_t s; FILE *f; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) return (1); failure_start(filename, line, "File should be empty: %s", f1); logprintf(" File size: %d\n", (int)st.st_size); logprintf(" Contents:\n"); f = fopen(f1, "rb"); if (f == NULL) { logprintf(" Unable to open %s\n", f1); } else { s = ((off_t)sizeof(buff) < st.st_size) ? (ssize_t)sizeof(buff) : (ssize_t)st.st_size; s = fread(buff, 1, s, f); hexdump(buff, NULL, s, 0); fclose(f); } failure_finish(NULL); return (0); } /* Verify that the named file exists and is not empty. */ int assertion_non_empty_file(const char *filename, int line, const char *f1) { struct stat st; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) { failure_start(filename, line, "File empty: %s", f1); failure_finish(NULL); return (0); } return (1); } /* Verify that two files have the same contents. */ /* TODO: hexdump the first bytes that actually differ. */ int assertion_equal_file(const char *filename, int line, const char *fn1, const char *fn2) { char buff1[1024]; char buff2[1024]; FILE *f1, *f2; int n1, n2; assertion_count(filename, line); f1 = fopen(fn1, "rb"); f2 = fopen(fn2, "rb"); for (;;) { n1 = fread(buff1, 1, sizeof(buff1), f1); n2 = fread(buff2, 1, sizeof(buff2), f2); if (n1 != n2) break; if (n1 == 0 && n2 == 0) { fclose(f1); fclose(f2); return (1); } if (memcmp(buff1, buff2, n1) != 0) break; } fclose(f1); fclose(f2); failure_start(filename, line, "Files not identical"); logprintf(" file1=\"%s\"\n", fn1); logprintf(" file2=\"%s\"\n", fn2); failure_finish(NULL); return (0); } /* Verify that the named file does exist. */ int assertion_file_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (!_access(f, 0)) return (1); #else if (!access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should exist: %s", f); failure_finish(NULL); return (0); } /* Verify that the named file doesn't exist. */ int assertion_file_not_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (_access(f, 0)) return (1); #else if (access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should not exist: %s", f); failure_finish(NULL); return (0); } /* Compare the contents of a file to a block of memory. */ int assertion_file_contents(const char *filename, int line, const void *buff, int s, const char *fn) { char *contents; FILE *f; int n; assertion_count(filename, line); f = fopen(fn, "rb"); if (f == NULL) { failure_start(filename, line, "File should exist: %s", fn); failure_finish(NULL); return (0); } contents = malloc(s * 2); n = fread(contents, 1, s * 2, f); fclose(f); if (n == s && memcmp(buff, contents, s) == 0) { free(contents); return (1); } failure_start(filename, line, "File contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) hexdump(contents, buff, n > 512 ? 512 : n, 0); else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s > 512 ? 512 : s, 0); } failure_finish(NULL); free(contents); return (0); } /* Check the contents of a text file, being tolerant of line endings. */ int assertion_text_file_contents(const char *filename, int line, const char *buff, const char *fn) { char *contents; const char *btxt, *ftxt; FILE *f; int n, s; assertion_count(filename, line); f = fopen(fn, "r"); if (f == NULL) { failure_start(filename, line, "File doesn't exist: %s", fn); failure_finish(NULL); return (0); } s = strlen(buff); contents = malloc(s * 2 + 128); n = fread(contents, 1, s * 2 + 128 - 1, f); if (n >= 0) contents[n] = '\0'; fclose(f); /* Compare texts. */ btxt = buff; ftxt = (const char *)contents; while (*btxt != '\0' && *ftxt != '\0') { if (*btxt == *ftxt) { ++btxt; ++ftxt; continue; } if (btxt[0] == '\n' && ftxt[0] == '\r' && ftxt[1] == '\n') { /* Pass over different new line characters. */ ++btxt; ftxt += 2; continue; } break; } if (*btxt == '\0' && *ftxt == '\0') { free(contents); return (1); } failure_start(filename, line, "Contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) { hexdump(contents, buff, n, 0); logprintf(" expected\n", fn); hexdump(buff, contents, s, 0); } else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s, 0); } failure_finish(NULL); free(contents); return (0); } /* Verify that a text file contains the specified lines, regardless of order */ /* This could be more efficient if we sorted both sets of lines, etc, but * since this is used only for testing and only ever deals with a dozen or so * lines at a time, this relatively crude approach is just fine. */ int assertion_file_contains_lines_any_order(const char *file, int line, const char *pathname, const char *lines[]) { char *buff; size_t buff_size; size_t expected_count, actual_count, i, j; char **expected; char *p, **actual; char c; int expected_failure = 0, actual_failure = 0; assertion_count(file, line); buff = slurpfile(&buff_size, "%s", pathname); if (buff == NULL) { failure_start(pathname, line, "Can't read file: %s", pathname); failure_finish(NULL); return (0); } /* Make a copy of the provided lines and count up the expected file size. */ expected_count = 0; for (i = 0; lines[i] != NULL; ++i) { } expected_count = i; expected = malloc(sizeof(char *) * expected_count); for (i = 0; lines[i] != NULL; ++i) { expected[i] = strdup(lines[i]); } /* Break the file into lines */ actual_count = 0; for (c = '\0', p = buff; p < buff + buff_size; ++p) { if (*p == '\x0d' || *p == '\x0a') *p = '\0'; if (c == '\0' && *p != '\0') ++actual_count; c = *p; } actual = malloc(sizeof(char *) * actual_count); for (j = 0, p = buff; p < buff + buff_size; p += 1 + strlen(p)) { if (*p != '\0') { actual[j] = p; ++j; } } /* Erase matching lines from both lists */ for (i = 0; i < expected_count; ++i) { if (expected[i] == NULL) continue; for (j = 0; j < actual_count; ++j) { if (actual[j] == NULL) continue; if (strcmp(expected[i], actual[j]) == 0) { free(expected[i]); expected[i] = NULL; actual[j] = NULL; break; } } } /* If there's anything left, it's a failure */ for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) ++expected_failure; } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) ++actual_failure; } if (expected_failure == 0 && actual_failure == 0) { free(buff); free(expected); free(actual); return (1); } failure_start(file, line, "File doesn't match: %s", pathname); for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) { logprintf(" Expected but not present: %s\n", expected[i]); free(expected[i]); } } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) logprintf(" Present but not expected: %s\n", actual[j]); } failure_finish(NULL); free(buff); free(expected); free(actual); return (0); } /* Test that two paths point to the same file. */ /* As a side-effect, asserts that both files exist. */ static int is_hardlink(const char *file, int line, const char *path1, const char *path2) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi1, bhfi2; int r; assertion_count(file, line); r = my_GetFileInformationByName(path1, &bhfi1); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path1); failure_finish(NULL); return (0); } r = my_GetFileInformationByName(path2, &bhfi2); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path2); failure_finish(NULL); return (0); } return (bhfi1.dwVolumeSerialNumber == bhfi2.dwVolumeSerialNumber && bhfi1.nFileIndexHigh == bhfi2.nFileIndexHigh && bhfi1.nFileIndexLow == bhfi2.nFileIndexLow); #else struct stat st1, st2; int r; assertion_count(file, line); r = lstat(path1, &st1); if (r != 0) { failure_start(file, line, "File should exist: %s", path1); failure_finish(NULL); return (0); } r = lstat(path2, &st2); if (r != 0) { failure_start(file, line, "File should exist: %s", path2); failure_finish(NULL); return (0); } return (st1.st_ino == st2.st_ino && st1.st_dev == st2.st_dev); #endif } int assertion_is_hardlink(const char *file, int line, const char *path1, const char *path2) { if (is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s are not hardlinked", path1, path2); failure_finish(NULL); return (0); } int assertion_is_not_hardlink(const char *file, int line, const char *path1, const char *path2) { if (!is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s should not be hardlinked", path1, path2); failure_finish(NULL); return (0); } /* Verify a/b/mtime of 'pathname'. */ /* If 'recent', verify that it's within last 10 seconds. */ static int assertion_file_time(const char *file, int line, const char *pathname, long t, long nsec, char type, int recent) { long long filet, filet_nsec; int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define EPOC_TIME (116444736000000000ULL) FILETIME ftime, fbirthtime, fatime, fmtime; ULARGE_INTEGER wintm; HANDLE h; ftime.dwLowDateTime = 0; ftime.dwHighDateTime = 0; assertion_count(file, line); /* Note: FILE_FLAG_BACKUP_SEMANTICS applies to open * a directory file. If not, CreateFile() will fail when * the pathname is a directory. */ h = CreateFile(pathname, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } r = GetFileTime(h, &fbirthtime, &fatime, &fmtime); switch (type) { case 'a': ftime = fatime; break; case 'b': ftime = fbirthtime; break; case 'm': ftime = fmtime; break; } CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't GetFileTime %s\n", pathname); failure_finish(NULL); return (0); } wintm.LowPart = ftime.dwLowDateTime; wintm.HighPart = ftime.dwHighDateTime; filet = (wintm.QuadPart - EPOC_TIME) / 10000000; filet_nsec = ((wintm.QuadPart - EPOC_TIME) % 10000000) * 100; nsec = (nsec / 100) * 100; /* Round the request */ #else struct stat st; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } switch (type) { case 'a': filet = st.st_atime; break; case 'm': filet = st.st_mtime; break; case 'b': filet = 0; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } #if defined(__FreeBSD__) switch (type) { case 'a': filet_nsec = st.st_atimespec.tv_nsec; break; case 'b': filet = st.st_birthtime; filet_nsec = st.st_birthtimespec.tv_nsec; break; case 'm': filet_nsec = st.st_mtimespec.tv_nsec; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } /* FreeBSD generally only stores to microsecond res, so round. */ filet_nsec = (filet_nsec / 1000) * 1000; nsec = (nsec / 1000) * 1000; #else filet_nsec = nsec = 0; /* Generic POSIX only has whole seconds. */ if (type == 'b') return (1); /* Generic POSIX doesn't have birthtime */ #if defined(__HAIKU__) if (type == 'a') return (1); /* Haiku doesn't have atime. */ #endif #endif #endif if (recent) { /* Check that requested time is up-to-date. */ time_t now = time(NULL); if (filet < now - 10 || filet > now + 1) { failure_start(file, line, "File %s has %ctime %lld, %lld seconds ago\n", pathname, type, filet, now - filet); failure_finish(NULL); return (0); } } else if (filet != t || filet_nsec != nsec) { failure_start(file, line, "File %s has %ctime %lld.%09lld, expected %lld.%09lld", pathname, type, filet, filet_nsec, t, nsec); failure_finish(NULL); return (0); } return (1); } /* Verify atime of 'pathname'. */ int assertion_file_atime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'a', 0); } /* Verify atime of 'pathname' is up-to-date. */ int assertion_file_atime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'a', 1); } /* Verify birthtime of 'pathname'. */ int assertion_file_birthtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'b', 0); } /* Verify birthtime of 'pathname' is up-to-date. */ int assertion_file_birthtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'b', 1); } /* Verify mtime of 'pathname'. */ int assertion_file_mtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'm', 0); } /* Verify mtime of 'pathname' is up-to-date. */ int assertion_file_mtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'm', 1); } /* Verify number of links to 'pathname'. */ int assertion_file_nlinks(const char *file, int line, const char *pathname, int nlinks) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; assertion_count(file, line); r = my_GetFileInformationByName(pathname, &bhfi); if (r != 0 && bhfi.nNumberOfLinks == (DWORD)nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, bhfi.nNumberOfLinks, nlinks); failure_finish(NULL); return (0); #else struct stat st; int r; assertion_count(file, line); r = lstat(pathname, &st); - if (r == 0 && st.st_nlink == nlinks) + if (r == 0 && (int)st.st_nlink == nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, st.st_nlink, nlinks); failure_finish(NULL); return (0); #endif } /* Verify size of 'pathname'. */ int assertion_file_size(const char *file, int line, const char *pathname, long size) { int64_t filesize; int r; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) { BY_HANDLE_FILE_INFORMATION bhfi; r = !my_GetFileInformationByName(pathname, &bhfi); filesize = ((int64_t)bhfi.nFileSizeHigh << 32) + bhfi.nFileSizeLow; } #else { struct stat st; r = lstat(pathname, &st); filesize = st.st_size; } #endif if (r == 0 && filesize == size) return (1); failure_start(file, line, "File %s has size %ld, expected %ld", pathname, (long)filesize, (long)size); failure_finish(NULL); return (0); } /* Assert that 'pathname' is a dir. If mode >= 0, verify that too. */ int assertion_is_dir(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Dir should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISDIR(st.st_mode)) { failure_start(file, line, "%s is not a dir", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "Dir %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Verify that 'pathname' is a regular file. If 'mode' is >= 0, * verify that too. */ int assertion_is_reg(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0 || !S_ISREG(st.st_mode)) { failure_start(file, line, "File should exist: %s", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "File %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Check whether 'pathname' is a symbolic link. If 'contents' is * non-NULL, verify that the symlink has those contents. */ static int is_symlink(const char *file, int line, const char *pathname, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) (void)pathname; /* UNUSED */ (void)contents; /* UNUSED */ assertion_count(file, line); /* Windows sort-of has real symlinks, but they're only usable * by privileged users and are crippled even then, so there's * really not much point in bothering with this. */ return (0); #else char buff[300]; struct stat st; ssize_t linklen; int r; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Symlink should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISLNK(st.st_mode)) return (0); if (contents == NULL) return (1); linklen = readlink(pathname, buff, sizeof(buff)); if (linklen < 0) { failure_start(file, line, "Can't read symlink %s", pathname); failure_finish(NULL); return (0); } buff[linklen] = '\0'; if (strcmp(buff, contents) != 0) return (0); return (1); #endif } /* Assert that path is a symlink that (optionally) contains contents. */ int assertion_is_symlink(const char *file, int line, const char *path, const char *contents) { if (is_symlink(file, line, path, contents)) return (1); if (contents) failure_start(file, line, "File %s is not a symlink to %s", path, contents); else failure_start(file, line, "File %s is not a symlink", path); failure_finish(NULL); return (0); } /* Create a directory and report any errors. */ int assertion_make_dir(const char *file, int line, const char *dirname, int mode) { assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ if (0 == _mkdir(dirname)) return (1); #else if (0 == mkdir(dirname, mode)) return (1); #endif failure_start(file, line, "Could not create directory %s", dirname); failure_finish(NULL); return(0); } /* Create a file with the specified contents and report any failures. */ int assertion_make_file(const char *file, int line, const char *path, int mode, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) /* TODO: Rework this to set file mode as well. */ FILE *f; (void)mode; /* UNUSED */ assertion_count(file, line); f = fopen(path, "wb"); if (f == NULL) { failure_start(file, line, "Could not create file %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if (strlen(contents) != fwrite(contents, 1, strlen(contents), f)) { fclose(f); failure_start(file, line, "Could not write file %s", path); failure_finish(NULL); return (0); } } fclose(f); return (1); #else int fd; assertion_count(file, line); fd = open(path, O_CREAT | O_WRONLY, mode >= 0 ? mode : 0644); if (fd < 0) { failure_start(file, line, "Could not create %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if ((ssize_t)strlen(contents) != write(fd, contents, strlen(contents))) { close(fd); failure_start(file, line, "Could not write to %s", path); failure_finish(NULL); return (0); } } close(fd); return (1); #endif } /* Create a hardlink and report any failures. */ int assertion_make_hardlink(const char *file, int line, const char *newpath, const char *linkto) { int succeeded; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) succeeded = my_CreateHardLinkA(newpath, linkto); #elif HAVE_LINK succeeded = !link(linkto, newpath); #else succeeded = 0; #endif if (succeeded) return (1); failure_start(file, line, "Could not create hardlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Create a symlink and report any failures. */ int assertion_make_symlink(const char *file, int line, const char *newpath, const char *linkto) { #if defined(_WIN32) && !defined(__CYGWIN__) int targetIsDir = 0; /* TODO: Fix this */ assertion_count(file, line); if (my_CreateSymbolicLinkA(newpath, linkto, targetIsDir)) return (1); #elif HAVE_SYMLINK assertion_count(file, line); if (0 == symlink(linkto, newpath)) return (1); #endif failure_start(file, line, "Could not create symlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Set umask, report failures. */ int assertion_umask(const char *file, int line, int mask) { assertion_count(file, line); (void)file; /* UNUSED */ (void)line; /* UNUSED */ umask(mask); return (1); } /* Set times, report failures. */ int assertion_utimes(const char *file, int line, const char *pathname, long at, long at_nsec, long mt, long mt_nsec) { int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define WINTIME(sec, nsec) ((Int32x32To64(sec, 10000000) + EPOC_TIME)\ + (((nsec)/1000)*10)) HANDLE h; ULARGE_INTEGER wintm; FILETIME fatime, fmtime; FILETIME *pat, *pmt; assertion_count(file, line); h = CreateFileA(pathname,GENERIC_READ | GENERIC_WRITE, FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } if (at > 0 || at_nsec > 0) { wintm.QuadPart = WINTIME(at, at_nsec); fatime.dwLowDateTime = wintm.LowPart; fatime.dwHighDateTime = wintm.HighPart; pat = &fatime; } else pat = NULL; if (mt > 0 || mt_nsec > 0) { wintm.QuadPart = WINTIME(mt, mt_nsec); fmtime.dwLowDateTime = wintm.LowPart; fmtime.dwHighDateTime = wintm.HighPart; pmt = &fmtime; } else pmt = NULL; if (pat != NULL || pmt != NULL) r = SetFileTime(h, NULL, pat, pmt); else r = 1; CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't SetFileTime %s\n", pathname); failure_finish(NULL); return (0); } return (1); #else /* defined(_WIN32) && !defined(__CYGWIN__) */ struct stat st; struct timeval times[2]; #if !defined(__FreeBSD__) mt_nsec = at_nsec = 0; /* Generic POSIX only has whole seconds. */ #endif if (mt == 0 && mt_nsec == 0 && at == 0 && at_nsec == 0) return (1); r = lstat(pathname, &st); if (r < 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } if (mt == 0 && mt_nsec == 0) { mt = st.st_mtime; #if defined(__FreeBSD__) mt_nsec = st.st_mtimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ mt_nsec = (mt_nsec / 1000) * 1000; #endif } if (at == 0 && at_nsec == 0) { at = st.st_atime; #if defined(__FreeBSD__) at_nsec = st.st_atimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ at_nsec = (at_nsec / 1000) * 1000; #endif } times[1].tv_sec = mt; times[1].tv_usec = mt_nsec / 1000; times[0].tv_sec = at; times[0].tv_usec = at_nsec / 1000; #ifdef HAVE_LUTIMES r = lutimes(pathname, times); #else r = utimes(pathname, times); #endif if (r < 0) { failure_start(file, line, "Can't utimes %s\n", pathname); failure_finish(NULL); return (0); } return (1); #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ } /* * * UTILITIES for use by tests. * */ /* * Check whether platform supports symlinks. This is intended * for tests to use in deciding whether to bother testing symlink * support; if the platform doesn't support symlinks, there's no point * in checking whether the program being tested can create them. * * Note that the first time this test is called, we actually go out to * disk to create and verify a symlink. This is necessary because * symlink support is actually a property of a particular filesystem * and can thus vary between directories on a single system. After * the first call, this returns the cached result from memory, so it's * safe to call it as often as you wish. */ int canSymlink(void) { /* Remember the test result */ static int value = 0, tested = 0; if (tested) return (value); ++tested; assertion_make_file(__FILE__, __LINE__, "canSymlink.0", 0644, "a"); /* Note: Cygwin has its own symlink() emulation that does not * use the Win32 CreateSymbolicLink() function. */ #if defined(_WIN32) && !defined(__CYGWIN__) value = my_CreateSymbolicLinkA("canSymlink.1", "canSymlink.0", 0) && is_symlink(__FILE__, __LINE__, "canSymlink.1", "canSymlink.0"); #elif HAVE_SYMLINK value = (0 == symlink("canSymlink.0", "canSymlink.1")) && is_symlink(__FILE__, __LINE__, "canSymlink.1","canSymlink.0"); #endif return (value); } /* * Can this platform run the gzip program? */ /* Platform-dependent options for hiding the output of a subcommand. */ #if defined(_WIN32) && !defined(__CYGWIN__) static const char *redirectArgs = ">NUL 2>NUL"; /* Win32 cmd.exe */ #else static const char *redirectArgs = ">/dev/null 2>/dev/null"; /* POSIX 'sh' */ #endif int canGzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the gunzip program? */ int canGunzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gunzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Sleep as needed; useful for verifying disk timestamp changes by * ensuring that the wall-clock time has actually changed before we * go back to re-read something from disk. */ void sleepUntilAfter(time_t t) { while (t >= time(NULL)) #if defined(_WIN32) && !defined(__CYGWIN__) Sleep(500); #else sleep(1); #endif } /* * Call standard system() call, but build up the command line using * sprintf() conventions. */ int systemf(const char *fmt, ...) { char buff[8192]; va_list ap; int r; va_start(ap, fmt); vsprintf(buff, fmt, ap); if (verbosity > VERBOSITY_FULL) logprintf("Cmd: %s\n", buff); r = system(buff); va_end(ap); return (r); } /* * Slurp a file into memory for ease of comparison and testing. * Returns size of file in 'sizep' if non-NULL, null-terminates * data in memory for ease of use. */ char * slurpfile(size_t * sizep, const char *fmt, ...) { char filename[8192]; struct stat st; va_list ap; char *p; ssize_t bytes_read; FILE *f; int r; va_start(ap, fmt); vsprintf(filename, fmt, ap); va_end(ap); f = fopen(filename, "rb"); if (f == NULL) { /* Note: No error; non-existent file is okay here. */ return (NULL); } r = fstat(fileno(f), &st); if (r != 0) { logprintf("Can't stat file %s\n", filename); fclose(f); return (NULL); } p = malloc((size_t)st.st_size + 1); if (p == NULL) { logprintf("Can't allocate %ld bytes of memory to read file %s\n", (long int)st.st_size, filename); fclose(f); return (NULL); } bytes_read = fread(p, 1, (size_t)st.st_size, f); if (bytes_read < st.st_size) { logprintf("Can't read file %s\n", filename); fclose(f); free(p); return (NULL); } p[st.st_size] = '\0'; if (sizep != NULL) *sizep = (size_t)st.st_size; fclose(f); return (p); } /* Read a uuencoded file from the reference directory, decode, and * write the result into the current directory. */ #define UUDECODE(c) (((c) - 0x20) & 0x3f) void extract_reference_file(const char *name) { char buff[1024]; FILE *in, *out; sprintf(buff, "%s/%s.uu", refdir, name); in = fopen(buff, "r"); failure("Couldn't open reference file %s", buff); assert(in != NULL); if (in == NULL) return; /* Read up to and including the 'begin' line. */ for (;;) { if (fgets(buff, sizeof(buff), in) == NULL) { /* TODO: This is a failure. */ return; } if (memcmp(buff, "begin ", 6) == 0) break; } /* Now, decode the rest and write it. */ /* Not a lot of error checking here; the input better be right. */ out = fopen(name, "wb"); while (fgets(buff, sizeof(buff), in) != NULL) { char *p = buff; int bytes; if (memcmp(buff, "end", 3) == 0) break; bytes = UUDECODE(*p++); while (bytes > 0) { int n = 0; /* Write out 1-3 bytes from that. */ if (bytes > 0) { n = UUDECODE(*p++) << 18; n |= UUDECODE(*p++) << 12; fputc(n >> 16, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++) << 6; fputc((n >> 8) & 0xFF, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++); fputc(n & 0xFF, out); --bytes; } } } fclose(out); fclose(in); } int is_LargeInode(const char *file) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; r = my_GetFileInformationByName(file, &bhfi); if (r != 0) return (0); return (bhfi.nFileIndexHigh & 0x0000FFFFUL); #else struct stat st; int64_t ino; if (stat(file, &st) < 0) return (0); ino = (int64_t)st.st_ino; return (ino > 0xffffffff); #endif } /* * * TEST management * */ /* * "list.h" is simply created by "grep DEFINE_TEST test_*.c"; it has * a line like * DEFINE_TEST(test_function) * for each test. */ /* Use "list.h" to declare all of the test functions. */ #undef DEFINE_TEST #define DEFINE_TEST(name) void name(void); #include "list.h" /* Use "list.h" to create a list of all tests (functions and names). */ #undef DEFINE_TEST #define DEFINE_TEST(n) { n, #n, 0 }, struct { void (*func)(void); const char *name; int failures; } tests[] = { #include "list.h" }; /* * Summarize repeated failures in the just-completed test. */ static void test_summarize(int failed) { unsigned int i; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: printf(failed ? "E" : "."); fflush(stdout); break; case VERBOSITY_PASSFAIL: printf(failed ? "FAIL\n" : "ok\n"); break; } log_console = (verbosity == VERBOSITY_LIGHT_REPORT); for (i = 0; i < sizeof(failed_lines)/sizeof(failed_lines[0]); i++) { if (failed_lines[i].count > 1 && !failed_lines[i].skip) logprintf("%s:%d: Summary: Failed %d times\n", failed_filename, i, failed_lines[i].count); } /* Clear the failure history for the next file. */ failed_filename = NULL; memset(failed_lines, 0, sizeof(failed_lines)); } /* * Actually run a single test, with appropriate setup and cleanup. */ static int test_run(int i, const char *tmpdir) { char workdir[1024]; char logfilename[64]; int failures_before = failures; int oldumask; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: /* No per-test reports at all */ break; case VERBOSITY_PASSFAIL: /* rest of line will include ok/FAIL marker */ printf("%3d: %-50s", i, tests[i].name); fflush(stdout); break; default: /* Title of test, details will follow */ printf("%3d: %s\n", i, tests[i].name); } /* Chdir to the top-level work directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Can't chdir to top work dir %s\n", tmpdir); exit(1); } /* Create a log file for this test. */ sprintf(logfilename, "%s.log", tests[i].name); logfile = fopen(logfilename, "w"); fprintf(logfile, "%s\n\n", tests[i].name); /* Chdir() to a work dir for this specific test. */ snprintf(workdir, sizeof(workdir), "%s/%s", tmpdir, tests[i].name); testworkdir = workdir; if (!assertMakeDir(testworkdir, 0755) || !assertChdir(testworkdir)) { fprintf(stderr, "ERROR: Can't chdir to work dir %s\n", testworkdir); exit(1); } /* Explicitly reset the locale before each test. */ setlocale(LC_ALL, "C"); /* Record the umask before we run the test. */ umask(oldumask = umask(0)); /* * Run the actual test. */ (*tests[i].func)(); /* * Clean up and report afterwards. */ testworkdir = NULL; /* Restore umask */ umask(oldumask); /* Reset locale. */ setlocale(LC_ALL, "C"); /* Reset directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Couldn't chdir to temp dir %s\n", tmpdir); exit(1); } /* Report per-test summaries. */ tests[i].failures = failures - failures_before; test_summarize(tests[i].failures); /* Close the per-test log file. */ fclose(logfile); logfile = NULL; /* If there were no failures, we can remove the work dir and logfile. */ if (tests[i].failures == 0) { if (!keep_temp_files && assertChdir(tmpdir)) { #if defined(_WIN32) && !defined(__CYGWIN__) /* Make sure not to leave empty directories. * Sometimes a processing of closing files used by tests * is not done, then rmdir will be failed and it will * leave a empty test directory. So we should wait a few * seconds and retry rmdir. */ int r, t; for (t = 0; t < 10; t++) { if (t > 0) Sleep(1000); r = systemf("rmdir /S /Q %s", tests[i].name); if (r == 0) break; } systemf("del %s", logfilename); #else systemf("rm -rf %s", tests[i].name); systemf("rm %s", logfilename); #endif } } /* Return appropriate status. */ return (tests[i].failures); } /* * * * MAIN and support routines. * * */ static void usage(const char *program) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i; printf("Usage: %s [options] ...\n", program); printf("Default is to run all tests.\n"); printf("Otherwise, specify the numbers of the tests you wish to run.\n"); printf("Options:\n"); printf(" -d Dump core after any failure, for debugging.\n"); printf(" -k Keep all temp files.\n"); printf(" Default: temp files for successful tests deleted.\n"); #ifdef PROGRAM printf(" -p Path to executable to be tested.\n"); printf(" Default: path taken from " ENVBASE " environment variable.\n"); #endif printf(" -q Quiet.\n"); printf(" -r Path to dir containing reference files.\n"); printf(" Default: Current directory.\n"); printf(" -u Keep running specifies tests until one fails.\n"); printf(" -v Verbose.\n"); printf("Available tests:\n"); for (i = 0; i < limit; i++) printf(" %d: %s\n", i, tests[i].name); exit(1); } static char * get_refdir(const char *d) { char tried[512] = { '\0' }; char buff[128]; char *pwd, *p; /* If a dir was specified, try that */ if (d != NULL) { pwd = NULL; snprintf(buff, sizeof(buff), "%s", d); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); goto failure; } /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; /* Look for a known file. */ snprintf(buff, sizeof(buff), "%s", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(LIBRARY) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, LIBRARY); #else snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM); #endif p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(PROGRAM_ALIAS) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM_ALIAS); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #endif if (memcmp(pwd, "/usr/obj", 8) == 0) { snprintf(buff, sizeof(buff), "%s", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); } failure: printf("Unable to locate known reference file %s\n", KNOWNREF); printf(" Checked following directories:\n%s\n", tried); #if defined(_WIN32) && !defined(__CYGWIN__) && defined(_DEBUG) DebugBreak(); #endif exit(1); success: free(p); free(pwd); return strdup(buff); } int main(int argc, char **argv) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i = 0, j = 0, start, end, tests_run = 0, tests_failed = 0, option; time_t now; char *refdir_alloc = NULL; const char *progname; char **saved_argv; const char *tmp, *option_arg, *p; char tmpdir[256], *pwd, *testprogdir, *tmp2 = NULL; char tmpdir_timestamp[256]; (void)argc; /* UNUSED */ /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; #if defined(HAVE__CrtSetReportMode) /* To stop to run the default invalid parameter handler. */ _set_invalid_parameter_handler(invalid_parameter_handler); /* Disable annoying assertion message box. */ _CrtSetReportMode(_CRT_ASSERT, 0); #endif /* * Name of this program, used to build root of our temp directory * tree. */ progname = p = argv[0]; if ((testprogdir = (char *)malloc(strlen(progname) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir, progname); while (*p != '\0') { /* Support \ or / dir separators for Windows compat. */ if (*p == '/' || *p == '\\') { progname = p + 1; i = j; } ++p; j++; } testprogdir[i] = '\0'; if (testprogdir[0] != '/') { /* Fixup path for relative directories. */ if ((testprogdir = (char *)realloc(testprogdir, strlen(pwd) + 1 + strlen(testprogdir) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir + strlen(pwd) + 1, testprogdir); strcpy(testprogdir, pwd); testprogdir[strlen(pwd)] = '/'; } #ifdef PROGRAM /* Get the target program from environment, if available. */ testprogfile = getenv(ENVBASE); #endif if (getenv("TMPDIR") != NULL) tmp = getenv("TMPDIR"); else if (getenv("TMP") != NULL) tmp = getenv("TMP"); else if (getenv("TEMP") != NULL) tmp = getenv("TEMP"); else if (getenv("TEMPDIR") != NULL) tmp = getenv("TEMPDIR"); else tmp = "/tmp"; /* Allow -d to be controlled through the environment. */ if (getenv(ENVBASE "_DEBUG") != NULL) dump_on_failure = 1; /* Get the directory holding test files from environment. */ refdir = getenv(ENVBASE "_TEST_FILES"); /* * Parse options, without using getopt(), which isn't available * on all platforms. */ ++argv; /* Skip program name */ while (*argv != NULL) { if (**argv != '-') break; p = *argv++; ++p; /* Skip '-' */ while (*p != '\0') { option = *p++; option_arg = NULL; /* If 'opt' takes an argument, parse that. */ if (option == 'p' || option == 'r') { if (*p != '\0') option_arg = p; else if (*argv == NULL) { fprintf(stderr, "Option -%c requires argument.\n", option); usage(progname); } else option_arg = *argv++; p = ""; /* End of this option word. */ } /* Now, handle the option. */ switch (option) { case 'd': dump_on_failure = 1; break; case 'k': keep_temp_files = 1; break; case 'p': #ifdef PROGRAM testprogfile = option_arg; #else fprintf(stderr, "-p option not permitted\n"); usage(progname); #endif break; case 'q': verbosity--; break; case 'r': refdir = option_arg; break; case 'u': until_failure++; break; case 'v': verbosity++; break; default: fprintf(stderr, "Unrecognized option '%c'\n", option); usage(progname); } } } /* * Sanity-check that our options make sense. */ #ifdef PROGRAM if (testprogfile == NULL) { if ((tmp2 = (char *)malloc(strlen(testprogdir) + 1 + strlen(PROGRAM) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(tmp2, testprogdir); strcat(tmp2, "/"); strcat(tmp2, PROGRAM); testprogfile = tmp2; } { char *testprg; #if defined(_WIN32) && !defined(__CYGWIN__) /* Command.com sometimes rejects '/' separators. */ testprg = strdup(testprogfile); for (i = 0; testprg[i] != '\0'; i++) { if (testprg[i] == '/') testprg[i] = '\\'; } testprogfile = testprg; #endif /* Quote the name that gets put into shell command lines. */ testprg = malloc(strlen(testprogfile) + 3); strcpy(testprg, "\""); strcat(testprg, testprogfile); strcat(testprg, "\""); testprog = testprg; } #endif #if !defined(_WIN32) && defined(SIGPIPE) { /* Ignore SIGPIPE signals */ struct sigaction sa; sa.sa_handler = SIG_IGN; sigemptyset(&sa.sa_mask); sa.sa_flags = 0; sigaction(SIGPIPE, &sa, NULL); } #endif /* * Create a temp directory for the following tests. * Include the time the tests started as part of the name, * to make it easier to track the results of multiple tests. */ now = time(NULL); for (i = 0; ; i++) { strftime(tmpdir_timestamp, sizeof(tmpdir_timestamp), "%Y-%m-%dT%H.%M.%S", localtime(&now)); sprintf(tmpdir, "%s/%s.%s-%03d", tmp, progname, tmpdir_timestamp, i); if (assertMakeDir(tmpdir,0755)) break; if (i >= 999) { fprintf(stderr, "ERROR: Unable to create temp directory %s\n", tmpdir); exit(1); } } /* * If the user didn't specify a directory for locating * reference files, try to find the reference files in * the "usual places." */ refdir = refdir_alloc = get_refdir(refdir); /* * Banner with basic information. */ printf("\n"); printf("If tests fail or crash, details will be in:\n"); printf(" %s\n", tmpdir); printf("\n"); if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("Reference files will be read from: %s\n", refdir); #ifdef PROGRAM printf("Running tests on: %s\n", testprog); #endif printf("Exercising: "); fflush(stdout); printf("%s\n", EXTRA_VERSION); } else { printf("Running "); fflush(stdout); } /* * Run some or all of the individual tests. */ saved_argv = argv; do { argv = saved_argv; if (*argv == NULL) { /* Default: Run all tests. */ for (i = 0; i < limit; i++) { tests_run++; if (test_run(i, tmpdir)) { tests_failed++; if (until_failure) goto finish; } } } else { while (*(argv) != NULL) { if (**argv >= '0' && **argv <= '9') { char *vp = *argv; start = 0; while (*vp >= '0' && *vp <= '9') { start *= 10; start += *vp - '0'; ++vp; } if (*vp == '\0') { end = start; } else if (*vp == '-') { ++vp; if (*vp == '\0') { end = limit - 1; } else { end = 0; while (*vp >= '0' && *vp <= '9') { end *= 10; end += *vp - '0'; ++vp; } } } else { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } if (start < 0 || end >= limit || start > end) { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } } else { for (start = 0; start < limit; ++start) { if (strcmp(*argv, tests[start].name) == 0) break; } end = start; if (start >= limit) { printf("*** INVALID Test ``%s''\n", *argv); free(refdir_alloc); usage(progname); /* usage() never returns */ } } while (start <= end) { tests_run++; if (test_run(start, tmpdir)) { tests_failed++; if (until_failure) goto finish; } ++start; } argv++; } } } while (until_failure); finish: /* Must be freed after all tests run */ free(tmp2); free(testprogdir); free(pwd); /* * Report summary statistics. */ if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("\n"); printf("Totals:\n"); printf(" Tests run: %8d\n", tests_run); printf(" Tests failed: %8d\n", tests_failed); printf(" Assertions checked:%8d\n", assertions); printf(" Assertions failed: %8d\n", failures); printf(" Skips reported: %8d\n", skips); } if (failures) { printf("\n"); printf("Failing tests:\n"); for (i = 0; i < limit; ++i) { if (tests[i].failures) printf(" %d: %s (%d failures)\n", i, tests[i].name, tests[i].failures); } printf("\n"); printf("Details for failing tests: %s\n", tmpdir); printf("\n"); } else { if (verbosity == VERBOSITY_SUMMARY_ONLY) printf("\n"); printf("%d tests passed, no failures\n", tests_run); } free(refdir_alloc); /* If the final tmpdir is empty, we can remove it. */ /* This should be the usual case when all tests succeed. */ assertChdir(".."); rmdir(tmpdir); return (tests_failed ? 1 : 0); } Index: vendor/libarchive/dist/libarchive/test/test_acl_pax.c =================================================================== --- vendor/libarchive/dist/libarchive/test/test_acl_pax.c (revision 231296) +++ vendor/libarchive/dist/libarchive/test/test_acl_pax.c (revision 231297) @@ -1,282 +1,282 @@ /*- * Copyright (c) 2003-2007 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" __FBSDID("$FreeBSD: head/lib/libarchive/test/test_acl_pax.c 201247 2009-12-30 05:59:21Z kientzle $"); /* * Exercise the system-independent portion of the ACL support. * Check that pax archive can save and restore ACL data. * * This should work on all systems, regardless of whether local * filesystems support ACLs or not. */ static unsigned char buff[16384]; struct acl_t { int type; /* Type of ACL: "access" or "default" */ int permset; /* Permissions for this class of users. */ int tag; /* Owner, User, Owning group, group, other, etc. */ int qual; /* GID or UID of user/group, depending on tag. */ const char *name; /* Name of user/group, depending on tag. */ }; static struct acl_t acls0[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_USER_OBJ, 0, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, 0, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE, ARCHIVE_ENTRY_ACL_OTHER, 0, "" }, }; static struct acl_t acls1[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_USER_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 77, "user77" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE, ARCHIVE_ENTRY_ACL_OTHER, -1, "" }, }; static struct acl_t acls2[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE | ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 77, "user77" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 0, ARCHIVE_ENTRY_ACL_USER, 78, "user78" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 0007, ARCHIVE_ENTRY_ACL_GROUP, 78, "group78" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE | ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_OTHER, -1, "" }, }; static void set_acls(struct archive_entry *ae, struct acl_t *acls, int n) { int i; archive_entry_acl_clear(ae); for (i = 0; i < n; i++) { archive_entry_acl_add_entry(ae, acls[i].type, acls[i].permset, acls[i].tag, acls[i].qual, acls[i].name); } } static int acl_match(struct acl_t *acl, int type, int permset, int tag, int qual, const char *name) { if (type != acl->type) return (0); if (permset != acl->permset) return (0); if (tag != acl->tag) return (0); if (tag == ARCHIVE_ENTRY_ACL_USER_OBJ) return (1); if (tag == ARCHIVE_ENTRY_ACL_GROUP_OBJ) return (1); if (tag == ARCHIVE_ENTRY_ACL_OTHER) return (1); if (qual != acl->qual) return (0); if (name == NULL) return (acl->name == NULL || acl->name[0] == '\0'); if (acl->name == NULL) return (name == NULL || name[0] == '\0'); return (0 == strcmp(name, acl->name)); } static void compare_acls(struct archive_entry *ae, struct acl_t *acls, int n, int mode) { int *marker = malloc(sizeof(marker[0]) * n); int i; int r; int type, permset, tag, qual; int matched; const char *name; for (i = 0; i < n; i++) marker[i] = i; while (0 == (r = archive_entry_acl_next(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, &type, &permset, &tag, &qual, &name))) { for (i = 0, matched = 0; i < n && !matched; i++) { if (acl_match(&acls[marker[i]], type, permset, tag, qual, name)) { /* We found a match; remove it. */ marker[i] = marker[n - 1]; n--; matched = 1; } } if (tag == ARCHIVE_ENTRY_ACL_USER_OBJ) { if (!matched) printf("No match for user_obj perm\n"); failure("USER_OBJ permset (%02o) != user mode (%02o)", permset, 07 & (mode >> 6)); assert((permset << 6) == (mode & 0700)); } else if (tag == ARCHIVE_ENTRY_ACL_GROUP_OBJ) { if (!matched) printf("No match for group_obj perm\n"); failure("GROUP_OBJ permset %02o != group mode %02o", permset, 07 & (mode >> 3)); assert((permset << 3) == (mode & 0070)); } else if (tag == ARCHIVE_ENTRY_ACL_OTHER) { if (!matched) printf("No match for other perm\n"); failure("OTHER permset (%02o) != other mode (%02o)", permset, mode & 07); assert((permset << 0) == (mode & 0007)); } else { failure("Could not find match for ACL " "(type=%d,permset=%d,tag=%d,qual=%d,name=``%s'')", type, permset, tag, qual, name); assert(matched == 1); } } assertEqualInt(ARCHIVE_EOF, r); - assert((mode & 0777) == (archive_entry_mode(ae) & 0777)); + assert((mode_t)(mode & 0777) == (archive_entry_mode(ae) & 0777)); failure("Could not find match for ACL " "(type=%d,permset=%d,tag=%d,qual=%d,name=``%s'')", acls[marker[0]].type, acls[marker[0]].permset, acls[marker[0]].tag, acls[marker[0]].qual, acls[marker[0]].name); assert(n == 0); /* Number of ACLs not matched should == 0 */ free(marker); } DEFINE_TEST(test_acl_pax) { struct archive *a; struct archive_entry *ae; size_t used; FILE *f; void *reference; size_t reference_size; /* Write an archive to memory. */ assert(NULL != (a = archive_write_new())); assertA(0 == archive_write_set_format_pax(a)); assertA(0 == archive_write_set_compression_none(a)); assertA(0 == archive_write_set_bytes_per_block(a, 1)); assertA(0 == archive_write_set_bytes_in_last_block(a, 1)); assertA(0 == archive_write_open_memory(a, buff, sizeof(buff), &used)); /* Write a series of files to the archive with different ACL info. */ /* Create a simple archive_entry. */ assert((ae = archive_entry_new()) != NULL); archive_entry_set_pathname(ae, "file"); archive_entry_set_mode(ae, S_IFREG | 0777); /* Basic owner/owning group should just update mode bits. */ set_acls(ae, acls0, sizeof(acls0)/sizeof(acls0[0])); assertA(0 == archive_write_header(a, ae)); /* With any extended ACL entry, we should read back a full set. */ set_acls(ae, acls1, sizeof(acls1)/sizeof(acls1[0])); assertA(0 == archive_write_header(a, ae)); /* A more extensive set of ACLs. */ set_acls(ae, acls2, sizeof(acls2)/sizeof(acls2[0])); assertA(0 == archive_write_header(a, ae)); /* * Check that clearing ACLs gets rid of them all by repeating * the first test. */ set_acls(ae, acls0, sizeof(acls0)/sizeof(acls0[0])); assertA(0 == archive_write_header(a, ae)); archive_entry_free(ae); /* Close out the archive. */ assertEqualIntA(a, ARCHIVE_OK, archive_write_close(a)); assertEqualInt(ARCHIVE_OK, archive_write_free(a)); /* Write out the data we generated to a file for manual inspection. */ assert(NULL != (f = fopen("testout", "wb"))); assertEqualInt(used, (size_t)fwrite(buff, 1, (unsigned int)used, f)); fclose(f); /* Write out the reference data to a file for manual inspection. */ extract_reference_file("test_acl_pax.tar"); reference = slurpfile(&reference_size, "test_acl_pax.tar"); /* Assert that the generated data matches the built-in reference data.*/ failure("Generated pax archive does not match reference; compare 'testout' to 'test_acl_pax.tar' reference file."); assertEqualMem(buff, reference, reference_size); failure("Generated pax archive does not match reference; compare 'testout' to 'test_acl_pax.tar' reference file."); assertEqualInt((int)used, reference_size); free(reference); /* Read back each entry and check that the ACL data is right. */ assert(NULL != (a = archive_read_new())); assertA(0 == archive_read_support_format_all(a)); assertA(0 == archive_read_support_filter_all(a)); assertA(0 == archive_read_open_memory(a, buff, used)); /* First item has no ACLs */ assertA(0 == archive_read_next_header(a, &ae)); failure("Basic ACLs shouldn't be stored as extended ACLs"); assert(0 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* Second item has a few ACLs */ assertA(0 == archive_read_next_header(a, &ae)); failure("One extended ACL should flag all ACLs to be returned."); assert(4 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); compare_acls(ae, acls1, sizeof(acls1)/sizeof(acls1[0]), 0142); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* Third item has pretty extensive ACLs */ assertA(0 == archive_read_next_header(a, &ae)); assertEqualInt(6, archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); compare_acls(ae, acls2, sizeof(acls2)/sizeof(acls2[0]), 0543); failure("Basic ACLs should set mode to 0543, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0543); /* Fourth item has no ACLs */ assertA(0 == archive_read_next_header(a, &ae)); failure("Basic ACLs shouldn't be stored as extended ACLs"); assert(0 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* Close the archive. */ assertEqualIntA(a, ARCHIVE_OK, archive_read_close(a)); assertEqualInt(ARCHIVE_OK, archive_read_free(a)); } Index: vendor/libarchive/dist/libarchive/test/test_acl_posix1e.c =================================================================== --- vendor/libarchive/dist/libarchive/test/test_acl_posix1e.c (revision 231296) +++ vendor/libarchive/dist/libarchive/test/test_acl_posix1e.c (revision 231297) @@ -1,277 +1,277 @@ /*- * Copyright (c) 2003-2007 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" __FBSDID("$FreeBSD: src/lib/libarchive/test/test_acl_basic.c,v 1.6 2008/10/19 00:13:57 kientzle Exp $"); /* * Exercise the system-independent portion of the ACL support. * Check that archive_entry objects can save and restore POSIX.1e-style ACL data. * * This should work on all systems, regardless of whether local * filesystems support ACLs or not. */ struct acl_t { int type; /* Type of ACL: "access" or "default" */ int permset; /* Permissions for this class of users. */ int tag; /* Owner, User, Owning group, group, other, etc. */ int qual; /* GID or UID of user/group, depending on tag. */ const char *name; /* Name of user/group, depending on tag. */ }; static struct acl_t acls0[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_USER_OBJ, 0, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, 0, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE, ARCHIVE_ENTRY_ACL_OTHER, 0, "" }, }; static struct acl_t acls1[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_USER_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 77, "user77" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE, ARCHIVE_ENTRY_ACL_OTHER, -1, "" }, }; static struct acl_t acls2[] = { { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_EXECUTE | ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 77, "user77" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 0, ARCHIVE_ENTRY_ACL_USER, 78, "user78" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_GROUP_OBJ, -1, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, 0007, ARCHIVE_ENTRY_ACL_GROUP, 78, "group78" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_WRITE | ARCHIVE_ENTRY_ACL_EXECUTE, ARCHIVE_ENTRY_ACL_OTHER, -1, "" }, }; /* * NFS4 entry types; attempts to set these on top of POSIX.1e * attributes should fail. */ static struct acl_t acls_nfs4[] = { /* NFS4 types */ { ARCHIVE_ENTRY_ACL_TYPE_ALLOW, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 78, "" }, { ARCHIVE_ENTRY_ACL_TYPE_DENY, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 78, "" }, { ARCHIVE_ENTRY_ACL_TYPE_AUDIT, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 78, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ALARM, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_USER, 78, "" }, /* NFS4 tags */ { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ, ARCHIVE_ENTRY_ACL_EVERYONE, -1, "" }, /* NFS4 inheritance markers */ { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ | ARCHIVE_ENTRY_ACL_ENTRY_FILE_INHERIT, ARCHIVE_ENTRY_ACL_USER, 79, "" }, { ARCHIVE_ENTRY_ACL_TYPE_ACCESS, ARCHIVE_ENTRY_ACL_READ | ARCHIVE_ENTRY_ACL_ENTRY_FILE_INHERIT, ARCHIVE_ENTRY_ACL_USER_OBJ, -1, "" }, }; static void set_acls(struct archive_entry *ae, struct acl_t *acls, int n) { int i; archive_entry_acl_clear(ae); for (i = 0; i < n; i++) { archive_entry_acl_add_entry(ae, acls[i].type, acls[i].permset, acls[i].tag, acls[i].qual, acls[i].name); } } static int acl_match(struct acl_t *acl, int type, int permset, int tag, int qual, const char *name) { if (type != acl->type) return (0); if (permset != acl->permset) return (0); if (tag != acl->tag) return (0); if (tag == ARCHIVE_ENTRY_ACL_USER_OBJ) return (1); if (tag == ARCHIVE_ENTRY_ACL_GROUP_OBJ) return (1); if (tag == ARCHIVE_ENTRY_ACL_OTHER) return (1); if (qual != acl->qual) return (0); if (name == NULL) { if (acl->name == NULL || acl->name[0] == '\0') return (1); } if (acl->name == NULL) { if (name[0] == '\0') return (1); } return (0 == strcmp(name, acl->name)); } static void compare_acls(struct archive_entry *ae, struct acl_t *acls, int n, int mode) { int *marker = malloc(sizeof(marker[0]) * n); int i; int r; int type, permset, tag, qual; int matched; const char *name; for (i = 0; i < n; i++) marker[i] = i; while (0 == (r = archive_entry_acl_next(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS, &type, &permset, &tag, &qual, &name))) { for (i = 0, matched = 0; i < n && !matched; i++) { if (acl_match(&acls[marker[i]], type, permset, tag, qual, name)) { /* We found a match; remove it. */ marker[i] = marker[n - 1]; n--; matched = 1; } } if (tag == ARCHIVE_ENTRY_ACL_USER_OBJ) { if (!matched) printf("No match for user_obj perm\n"); failure("USER_OBJ permset (%02o) != user mode (%02o)", permset, 07 & (mode >> 6)); assert((permset << 6) == (mode & 0700)); } else if (tag == ARCHIVE_ENTRY_ACL_GROUP_OBJ) { if (!matched) printf("No match for group_obj perm\n"); failure("GROUP_OBJ permset %02o != group mode %02o", permset, 07 & (mode >> 3)); assert((permset << 3) == (mode & 0070)); } else if (tag == ARCHIVE_ENTRY_ACL_OTHER) { if (!matched) printf("No match for other perm\n"); failure("OTHER permset (%02o) != other mode (%02o)", permset, mode & 07); assert((permset << 0) == (mode & 0007)); } else { failure("Could not find match for ACL " "(type=%d,permset=%d,tag=%d,qual=%d,name=``%s'')", type, permset, tag, qual, name); assert(matched == 1); } } assertEqualInt(ARCHIVE_EOF, r); - assert((mode & 0777) == (archive_entry_mode(ae) & 0777)); + assert((mode_t)(mode & 0777) == (archive_entry_mode(ae) & 0777)); failure("Could not find match for ACL " "(type=%d,permset=%d,tag=%d,qual=%d,name=``%s'')", acls[marker[0]].type, acls[marker[0]].permset, acls[marker[0]].tag, acls[marker[0]].qual, acls[marker[0]].name); assert(n == 0); /* Number of ACLs not matched should == 0 */ free(marker); } DEFINE_TEST(test_acl_posix1e) { struct archive_entry *ae; int i; /* Create a simple archive_entry. */ assert((ae = archive_entry_new()) != NULL); archive_entry_set_pathname(ae, "file"); archive_entry_set_mode(ae, S_IFREG | 0777); /* Basic owner/owning group should just update mode bits. */ /* * Note: This features of libarchive's ACL implementation * shouldn't be relied on and should probably be removed. It * was done to identify trivial ACLs so we could avoid * triggering unnecessary extensions. It's better to identify * trivial ACLs at the point they are being read from disk. */ set_acls(ae, acls0, sizeof(acls0)/sizeof(acls0[0])); failure("Basic ACLs shouldn't be stored as extended ACLs"); assert(0 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* With any extended ACL entry, we should read back a full set. */ set_acls(ae, acls1, sizeof(acls1)/sizeof(acls1[0])); failure("One extended ACL should flag all ACLs to be returned."); assert(4 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); compare_acls(ae, acls1, sizeof(acls1)/sizeof(acls1[0]), 0142); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* A more extensive set of ACLs. */ set_acls(ae, acls2, sizeof(acls2)/sizeof(acls2[0])); assertEqualInt(6, archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); compare_acls(ae, acls2, sizeof(acls2)/sizeof(acls2[0]), 0543); failure("Basic ACLs should set mode to 0543, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0543); /* * Check that clearing ACLs gets rid of them all by repeating * the first test. */ set_acls(ae, acls0, sizeof(acls0)/sizeof(acls0[0])); failure("Basic ACLs shouldn't be stored as extended ACLs"); assert(0 == archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); failure("Basic ACLs should set mode to 0142, not %04o", archive_entry_mode(ae)&0777); assert((archive_entry_mode(ae) & 0777) == 0142); /* * Different types of malformed ACL entries that should * fail when added to existing POSIX.1e ACLs. */ set_acls(ae, acls2, sizeof(acls2)/sizeof(acls2[0])); for (i = 0; i < (int)(sizeof(acls_nfs4)/sizeof(acls_nfs4[0])); ++i) { struct acl_t *p = &acls_nfs4[i]; failure("Malformed ACL test #%d", i); assertEqualInt(ARCHIVE_FAILED, archive_entry_acl_add_entry(ae, p->type, p->permset, p->tag, p->qual, p->name)); assertEqualInt(6, archive_entry_acl_reset(ae, ARCHIVE_ENTRY_ACL_TYPE_ACCESS)); } archive_entry_free(ae); } Index: vendor/libarchive/dist/libarchive/test/test_sparse_basic.c =================================================================== --- vendor/libarchive/dist/libarchive/test/test_sparse_basic.c (revision 231296) +++ vendor/libarchive/dist/libarchive/test/test_sparse_basic.c (revision 231297) @@ -1,462 +1,464 @@ /*- * Copyright (c) 2010 Michihiro NAKAJIMA * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" __FBSDID("$FreeBSD$"); #ifdef HAVE_SYS_IOCTL_H #include #endif #ifdef HAVE_SYS_PARAM_H #include #endif #ifdef HAVE_SYS_UTSNAME_H #include #endif #ifdef HAVE_FCNTL_H #include #endif #ifdef HAVE_LIMITS_H #include #endif #ifdef HAVE_UNISTD_H #include #endif #ifdef HAVE_LINUX_FIEMAP_H #include #endif #ifdef HAVE_LINUX_FS_H #include #endif /* * NOTE: On FreeBSD and Solaris, this test needs ZFS. * You may should perfom this test as * 'TMPDIR= libarchive_test'. */ struct sparse { enum { DATA, HOLE, END } type; size_t size; }; static void create_sparse_file(const char *, const struct sparse *); #if defined(_WIN32) && !defined(__CYGWIN__) #include /* * Create a sparse file on Windows. */ #if !defined(PATH_MAX) #define PATH_MAX MAX_PATH #endif #if !defined(__BORLANDC__) #define getcwd _getcwd #endif static int is_sparse_supported(const char *path) { char root[MAX_PATH+1]; char vol[MAX_PATH+1]; char sys[MAX_PATH+1]; DWORD flags; BOOL r; strncpy(root, path, sizeof(root)-1); if (((root[0] >= 'c' && root[0] <= 'z') || (root[0] >= 'C' && root[0] <= 'Z')) && root[1] == ':' && (root[2] == '\\' || root[2] == '/')) root[3] = '\0'; else return (0); assertEqualInt((r = GetVolumeInformation(root, vol, sizeof(vol), NULL, NULL, &flags, sys, sizeof(sys))), 1); return (r != 0 && (flags & FILE_SUPPORTS_SPARSE_FILES) != 0); } static void create_sparse_file(const char *path, const struct sparse *s) { char buff[1024]; HANDLE handle; DWORD dmy; memset(buff, ' ', sizeof(buff)); handle = CreateFileA(path, GENERIC_WRITE, 0, NULL, CREATE_NEW, FILE_ATTRIBUTE_NORMAL, NULL); assert(handle != INVALID_HANDLE_VALUE); assert(DeviceIoControl(handle, FSCTL_SET_SPARSE, NULL, 0, NULL, 0, &dmy, NULL) != 0); while (s->type != END) { if (s->type == HOLE) { LARGE_INTEGER distance; distance.QuadPart = s->size; assert(SetFilePointerEx(handle, distance, NULL, FILE_CURRENT) != 0); } else { DWORD w, wr, size; size = s->size; while (size) { if (size > sizeof(buff)) w = sizeof(buff); else w = size; assert(WriteFile(handle, buff, w, &wr, NULL) != 0); size -= wr; } } s++; } assertEqualInt(CloseHandle(handle), 1); } #else #if defined(_PC_MIN_HOLE_SIZE) /* * FreeBSD and Solaris can detect 'hole' of a sparse file * through lseek(HOLE) on ZFS. (UFS does not support yet) */ static int is_sparse_supported(const char *path) { return (pathconf(path, _PC_MIN_HOLE_SIZE) > 0); } #elif defined(__linux__)&& defined(HAVE_LINUX_FIEMAP_H) /* * FIEMAP, which can detect 'hole' of a sparse file, has * been supported from 2.6.28 */ static int is_sparse_supported(const char *path) { const struct sparse sparse_file[] = { /* This hole size is too small to create a sparse * files for almost filesystem. */ { HOLE, 1024 }, { DATA, 10240 }, { END, 0 } }; struct utsname ut; char *p, *e; long d; int fd, r; struct fiemap *fm; char buff[1024]; const char *testfile = "can_sparse"; + (void)path; /* UNUSED */ memset(&ut, 0, sizeof(ut)); assertEqualInt(uname(&ut), 0); p = ut.release; d = strtol(p, &e, 10); if (d < 2 || *e != '.') return (0); if (d == 2) { p = e + 1; d = strtol(p, &e, 10); if (d < 6 || *e != '.') return (0); p = e + 1; d = strtol(p, NULL, 10); if (d < 28) return (0); } create_sparse_file(testfile, sparse_file); fd = open(testfile, O_RDWR); if (fd < 0) return (0); fm = (struct fiemap *)buff; fm->fm_start = 0; fm->fm_length = ~0ULL;; fm->fm_flags = FIEMAP_FLAG_SYNC; fm->fm_extent_count = (sizeof(buff) - sizeof(*fm))/ sizeof(struct fiemap_extent); r = ioctl(fd, FS_IOC_FIEMAP, fm); if (r < 0 && (errno == ENOTTY || errno == EOPNOTSUPP)) return (0);/* Not supported. */ close(fd); unlink(testfile); return (1); } #else /* * Other system may do not have the API such as lseek(HOLE), * which detect 'hole' of a sparse file. */ static int is_sparse_supported(const char *path) { + (void)path; /* UNUSED */ return (0); } #endif /* * Create a sparse file on POSIX like system. */ static void create_sparse_file(const char *path, const struct sparse *s) { char buff[1024]; int fd; memset(buff, ' ', sizeof(buff)); assert((fd = open(path, O_CREAT | O_WRONLY, 0600)) != -1); while (s->type != END) { if (s->type == HOLE) { assert(lseek(fd, s->size, SEEK_CUR) != (off_t)-1); } else { size_t w, size; size = s->size; while (size) { if (size > sizeof(buff)) w = sizeof(buff); else w = size; assert(write(fd, buff, w) != (ssize_t)-1); size -= w; } } s++; } close(fd); } #endif /* * Sparse test with directory traversals. */ static void verify_sparse_file(struct archive *a, const char *path, const struct sparse *sparse, int blocks) { struct archive_entry *ae; const void *buff; size_t bytes_read; int64_t offset; int64_t total; int data_blocks, hole; create_sparse_file(path, sparse); assert((ae = archive_entry_new()) != NULL); assertEqualIntA(a, ARCHIVE_OK, archive_read_disk_open(a, path)); assertEqualIntA(a, ARCHIVE_OK, archive_read_next_header2(a, ae)); /* Verify the number of holes only, not its offset nor its * length because those alignments are deeply dependence on * its filesystem. */ assertEqualInt(blocks, archive_entry_sparse_count(ae)); total = 0; data_blocks = 0; hole = 0; while (ARCHIVE_OK == archive_read_data_block(a, &buff, &bytes_read, &offset)) { if (offset > total || offset == 0) { if (offset > total) hole = 1; data_blocks++; } total = offset + bytes_read; } if (!hole && data_blocks == 1) data_blocks = 0;/* There are no holes */ assertEqualInt(blocks, data_blocks); assertEqualIntA(a, ARCHIVE_OK, archive_read_close(a)); archive_entry_free(ae); } #if defined(_WIN32) && !defined(__CYGWIN__) #define close _close #define open _open #endif /* * Sparse test without directory traversals. */ static void verify_sparse_file2(struct archive *a, const char *path, const struct sparse *sparse, int blocks, int preopen) { struct archive_entry *ae; int fd; (void)sparse; /* UNUSED */ assert((ae = archive_entry_new()) != NULL); archive_entry_set_pathname(ae, path); if (preopen) fd = open(path, O_RDONLY | O_BINARY); else fd = -1; assertEqualIntA(a, ARCHIVE_OK, archive_read_disk_entry_from_file(a, ae, fd, NULL)); if (fd >= 0) close(fd); /* Verify the number of holes only, not its offset nor its * length because those alignments are deeply dependence on * its filesystem. */ assertEqualInt(blocks, archive_entry_sparse_count(ae)); archive_entry_free(ae); } static void test_sparse_whole_file_data() { struct archive_entry *ae; int64_t offset; int i; assert((ae = archive_entry_new()) != NULL); archive_entry_set_size(ae, 1024*10); /* * Add sparse block data up to the file size. */ offset = 0; for (i = 0; i < 10; i++) { archive_entry_sparse_add_entry(ae, offset, 1024); offset += 1024; } failure("There should be no sparse"); assertEqualInt(0, archive_entry_sparse_count(ae)); archive_entry_free(ae); } DEFINE_TEST(test_sparse_basic) { char *cwd; struct archive *a; /* * The alignment of the hole of sparse files deeply depends * on filesystem. In my experience, sparse_file2 test with * 204800 bytes hole size did not pass on ZFS and the result * of that test seemed the size was too small, thus you should * keep a hole size more than 409600 bytes to pass this test * on all platform. */ const struct sparse sparse_file0[] = { { DATA, 1024 }, { HOLE, 2048000 }, { DATA, 2048 }, { HOLE, 2048000 }, { DATA, 4096 }, { HOLE, 20480000 }, { DATA, 8192 }, { HOLE, 204800000 }, { DATA, 1 }, { END, 0 } }; const struct sparse sparse_file1[] = { { HOLE, 409600 }, { DATA, 1 }, { HOLE, 409600 }, { DATA, 1 }, { HOLE, 409600 }, { END, 0 } }; const struct sparse sparse_file2[] = { { HOLE, 409600 * 1 }, { DATA, 1024 }, { HOLE, 409600 * 2 }, { DATA, 1024 }, { HOLE, 409600 * 3 }, { DATA, 1024 }, { HOLE, 409600 * 4 }, { DATA, 1024 }, { HOLE, 409600 * 5 }, { DATA, 1024 }, { HOLE, 409600 * 6 }, { DATA, 1024 }, { HOLE, 409600 * 7 }, { DATA, 1024 }, { HOLE, 409600 * 8 }, { DATA, 1024 }, { HOLE, 409600 * 9 }, { DATA, 1024 }, { HOLE, 409600 * 10}, { DATA, 1024 },/* 10 */ { HOLE, 409600 * 1 }, { DATA, 1024 * 1 }, { HOLE, 409600 * 2 }, { DATA, 1024 * 2 }, { HOLE, 409600 * 3 }, { DATA, 1024 * 3 }, { HOLE, 409600 * 4 }, { DATA, 1024 * 4 }, { HOLE, 409600 * 5 }, { DATA, 1024 * 5 }, { HOLE, 409600 * 6 }, { DATA, 1024 * 6 }, { HOLE, 409600 * 7 }, { DATA, 1024 * 7 }, { HOLE, 409600 * 8 }, { DATA, 1024 * 8 }, { HOLE, 409600 * 9 }, { DATA, 1024 * 9 }, { HOLE, 409600 * 10}, { DATA, 1024 * 10},/* 20 */ { END, 0 } }; const struct sparse sparse_file3[] = { /* This hole size is too small to create a sparse * files for almost filesystem. */ { HOLE, 1024 }, { DATA, 10240 }, { END, 0 } }; /* * Test for the case that sparse data indicates just the whole file * data. */ test_sparse_whole_file_data(); /* Check if the filesystem where CWD on can * report the number of the holes of a sparse file. */ #ifdef PATH_MAX cwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else cwd = getcwd(NULL, 0); #endif if (!assert(cwd != NULL)) return; if (!is_sparse_supported(cwd)) { free(cwd); skipping("This filesystem or platform do not support " "the reporting of the holes of a sparse file through " "API such as lseek(HOLE)"); return; } /* * Get sparse data through directory traversals. */ assert((a = archive_read_disk_new()) != NULL); verify_sparse_file(a, "file0", sparse_file0, 5); verify_sparse_file(a, "file1", sparse_file1, 2); verify_sparse_file(a, "file2", sparse_file2, 20); verify_sparse_file(a, "file3", sparse_file3, 0); assertEqualInt(ARCHIVE_OK, archive_read_free(a)); /* * Get sparse data through archive_read_disk_entry_from_file(). */ assert((a = archive_read_disk_new()) != NULL); verify_sparse_file2(a, "file0", sparse_file0, 5, 0); verify_sparse_file2(a, "file0", sparse_file0, 5, 1); assertEqualInt(ARCHIVE_OK, archive_read_free(a)); free(cwd); } Index: vendor/libarchive/dist/tar/test/main.c =================================================================== --- vendor/libarchive/dist/tar/test/main.c (revision 231296) +++ vendor/libarchive/dist/tar/test/main.c (revision 231297) @@ -1,2612 +1,2612 @@ /* * Copyright (c) 2003-2009 Tim Kientzle * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "test.h" #ifdef HAVE_SYS_TIME_H #include #endif #include #ifdef HAVE_ICONV_H #include #endif #include #include #ifdef HAVE_SIGNAL_H #include #endif #include #include /* * This same file is used pretty much verbatim for all test harnesses. * * The next few lines are the only differences. * TODO: Move this into a separate configuration header, have all test * suites share one copy of this file. */ __FBSDID("$FreeBSD: src/usr.bin/tar/test/main.c,v 1.6 2008/11/05 06:40:53 kientzle Exp $"); #define KNOWNREF "test_patterns_2.tar.uu" #define ENVBASE "BSDTAR" /* Prefix for environment variables. */ #define PROGRAM "bsdtar" /* Name of program being tested. */ #define PROGRAM_ALIAS "tar" /* Generic alias for program */ #undef LIBRARY /* Not testing a library. */ #undef EXTRA_DUMP /* How to dump extra data */ #undef EXTRA_ERRNO /* How to dump errno */ /* How to generate extra version info. */ #define EXTRA_VERSION (systemf("%s --version", testprog) ? "" : "") /* * * Windows support routines * * Note: Configuration is a tricky issue. Using HAVE_* feature macros * in the test harness is dangerous because they cover up * configuration errors. The classic example of this is omitting a * configure check. If libarchive and libarchive_test both look for * the same feature macro, such errors are hard to detect. Platform * macros (e.g., _WIN32 or __GNUC__) are a little better, but can * easily lead to very messy code. It's best to limit yourself * to only the most generic programming techniques in the test harness * and thus avoid conditionals altogether. Where that's not possible, * try to minimize conditionals by grouping platform-specific tests in * one place (e.g., test_acl_freebsd) or by adding new assert() * functions (e.g., assertMakeHardlink()) to cover up platform * differences. Platform-specific coding in libarchive_test is often * a symptom that some capability is missing from libarchive itself. */ #if defined(_WIN32) && !defined(__CYGWIN__) #include #include #ifndef F_OK #define F_OK (0) #endif #ifndef S_ISDIR #define S_ISDIR(m) ((m) & _S_IFDIR) #endif #ifndef S_ISREG #define S_ISREG(m) ((m) & _S_IFREG) #endif #if !defined(__BORLANDC__) #define access _access #undef chdir #define chdir _chdir #endif #ifndef fileno #define fileno _fileno #endif /*#define fstat _fstat64*/ #if !defined(__BORLANDC__) #define getcwd _getcwd #endif #define lstat stat /*#define lstat _stat64*/ /*#define stat _stat64*/ #define rmdir _rmdir #if !defined(__BORLANDC__) #define strdup _strdup #define umask _umask #endif #define int64_t __int64 #endif #if defined(HAVE__CrtSetReportMode) # include #endif #if defined(_WIN32) && !defined(__CYGWIN__) void *GetFunctionKernel32(const char *name) { static HINSTANCE lib; static int set; if (!set) { set = 1; lib = LoadLibrary("kernel32.dll"); } if (lib == NULL) { fprintf(stderr, "Can't load kernel32.dll?!\n"); exit(1); } return (void *)GetProcAddress(lib, name); } static int my_CreateSymbolicLinkA(const char *linkname, const char *target, int flags) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, DWORD); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateSymbolicLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, flags); } static int my_CreateHardLinkA(const char *linkname, const char *target) { static BOOLEAN (WINAPI *f)(LPCSTR, LPCSTR, LPSECURITY_ATTRIBUTES); static int set; if (!set) { set = 1; f = GetFunctionKernel32("CreateHardLinkA"); } return f == NULL ? 0 : (*f)(linkname, target, NULL); } int my_GetFileInformationByName(const char *path, BY_HANDLE_FILE_INFORMATION *bhfi) { HANDLE h; int r; memset(bhfi, 0, sizeof(*bhfi)); h = CreateFile(path, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) return (0); r = GetFileInformationByHandle(h, bhfi); CloseHandle(h); return (r); } #endif #if defined(HAVE__CrtSetReportMode) static void invalid_parameter_handler(const wchar_t * expression, const wchar_t * function, const wchar_t * file, unsigned int line, uintptr_t pReserved) { /* nop */ } #endif /* * * OPTIONS FLAGS * */ /* Enable core dump on failure. */ static int dump_on_failure = 0; /* Default is to remove temp dirs and log data for successful tests. */ static int keep_temp_files = 0; /* Default is to run the specified tests once and report errors. */ static int until_failure = 0; /* Default is to just report pass/fail for each test. */ static int verbosity = 0; #define VERBOSITY_SUMMARY_ONLY -1 /* -q */ #define VERBOSITY_PASSFAIL 0 /* Default */ #define VERBOSITY_LIGHT_REPORT 1 /* -v */ #define VERBOSITY_FULL 2 /* -vv */ /* A few places generate even more output for verbosity > VERBOSITY_FULL, * mostly for debugging the test harness itself. */ /* Cumulative count of assertion failures. */ static int failures = 0; /* Cumulative count of reported skips. */ static int skips = 0; /* Cumulative count of assertions checked. */ static int assertions = 0; /* Directory where uuencoded reference files can be found. */ static const char *refdir; /* * Report log information selectively to console and/or disk log. */ static int log_console = 0; static FILE *logfile; static void vlogprintf(const char *fmt, va_list ap) { #ifdef va_copy va_list lfap; va_copy(lfap, ap); #endif if (log_console) vfprintf(stdout, fmt, ap); if (logfile != NULL) #ifdef va_copy vfprintf(logfile, fmt, lfap); va_end(lfap); #else vfprintf(logfile, fmt, ap); #endif } static void logprintf(const char *fmt, ...) { va_list ap; va_start(ap, fmt); vlogprintf(fmt, ap); va_end(ap); } /* Set up a message to display only if next assertion fails. */ static char msgbuff[4096]; static const char *msg, *nextmsg; void failure(const char *fmt, ...) { va_list ap; if (fmt == NULL) { nextmsg = NULL; } else { va_start(ap, fmt); vsprintf(msgbuff, fmt, ap); va_end(ap); nextmsg = msgbuff; } } /* * Copy arguments into file-local variables. * This was added to permit vararg assert() functions without needing * variadic wrapper macros. Turns out that the vararg capability is almost * never used, so almost all of the vararg assertions can be simplified * by removing the vararg capability and reworking the wrapper macro to * pass __FILE__, __LINE__ directly into the function instead of using * this hook. I suspect this machinery is used so rarely that we * would be better off just removing it entirely. That would simplify * the code here noticeably. */ static const char *skipping_filename; static int skipping_line; void skipping_setup(const char *filename, int line) { skipping_filename = filename; skipping_line = line; } /* Called at the beginning of each assert() function. */ static void assertion_count(const char *file, int line) { (void)file; /* UNUSED */ (void)line; /* UNUSED */ ++assertions; /* Proper handling of "failure()" message. */ msg = nextmsg; nextmsg = NULL; /* Uncomment to print file:line after every assertion. * Verbose, but occasionally useful in tracking down crashes. */ /* printf("Checked %s:%d\n", file, line); */ } /* * For each test source file, we remember how many times each * assertion was reported. Cleared before each new test, * used by test_summarize(). */ static struct line { int count; int skip; } failed_lines[10000]; const char *failed_filename; /* Count this failure, setup up log destination and handle initial report. */ static void failure_start(const char *filename, int line, const char *fmt, ...) { va_list ap; /* Record another failure for this line. */ ++failures; failed_filename = filename; failed_lines[line].count++; /* Determine whether to log header to console. */ switch (verbosity) { case VERBOSITY_LIGHT_REPORT: log_console = (failed_lines[line].count < 2); break; default: log_console = (verbosity >= VERBOSITY_FULL); } /* Log file:line header for this failure */ va_start(ap, fmt); #if _MSC_VER logprintf("%s(%d): ", filename, line); #else logprintf("%s:%d: ", filename, line); #endif vlogprintf(fmt, ap); va_end(ap); logprintf("\n"); if (msg != NULL && msg[0] != '\0') { logprintf(" Description: %s\n", msg); msg = NULL; } /* Determine whether to log details to console. */ if (verbosity == VERBOSITY_LIGHT_REPORT) log_console = 0; } /* Complete reporting of failed tests. */ /* * The 'extra' hook here is used by libarchive to include libarchive * error messages with assertion failures. It could also be used * to add strerror() output, for example. Just define the EXTRA_DUMP() * macro appropriately. */ static void failure_finish(void *extra) { (void)extra; /* UNUSED (maybe) */ #ifdef EXTRA_DUMP if (extra != NULL) { logprintf(" errno: %d\n", EXTRA_ERRNO(extra)); logprintf(" detail: %s\n", EXTRA_DUMP(extra)); } #endif if (dump_on_failure) { fprintf(stderr, " *** forcing core dump so failure can be debugged ***\n"); abort(); exit(1); } } /* Inform user that we're skipping some checks. */ void test_skipping(const char *fmt, ...) { char buff[1024]; va_list ap; va_start(ap, fmt); vsprintf(buff, fmt, ap); va_end(ap); /* Use failure() message if set. */ msg = nextmsg; nextmsg = NULL; /* failure_start() isn't quite right, but is awfully convenient. */ failure_start(skipping_filename, skipping_line, "SKIPPING: %s", buff); --failures; /* Undo failures++ in failure_start() */ /* Don't failure_finish() here. */ /* Mark as skip, so doesn't count as failed test. */ failed_lines[skipping_line].skip = 1; ++skips; } /* * * ASSERTIONS * */ /* Generic assert() just displays the failed condition. */ int assertion_assert(const char *file, int line, int value, const char *condition, void *extra) { assertion_count(file, line); if (!value) { failure_start(file, line, "Assertion failed: %s", condition); failure_finish(extra); } return (value); } /* chdir() and report any errors */ int assertion_chdir(const char *file, int line, const char *pathname) { assertion_count(file, line); if (chdir(pathname) == 0) return (1); failure_start(file, line, "chdir(\"%s\")", pathname); failure_finish(NULL); return (0); } /* Verify two integers are equal. */ int assertion_equal_int(const char *file, int line, long long v1, const char *e1, long long v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e1, v1, v1, v1); logprintf(" %s=%lld (0x%llx, 0%llo)\n", e2, v2, v2, v2); failure_finish(extra); return (0); } /* * Utility to convert a single UTF-8 sequence. */ static int _utf8_to_unicode(uint32_t *pwc, const char *s, size_t n) { static const char utf8_count[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 00 - 0F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 10 - 1F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 20 - 2F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 30 - 3F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 40 - 4F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 50 - 5F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 60 - 6F */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,/* 70 - 7F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 80 - 8F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* 90 - 9F */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* A0 - AF */ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,/* B0 - BF */ 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* C0 - CF */ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,/* D0 - DF */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,/* E0 - EF */ 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 /* F0 - FF */ }; int ch; int cnt; uint32_t wc; *pwc = 0; /* Sanity check. */ if (n == 0) return (0); /* * Decode 1-4 bytes depending on the value of the first byte. */ ch = (unsigned char)*s; if (ch == 0) return (0); /* Standard: return 0 for end-of-string. */ cnt = utf8_count[ch]; /* Invalide sequence or there are not plenty bytes. */ if (n < (size_t)cnt) return (-1); /* Make a Unicode code point from a single UTF-8 sequence. */ switch (cnt) { case 1: /* 1 byte sequence. */ *pwc = ch & 0x7f; return (cnt); case 2: /* 2 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); *pwc = ((ch & 0x1f) << 6) | (s[1] & 0x3f); return (cnt); case 3: /* 3 bytes sequence. */ if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x0f) << 12) | ((s[1] & 0x3f) << 6) | (s[2] & 0x3f); if (wc < 0x800) return (-1);/* Overlong sequence. */ break; case 4: /* 4 bytes sequence. */ if (n < 4) return (-1); if ((s[1] & 0xc0) != 0x80) return (-1); if ((s[2] & 0xc0) != 0x80) return (-1); if ((s[3] & 0xc0) != 0x80) return (-1); wc = ((ch & 0x07) << 18) | ((s[1] & 0x3f) << 12) | ((s[2] & 0x3f) << 6) | (s[3] & 0x3f); if (wc < 0x10000) return (-1);/* Overlong sequence. */ break; default: return (-1); } /* The code point larger than 0x10FFFF is not leagal * Unicode values. */ if (wc > 0x10FFFF) return (-1); /* Correctly gets a Unicode, returns used bytes. */ *pwc = wc; return (cnt); } static void strdump(const char *e, const char *p, int ewidth, int utf8) { const char *q = p; logprintf(" %*s = ", ewidth, e); if (p == NULL) { logprintf("NULL\n"); return; } logprintf("\""); while (*p != '\0') { unsigned int c = 0xff & *p++; switch (c) { case '\a': printf("\a"); break; case '\b': printf("\b"); break; case '\n': printf("\n"); break; case '\r': printf("\r"); break; default: if (c >= 32 && c < 127) logprintf("%c", c); else logprintf("\\x%02X", c); } } logprintf("\""); logprintf(" (length %d)", q == NULL ? -1 : (int)strlen(q)); /* * If the current string is UTF-8, dump its code points. */ if (utf8) { size_t len; uint32_t uc; int n; int cnt = 0; p = q; len = strlen(p); logprintf(" ["); while ((n = _utf8_to_unicode(&uc, p, len)) > 0) { if (p != q) logprintf(" "); logprintf("%04X", uc); p += n; len -= n; cnt++; } logprintf("]"); logprintf(" (count %d", cnt); if (n < 0) { logprintf(",unknown %d bytes", len); } logprintf(")"); } logprintf("\n"); } /* Verify two strings are equal, dump them if not. */ int assertion_equal_string(const char *file, int line, const char *v1, const char *e1, const char *v2, const char *e2, void *extra, int utf8) { int l1, l2; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && strcmp(v1, v2) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); l1 = strlen(e1); l2 = strlen(e2); if (l1 < l2) l1 = l2; strdump(e1, v1, l1, utf8); strdump(e2, v2, l1, utf8); failure_finish(extra); return (0); } static void wcsdump(const char *e, const wchar_t *w) { logprintf(" %s = ", e); if (w == NULL) { logprintf("(null)"); return; } logprintf("\""); while (*w != L'\0') { unsigned int c = *w++; if (c >= 32 && c < 127) logprintf("%c", c); else if (c < 256) logprintf("\\x%02X", c); else if (c < 0x10000) logprintf("\\u%04X", c); else logprintf("\\U%08X", c); } logprintf("\"\n"); } #ifndef HAVE_WCSCMP static int wcscmp(const wchar_t *s1, const wchar_t *s2) { while (*s1 == *s2++) { if (*s1++ == L'\0') return 0; } if (*s1 > *--s2) return 1; else return -1; } #endif /* Verify that two wide strings are equal, dump them if not. */ int assertion_equal_wstring(const char *file, int line, const wchar_t *v1, const char *e1, const wchar_t *v2, const char *e2, void *extra) { assertion_count(file, line); if (v1 == v2) return (1); if (v1 != NULL && v2 != NULL && wcscmp(v1, v2) == 0) return (1); failure_start(file, line, "%s != %s", e1, e2); wcsdump(e1, v1); wcsdump(e2, v2); failure_finish(extra); return (0); } /* * Pretty standard hexdump routine. As a bonus, if ref != NULL, then * any bytes in p that differ from ref will be highlighted with '_' * before and after the hex value. */ static void hexdump(const char *p, const char *ref, size_t l, size_t offset) { size_t i, j; char sep; if (p == NULL) { logprintf("(null)\n"); return; } for(i=0; i < l; i+=16) { logprintf("%04x", (unsigned)(i + offset)); sep = ' '; for (j = 0; j < 16 && i + j < l; j++) { if (ref != NULL && p[i + j] != ref[i + j]) sep = '_'; logprintf("%c%02x", sep, 0xff & (int)p[i+j]); if (ref != NULL && p[i + j] == ref[i + j]) sep = ' '; } for (; j < 16; j++) { logprintf("%c ", sep); sep = ' '; } logprintf("%c", sep); for (j=0; j < 16 && i + j < l; j++) { int c = p[i + j]; if (c >= ' ' && c <= 126) logprintf("%c", c); else logprintf("."); } logprintf("\n"); } } /* Verify that two blocks of memory are the same, display the first * block of differences if they're not. */ int assertion_equal_mem(const char *file, int line, const void *_v1, const char *e1, const void *_v2, const char *e2, size_t l, const char *ld, void *extra) { const char *v1 = (const char *)_v1; const char *v2 = (const char *)_v2; size_t offset; assertion_count(file, line); if (v1 == v2 || (v1 != NULL && v2 != NULL && memcmp(v1, v2, l) == 0)) return (1); failure_start(file, line, "%s != %s", e1, e2); logprintf(" size %s = %d\n", ld, (int)l); /* Dump 48 bytes (3 lines) so that the first difference is * in the second line. */ offset = 0; while (l > 64 && memcmp(v1, v2, 32) == 0) { /* Two lines agree, so step forward one line. */ v1 += 16; v2 += 16; l -= 16; offset += 16; } logprintf(" Dump of %s\n", e1); hexdump(v1, v2, l < 128 ? l : 128, offset); logprintf(" Dump of %s\n", e2); hexdump(v2, v1, l < 128 ? l : 128, offset); logprintf("\n"); failure_finish(extra); return (0); } /* Verify that the named file exists and is empty. */ int assertion_empty_file(const char *filename, int line, const char *f1) { char buff[1024]; struct stat st; ssize_t s; FILE *f; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) return (1); failure_start(filename, line, "File should be empty: %s", f1); logprintf(" File size: %d\n", (int)st.st_size); logprintf(" Contents:\n"); f = fopen(f1, "rb"); if (f == NULL) { logprintf(" Unable to open %s\n", f1); } else { s = ((off_t)sizeof(buff) < st.st_size) ? (ssize_t)sizeof(buff) : (ssize_t)st.st_size; s = fread(buff, 1, s, f); hexdump(buff, NULL, s, 0); fclose(f); } failure_finish(NULL); return (0); } /* Verify that the named file exists and is not empty. */ int assertion_non_empty_file(const char *filename, int line, const char *f1) { struct stat st; assertion_count(filename, line); if (stat(f1, &st) != 0) { failure_start(filename, line, "Stat failed: %s", f1); failure_finish(NULL); return (0); } if (st.st_size == 0) { failure_start(filename, line, "File empty: %s", f1); failure_finish(NULL); return (0); } return (1); } /* Verify that two files have the same contents. */ /* TODO: hexdump the first bytes that actually differ. */ int assertion_equal_file(const char *filename, int line, const char *fn1, const char *fn2) { char buff1[1024]; char buff2[1024]; FILE *f1, *f2; int n1, n2; assertion_count(filename, line); f1 = fopen(fn1, "rb"); f2 = fopen(fn2, "rb"); for (;;) { n1 = fread(buff1, 1, sizeof(buff1), f1); n2 = fread(buff2, 1, sizeof(buff2), f2); if (n1 != n2) break; if (n1 == 0 && n2 == 0) { fclose(f1); fclose(f2); return (1); } if (memcmp(buff1, buff2, n1) != 0) break; } fclose(f1); fclose(f2); failure_start(filename, line, "Files not identical"); logprintf(" file1=\"%s\"\n", fn1); logprintf(" file2=\"%s\"\n", fn2); failure_finish(NULL); return (0); } /* Verify that the named file does exist. */ int assertion_file_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (!_access(f, 0)) return (1); #else if (!access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should exist: %s", f); failure_finish(NULL); return (0); } /* Verify that the named file doesn't exist. */ int assertion_file_not_exists(const char *filename, int line, const char *f) { assertion_count(filename, line); #if defined(_WIN32) && !defined(__CYGWIN__) if (_access(f, 0)) return (1); #else if (access(f, F_OK)) return (1); #endif failure_start(filename, line, "File should not exist: %s", f); failure_finish(NULL); return (0); } /* Compare the contents of a file to a block of memory. */ int assertion_file_contents(const char *filename, int line, const void *buff, int s, const char *fn) { char *contents; FILE *f; int n; assertion_count(filename, line); f = fopen(fn, "rb"); if (f == NULL) { failure_start(filename, line, "File should exist: %s", fn); failure_finish(NULL); return (0); } contents = malloc(s * 2); n = fread(contents, 1, s * 2, f); fclose(f); if (n == s && memcmp(buff, contents, s) == 0) { free(contents); return (1); } failure_start(filename, line, "File contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) hexdump(contents, buff, n > 512 ? 512 : n, 0); else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s > 512 ? 512 : s, 0); } failure_finish(NULL); free(contents); return (0); } /* Check the contents of a text file, being tolerant of line endings. */ int assertion_text_file_contents(const char *filename, int line, const char *buff, const char *fn) { char *contents; const char *btxt, *ftxt; FILE *f; int n, s; assertion_count(filename, line); f = fopen(fn, "r"); if (f == NULL) { failure_start(filename, line, "File doesn't exist: %s", fn); failure_finish(NULL); return (0); } s = strlen(buff); contents = malloc(s * 2 + 128); n = fread(contents, 1, s * 2 + 128 - 1, f); if (n >= 0) contents[n] = '\0'; fclose(f); /* Compare texts. */ btxt = buff; ftxt = (const char *)contents; while (*btxt != '\0' && *ftxt != '\0') { if (*btxt == *ftxt) { ++btxt; ++ftxt; continue; } if (btxt[0] == '\n' && ftxt[0] == '\r' && ftxt[1] == '\n') { /* Pass over different new line characters. */ ++btxt; ftxt += 2; continue; } break; } if (*btxt == '\0' && *ftxt == '\0') { free(contents); return (1); } failure_start(filename, line, "Contents don't match"); logprintf(" file=\"%s\"\n", fn); if (n > 0) { hexdump(contents, buff, n, 0); logprintf(" expected\n", fn); hexdump(buff, contents, s, 0); } else { logprintf(" File empty, contents should be:\n"); hexdump(buff, NULL, s, 0); } failure_finish(NULL); free(contents); return (0); } /* Verify that a text file contains the specified lines, regardless of order */ /* This could be more efficient if we sorted both sets of lines, etc, but * since this is used only for testing and only ever deals with a dozen or so * lines at a time, this relatively crude approach is just fine. */ int assertion_file_contains_lines_any_order(const char *file, int line, const char *pathname, const char *lines[]) { char *buff; size_t buff_size; size_t expected_count, actual_count, i, j; char **expected; char *p, **actual; char c; int expected_failure = 0, actual_failure = 0; assertion_count(file, line); buff = slurpfile(&buff_size, "%s", pathname); if (buff == NULL) { failure_start(pathname, line, "Can't read file: %s", pathname); failure_finish(NULL); return (0); } /* Make a copy of the provided lines and count up the expected file size. */ expected_count = 0; for (i = 0; lines[i] != NULL; ++i) { } expected_count = i; expected = malloc(sizeof(char *) * expected_count); for (i = 0; lines[i] != NULL; ++i) { expected[i] = strdup(lines[i]); } /* Break the file into lines */ actual_count = 0; for (c = '\0', p = buff; p < buff + buff_size; ++p) { if (*p == '\x0d' || *p == '\x0a') *p = '\0'; if (c == '\0' && *p != '\0') ++actual_count; c = *p; } actual = malloc(sizeof(char *) * actual_count); for (j = 0, p = buff; p < buff + buff_size; p += 1 + strlen(p)) { if (*p != '\0') { actual[j] = p; ++j; } } /* Erase matching lines from both lists */ for (i = 0; i < expected_count; ++i) { if (expected[i] == NULL) continue; for (j = 0; j < actual_count; ++j) { if (actual[j] == NULL) continue; if (strcmp(expected[i], actual[j]) == 0) { free(expected[i]); expected[i] = NULL; actual[j] = NULL; break; } } } /* If there's anything left, it's a failure */ for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) ++expected_failure; } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) ++actual_failure; } if (expected_failure == 0 && actual_failure == 0) { free(buff); free(expected); free(actual); return (1); } failure_start(file, line, "File doesn't match: %s", pathname); for (i = 0; i < expected_count; ++i) { if (expected[i] != NULL) { logprintf(" Expected but not present: %s\n", expected[i]); free(expected[i]); } } for (j = 0; j < actual_count; ++j) { if (actual[j] != NULL) logprintf(" Present but not expected: %s\n", actual[j]); } failure_finish(NULL); free(buff); free(expected); free(actual); return (0); } /* Test that two paths point to the same file. */ /* As a side-effect, asserts that both files exist. */ static int is_hardlink(const char *file, int line, const char *path1, const char *path2) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi1, bhfi2; int r; assertion_count(file, line); r = my_GetFileInformationByName(path1, &bhfi1); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path1); failure_finish(NULL); return (0); } r = my_GetFileInformationByName(path2, &bhfi2); if (r == 0) { failure_start(file, line, "File %s can't be inspected?", path2); failure_finish(NULL); return (0); } return (bhfi1.dwVolumeSerialNumber == bhfi2.dwVolumeSerialNumber && bhfi1.nFileIndexHigh == bhfi2.nFileIndexHigh && bhfi1.nFileIndexLow == bhfi2.nFileIndexLow); #else struct stat st1, st2; int r; assertion_count(file, line); r = lstat(path1, &st1); if (r != 0) { failure_start(file, line, "File should exist: %s", path1); failure_finish(NULL); return (0); } r = lstat(path2, &st2); if (r != 0) { failure_start(file, line, "File should exist: %s", path2); failure_finish(NULL); return (0); } return (st1.st_ino == st2.st_ino && st1.st_dev == st2.st_dev); #endif } int assertion_is_hardlink(const char *file, int line, const char *path1, const char *path2) { if (is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s are not hardlinked", path1, path2); failure_finish(NULL); return (0); } int assertion_is_not_hardlink(const char *file, int line, const char *path1, const char *path2) { if (!is_hardlink(file, line, path1, path2)) return (1); failure_start(file, line, "Files %s and %s should not be hardlinked", path1, path2); failure_finish(NULL); return (0); } /* Verify a/b/mtime of 'pathname'. */ /* If 'recent', verify that it's within last 10 seconds. */ static int assertion_file_time(const char *file, int line, const char *pathname, long t, long nsec, char type, int recent) { long long filet, filet_nsec; int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define EPOC_TIME (116444736000000000ULL) FILETIME ftime, fbirthtime, fatime, fmtime; ULARGE_INTEGER wintm; HANDLE h; ftime.dwLowDateTime = 0; ftime.dwHighDateTime = 0; assertion_count(file, line); /* Note: FILE_FLAG_BACKUP_SEMANTICS applies to open * a directory file. If not, CreateFile() will fail when * the pathname is a directory. */ h = CreateFile(pathname, FILE_READ_ATTRIBUTES, 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } r = GetFileTime(h, &fbirthtime, &fatime, &fmtime); switch (type) { case 'a': ftime = fatime; break; case 'b': ftime = fbirthtime; break; case 'm': ftime = fmtime; break; } CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't GetFileTime %s\n", pathname); failure_finish(NULL); return (0); } wintm.LowPart = ftime.dwLowDateTime; wintm.HighPart = ftime.dwHighDateTime; filet = (wintm.QuadPart - EPOC_TIME) / 10000000; filet_nsec = ((wintm.QuadPart - EPOC_TIME) % 10000000) * 100; nsec = (nsec / 100) * 100; /* Round the request */ #else struct stat st; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } switch (type) { case 'a': filet = st.st_atime; break; case 'm': filet = st.st_mtime; break; case 'b': filet = 0; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } #if defined(__FreeBSD__) switch (type) { case 'a': filet_nsec = st.st_atimespec.tv_nsec; break; case 'b': filet = st.st_birthtime; filet_nsec = st.st_birthtimespec.tv_nsec; break; case 'm': filet_nsec = st.st_mtimespec.tv_nsec; break; default: fprintf(stderr, "INTERNAL: Bad type %c for file time", type); exit(1); } /* FreeBSD generally only stores to microsecond res, so round. */ filet_nsec = (filet_nsec / 1000) * 1000; nsec = (nsec / 1000) * 1000; #else filet_nsec = nsec = 0; /* Generic POSIX only has whole seconds. */ if (type == 'b') return (1); /* Generic POSIX doesn't have birthtime */ #if defined(__HAIKU__) if (type == 'a') return (1); /* Haiku doesn't have atime. */ #endif #endif #endif if (recent) { /* Check that requested time is up-to-date. */ time_t now = time(NULL); if (filet < now - 10 || filet > now + 1) { failure_start(file, line, "File %s has %ctime %lld, %lld seconds ago\n", pathname, type, filet, now - filet); failure_finish(NULL); return (0); } } else if (filet != t || filet_nsec != nsec) { failure_start(file, line, "File %s has %ctime %lld.%09lld, expected %lld.%09lld", pathname, type, filet, filet_nsec, t, nsec); failure_finish(NULL); return (0); } return (1); } /* Verify atime of 'pathname'. */ int assertion_file_atime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'a', 0); } /* Verify atime of 'pathname' is up-to-date. */ int assertion_file_atime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'a', 1); } /* Verify birthtime of 'pathname'. */ int assertion_file_birthtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'b', 0); } /* Verify birthtime of 'pathname' is up-to-date. */ int assertion_file_birthtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'b', 1); } /* Verify mtime of 'pathname'. */ int assertion_file_mtime(const char *file, int line, const char *pathname, long t, long nsec) { return assertion_file_time(file, line, pathname, t, nsec, 'm', 0); } /* Verify mtime of 'pathname' is up-to-date. */ int assertion_file_mtime_recent(const char *file, int line, const char *pathname) { return assertion_file_time(file, line, pathname, 0, 0, 'm', 1); } /* Verify number of links to 'pathname'. */ int assertion_file_nlinks(const char *file, int line, const char *pathname, int nlinks) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; assertion_count(file, line); r = my_GetFileInformationByName(pathname, &bhfi); if (r != 0 && bhfi.nNumberOfLinks == (DWORD)nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, bhfi.nNumberOfLinks, nlinks); failure_finish(NULL); return (0); #else struct stat st; int r; assertion_count(file, line); r = lstat(pathname, &st); - if (r == 0 && st.st_nlink == nlinks) + if (r == 0 && (int)st.st_nlink == nlinks) return (1); failure_start(file, line, "File %s has %d links, expected %d", pathname, st.st_nlink, nlinks); failure_finish(NULL); return (0); #endif } /* Verify size of 'pathname'. */ int assertion_file_size(const char *file, int line, const char *pathname, long size) { int64_t filesize; int r; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) { BY_HANDLE_FILE_INFORMATION bhfi; r = !my_GetFileInformationByName(pathname, &bhfi); filesize = ((int64_t)bhfi.nFileSizeHigh << 32) + bhfi.nFileSizeLow; } #else { struct stat st; r = lstat(pathname, &st); filesize = st.st_size; } #endif if (r == 0 && filesize == size) return (1); failure_start(file, line, "File %s has size %ld, expected %ld", pathname, (long)filesize, (long)size); failure_finish(NULL); return (0); } /* Assert that 'pathname' is a dir. If mode >= 0, verify that too. */ int assertion_is_dir(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Dir should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISDIR(st.st_mode)) { failure_start(file, line, "%s is not a dir", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "Dir %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Verify that 'pathname' is a regular file. If 'mode' is >= 0, * verify that too. */ int assertion_is_reg(const char *file, int line, const char *pathname, int mode) { struct stat st; int r; #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ #endif assertion_count(file, line); r = lstat(pathname, &st); if (r != 0 || !S_ISREG(st.st_mode)) { failure_start(file, line, "File should exist: %s", pathname); failure_finish(NULL); return (0); } #if !defined(_WIN32) || defined(__CYGWIN__) /* Windows doesn't handle permissions the same way as POSIX, * so just ignore the mode tests. */ /* TODO: Can we do better here? */ - if (mode >= 0 && mode != (st.st_mode & 07777)) { + if (mode >= 0 && (mode_t)mode != (st.st_mode & 07777)) { failure_start(file, line, "File %s has wrong mode", pathname); logprintf(" Expected: 0%3o\n", mode); logprintf(" Found: 0%3o\n", st.st_mode & 07777); failure_finish(NULL); return (0); } #endif return (1); } /* Check whether 'pathname' is a symbolic link. If 'contents' is * non-NULL, verify that the symlink has those contents. */ static int is_symlink(const char *file, int line, const char *pathname, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) (void)pathname; /* UNUSED */ (void)contents; /* UNUSED */ assertion_count(file, line); /* Windows sort-of has real symlinks, but they're only usable * by privileged users and are crippled even then, so there's * really not much point in bothering with this. */ return (0); #else char buff[300]; struct stat st; ssize_t linklen; int r; assertion_count(file, line); r = lstat(pathname, &st); if (r != 0) { failure_start(file, line, "Symlink should exist: %s", pathname); failure_finish(NULL); return (0); } if (!S_ISLNK(st.st_mode)) return (0); if (contents == NULL) return (1); linklen = readlink(pathname, buff, sizeof(buff)); if (linklen < 0) { failure_start(file, line, "Can't read symlink %s", pathname); failure_finish(NULL); return (0); } buff[linklen] = '\0'; if (strcmp(buff, contents) != 0) return (0); return (1); #endif } /* Assert that path is a symlink that (optionally) contains contents. */ int assertion_is_symlink(const char *file, int line, const char *path, const char *contents) { if (is_symlink(file, line, path, contents)) return (1); if (contents) failure_start(file, line, "File %s is not a symlink to %s", path, contents); else failure_start(file, line, "File %s is not a symlink", path); failure_finish(NULL); return (0); } /* Create a directory and report any errors. */ int assertion_make_dir(const char *file, int line, const char *dirname, int mode) { assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) (void)mode; /* UNUSED */ if (0 == _mkdir(dirname)) return (1); #else if (0 == mkdir(dirname, mode)) return (1); #endif failure_start(file, line, "Could not create directory %s", dirname); failure_finish(NULL); return(0); } /* Create a file with the specified contents and report any failures. */ int assertion_make_file(const char *file, int line, const char *path, int mode, const char *contents) { #if defined(_WIN32) && !defined(__CYGWIN__) /* TODO: Rework this to set file mode as well. */ FILE *f; (void)mode; /* UNUSED */ assertion_count(file, line); f = fopen(path, "wb"); if (f == NULL) { failure_start(file, line, "Could not create file %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if (strlen(contents) != fwrite(contents, 1, strlen(contents), f)) { fclose(f); failure_start(file, line, "Could not write file %s", path); failure_finish(NULL); return (0); } } fclose(f); return (1); #else int fd; assertion_count(file, line); fd = open(path, O_CREAT | O_WRONLY, mode >= 0 ? mode : 0644); if (fd < 0) { failure_start(file, line, "Could not create %s", path); failure_finish(NULL); return (0); } if (contents != NULL) { if ((ssize_t)strlen(contents) != write(fd, contents, strlen(contents))) { close(fd); failure_start(file, line, "Could not write to %s", path); failure_finish(NULL); return (0); } } close(fd); return (1); #endif } /* Create a hardlink and report any failures. */ int assertion_make_hardlink(const char *file, int line, const char *newpath, const char *linkto) { int succeeded; assertion_count(file, line); #if defined(_WIN32) && !defined(__CYGWIN__) succeeded = my_CreateHardLinkA(newpath, linkto); #elif HAVE_LINK succeeded = !link(linkto, newpath); #else succeeded = 0; #endif if (succeeded) return (1); failure_start(file, line, "Could not create hardlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Create a symlink and report any failures. */ int assertion_make_symlink(const char *file, int line, const char *newpath, const char *linkto) { #if defined(_WIN32) && !defined(__CYGWIN__) int targetIsDir = 0; /* TODO: Fix this */ assertion_count(file, line); if (my_CreateSymbolicLinkA(newpath, linkto, targetIsDir)) return (1); #elif HAVE_SYMLINK assertion_count(file, line); if (0 == symlink(linkto, newpath)) return (1); #endif failure_start(file, line, "Could not create symlink"); logprintf(" New link: %s\n", newpath); logprintf(" Old name: %s\n", linkto); failure_finish(NULL); return(0); } /* Set umask, report failures. */ int assertion_umask(const char *file, int line, int mask) { assertion_count(file, line); (void)file; /* UNUSED */ (void)line; /* UNUSED */ umask(mask); return (1); } /* Set times, report failures. */ int assertion_utimes(const char *file, int line, const char *pathname, long at, long at_nsec, long mt, long mt_nsec) { int r; #if defined(_WIN32) && !defined(__CYGWIN__) #define WINTIME(sec, nsec) ((Int32x32To64(sec, 10000000) + EPOC_TIME)\ + (((nsec)/1000)*10)) HANDLE h; ULARGE_INTEGER wintm; FILETIME fatime, fmtime; FILETIME *pat, *pmt; assertion_count(file, line); h = CreateFileA(pathname,GENERIC_READ | GENERIC_WRITE, FILE_SHARE_READ | FILE_SHARE_WRITE, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); if (h == INVALID_HANDLE_VALUE) { failure_start(file, line, "Can't access %s\n", pathname); failure_finish(NULL); return (0); } if (at > 0 || at_nsec > 0) { wintm.QuadPart = WINTIME(at, at_nsec); fatime.dwLowDateTime = wintm.LowPart; fatime.dwHighDateTime = wintm.HighPart; pat = &fatime; } else pat = NULL; if (mt > 0 || mt_nsec > 0) { wintm.QuadPart = WINTIME(mt, mt_nsec); fmtime.dwLowDateTime = wintm.LowPart; fmtime.dwHighDateTime = wintm.HighPart; pmt = &fmtime; } else pmt = NULL; if (pat != NULL || pmt != NULL) r = SetFileTime(h, NULL, pat, pmt); else r = 1; CloseHandle(h); if (r == 0) { failure_start(file, line, "Can't SetFileTime %s\n", pathname); failure_finish(NULL); return (0); } return (1); #else /* defined(_WIN32) && !defined(__CYGWIN__) */ struct stat st; struct timeval times[2]; #if !defined(__FreeBSD__) mt_nsec = at_nsec = 0; /* Generic POSIX only has whole seconds. */ #endif if (mt == 0 && mt_nsec == 0 && at == 0 && at_nsec == 0) return (1); r = lstat(pathname, &st); if (r < 0) { failure_start(file, line, "Can't stat %s\n", pathname); failure_finish(NULL); return (0); } if (mt == 0 && mt_nsec == 0) { mt = st.st_mtime; #if defined(__FreeBSD__) mt_nsec = st.st_mtimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ mt_nsec = (mt_nsec / 1000) * 1000; #endif } if (at == 0 && at_nsec == 0) { at = st.st_atime; #if defined(__FreeBSD__) at_nsec = st.st_atimespec.tv_nsec; /* FreeBSD generally only stores to microsecond res, so round. */ at_nsec = (at_nsec / 1000) * 1000; #endif } times[1].tv_sec = mt; times[1].tv_usec = mt_nsec / 1000; times[0].tv_sec = at; times[0].tv_usec = at_nsec / 1000; #ifdef HAVE_LUTIMES r = lutimes(pathname, times); #else r = utimes(pathname, times); #endif if (r < 0) { failure_start(file, line, "Can't utimes %s\n", pathname); failure_finish(NULL); return (0); } return (1); #endif /* defined(_WIN32) && !defined(__CYGWIN__) */ } /* * * UTILITIES for use by tests. * */ /* * Check whether platform supports symlinks. This is intended * for tests to use in deciding whether to bother testing symlink * support; if the platform doesn't support symlinks, there's no point * in checking whether the program being tested can create them. * * Note that the first time this test is called, we actually go out to * disk to create and verify a symlink. This is necessary because * symlink support is actually a property of a particular filesystem * and can thus vary between directories on a single system. After * the first call, this returns the cached result from memory, so it's * safe to call it as often as you wish. */ int canSymlink(void) { /* Remember the test result */ static int value = 0, tested = 0; if (tested) return (value); ++tested; assertion_make_file(__FILE__, __LINE__, "canSymlink.0", 0644, "a"); /* Note: Cygwin has its own symlink() emulation that does not * use the Win32 CreateSymbolicLink() function. */ #if defined(_WIN32) && !defined(__CYGWIN__) value = my_CreateSymbolicLinkA("canSymlink.1", "canSymlink.0", 0) && is_symlink(__FILE__, __LINE__, "canSymlink.1", "canSymlink.0"); #elif HAVE_SYMLINK value = (0 == symlink("canSymlink.0", "canSymlink.1")) && is_symlink(__FILE__, __LINE__, "canSymlink.1","canSymlink.0"); #endif return (value); } /* * Can this platform run the gzip program? */ /* Platform-dependent options for hiding the output of a subcommand. */ #if defined(_WIN32) && !defined(__CYGWIN__) static const char *redirectArgs = ">NUL 2>NUL"; /* Win32 cmd.exe */ #else static const char *redirectArgs = ">/dev/null 2>/dev/null"; /* POSIX 'sh' */ #endif int canGzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Can this platform run the gunzip program? */ int canGunzip(void) { static int tested = 0, value = 0; if (!tested) { tested = 1; if (systemf("gunzip -V %s", redirectArgs) == 0) value = 1; } return (value); } /* * Sleep as needed; useful for verifying disk timestamp changes by * ensuring that the wall-clock time has actually changed before we * go back to re-read something from disk. */ void sleepUntilAfter(time_t t) { while (t >= time(NULL)) #if defined(_WIN32) && !defined(__CYGWIN__) Sleep(500); #else sleep(1); #endif } /* * Call standard system() call, but build up the command line using * sprintf() conventions. */ int systemf(const char *fmt, ...) { char buff[8192]; va_list ap; int r; va_start(ap, fmt); vsprintf(buff, fmt, ap); if (verbosity > VERBOSITY_FULL) logprintf("Cmd: %s\n", buff); r = system(buff); va_end(ap); return (r); } /* * Slurp a file into memory for ease of comparison and testing. * Returns size of file in 'sizep' if non-NULL, null-terminates * data in memory for ease of use. */ char * slurpfile(size_t * sizep, const char *fmt, ...) { char filename[8192]; struct stat st; va_list ap; char *p; ssize_t bytes_read; FILE *f; int r; va_start(ap, fmt); vsprintf(filename, fmt, ap); va_end(ap); f = fopen(filename, "rb"); if (f == NULL) { /* Note: No error; non-existent file is okay here. */ return (NULL); } r = fstat(fileno(f), &st); if (r != 0) { logprintf("Can't stat file %s\n", filename); fclose(f); return (NULL); } p = malloc((size_t)st.st_size + 1); if (p == NULL) { logprintf("Can't allocate %ld bytes of memory to read file %s\n", (long int)st.st_size, filename); fclose(f); return (NULL); } bytes_read = fread(p, 1, (size_t)st.st_size, f); if (bytes_read < st.st_size) { logprintf("Can't read file %s\n", filename); fclose(f); free(p); return (NULL); } p[st.st_size] = '\0'; if (sizep != NULL) *sizep = (size_t)st.st_size; fclose(f); return (p); } /* Read a uuencoded file from the reference directory, decode, and * write the result into the current directory. */ #define UUDECODE(c) (((c) - 0x20) & 0x3f) void extract_reference_file(const char *name) { char buff[1024]; FILE *in, *out; sprintf(buff, "%s/%s.uu", refdir, name); in = fopen(buff, "r"); failure("Couldn't open reference file %s", buff); assert(in != NULL); if (in == NULL) return; /* Read up to and including the 'begin' line. */ for (;;) { if (fgets(buff, sizeof(buff), in) == NULL) { /* TODO: This is a failure. */ return; } if (memcmp(buff, "begin ", 6) == 0) break; } /* Now, decode the rest and write it. */ /* Not a lot of error checking here; the input better be right. */ out = fopen(name, "wb"); while (fgets(buff, sizeof(buff), in) != NULL) { char *p = buff; int bytes; if (memcmp(buff, "end", 3) == 0) break; bytes = UUDECODE(*p++); while (bytes > 0) { int n = 0; /* Write out 1-3 bytes from that. */ if (bytes > 0) { n = UUDECODE(*p++) << 18; n |= UUDECODE(*p++) << 12; fputc(n >> 16, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++) << 6; fputc((n >> 8) & 0xFF, out); --bytes; } if (bytes > 0) { n |= UUDECODE(*p++); fputc(n & 0xFF, out); --bytes; } } } fclose(out); fclose(in); } int is_LargeInode(const char *file) { #if defined(_WIN32) && !defined(__CYGWIN__) BY_HANDLE_FILE_INFORMATION bhfi; int r; r = my_GetFileInformationByName(file, &bhfi); if (r != 0) return (0); return (bhfi.nFileIndexHigh & 0x0000FFFFUL); #else struct stat st; int64_t ino; if (stat(file, &st) < 0) return (0); ino = (int64_t)st.st_ino; return (ino > 0xffffffff); #endif } /* * * TEST management * */ /* * "list.h" is simply created by "grep DEFINE_TEST test_*.c"; it has * a line like * DEFINE_TEST(test_function) * for each test. */ /* Use "list.h" to declare all of the test functions. */ #undef DEFINE_TEST #define DEFINE_TEST(name) void name(void); #include "list.h" /* Use "list.h" to create a list of all tests (functions and names). */ #undef DEFINE_TEST #define DEFINE_TEST(n) { n, #n, 0 }, struct { void (*func)(void); const char *name; int failures; } tests[] = { #include "list.h" }; /* * Summarize repeated failures in the just-completed test. */ static void test_summarize(int failed) { unsigned int i; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: printf(failed ? "E" : "."); fflush(stdout); break; case VERBOSITY_PASSFAIL: printf(failed ? "FAIL\n" : "ok\n"); break; } log_console = (verbosity == VERBOSITY_LIGHT_REPORT); for (i = 0; i < sizeof(failed_lines)/sizeof(failed_lines[0]); i++) { if (failed_lines[i].count > 1 && !failed_lines[i].skip) logprintf("%s:%d: Summary: Failed %d times\n", failed_filename, i, failed_lines[i].count); } /* Clear the failure history for the next file. */ failed_filename = NULL; memset(failed_lines, 0, sizeof(failed_lines)); } /* * Actually run a single test, with appropriate setup and cleanup. */ static int test_run(int i, const char *tmpdir) { char workdir[1024]; char logfilename[64]; int failures_before = failures; int oldumask; switch (verbosity) { case VERBOSITY_SUMMARY_ONLY: /* No per-test reports at all */ break; case VERBOSITY_PASSFAIL: /* rest of line will include ok/FAIL marker */ printf("%3d: %-50s", i, tests[i].name); fflush(stdout); break; default: /* Title of test, details will follow */ printf("%3d: %s\n", i, tests[i].name); } /* Chdir to the top-level work directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Can't chdir to top work dir %s\n", tmpdir); exit(1); } /* Create a log file for this test. */ sprintf(logfilename, "%s.log", tests[i].name); logfile = fopen(logfilename, "w"); fprintf(logfile, "%s\n\n", tests[i].name); /* Chdir() to a work dir for this specific test. */ snprintf(workdir, sizeof(workdir), "%s/%s", tmpdir, tests[i].name); testworkdir = workdir; if (!assertMakeDir(testworkdir, 0755) || !assertChdir(testworkdir)) { fprintf(stderr, "ERROR: Can't chdir to work dir %s\n", testworkdir); exit(1); } /* Explicitly reset the locale before each test. */ setlocale(LC_ALL, "C"); /* Record the umask before we run the test. */ umask(oldumask = umask(0)); /* * Run the actual test. */ (*tests[i].func)(); /* * Clean up and report afterwards. */ testworkdir = NULL; /* Restore umask */ umask(oldumask); /* Reset locale. */ setlocale(LC_ALL, "C"); /* Reset directory. */ if (!assertChdir(tmpdir)) { fprintf(stderr, "ERROR: Couldn't chdir to temp dir %s\n", tmpdir); exit(1); } /* Report per-test summaries. */ tests[i].failures = failures - failures_before; test_summarize(tests[i].failures); /* Close the per-test log file. */ fclose(logfile); logfile = NULL; /* If there were no failures, we can remove the work dir and logfile. */ if (tests[i].failures == 0) { if (!keep_temp_files && assertChdir(tmpdir)) { #if defined(_WIN32) && !defined(__CYGWIN__) /* Make sure not to leave empty directories. * Sometimes a processing of closing files used by tests * is not done, then rmdir will be failed and it will * leave a empty test directory. So we should wait a few * seconds and retry rmdir. */ int r, t; for (t = 0; t < 10; t++) { if (t > 0) Sleep(1000); r = systemf("rmdir /S /Q %s", tests[i].name); if (r == 0) break; } systemf("del %s", logfilename); #else systemf("rm -rf %s", tests[i].name); systemf("rm %s", logfilename); #endif } } /* Return appropriate status. */ return (tests[i].failures); } /* * * * MAIN and support routines. * * */ static void usage(const char *program) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i; printf("Usage: %s [options] ...\n", program); printf("Default is to run all tests.\n"); printf("Otherwise, specify the numbers of the tests you wish to run.\n"); printf("Options:\n"); printf(" -d Dump core after any failure, for debugging.\n"); printf(" -k Keep all temp files.\n"); printf(" Default: temp files for successful tests deleted.\n"); #ifdef PROGRAM printf(" -p Path to executable to be tested.\n"); printf(" Default: path taken from " ENVBASE " environment variable.\n"); #endif printf(" -q Quiet.\n"); printf(" -r Path to dir containing reference files.\n"); printf(" Default: Current directory.\n"); printf(" -u Keep running specifies tests until one fails.\n"); printf(" -v Verbose.\n"); printf("Available tests:\n"); for (i = 0; i < limit; i++) printf(" %d: %s\n", i, tests[i].name); exit(1); } static char * get_refdir(const char *d) { char tried[512] = { '\0' }; char buff[128]; char *pwd, *p; /* If a dir was specified, try that */ if (d != NULL) { pwd = NULL; snprintf(buff, sizeof(buff), "%s", d); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); goto failure; } /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; /* Look for a known file. */ snprintf(buff, sizeof(buff), "%s", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(LIBRARY) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, LIBRARY); #else snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM); #endif p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #if defined(PROGRAM_ALIAS) snprintf(buff, sizeof(buff), "%s/%s/test", pwd, PROGRAM_ALIAS); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); #endif if (memcmp(pwd, "/usr/obj", 8) == 0) { snprintf(buff, sizeof(buff), "%s", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); snprintf(buff, sizeof(buff), "%s/test", pwd + 8); p = slurpfile(NULL, "%s/%s", buff, KNOWNREF); if (p != NULL) goto success; strncat(tried, buff, sizeof(tried) - strlen(tried) - 1); strncat(tried, "\n", sizeof(tried) - strlen(tried) - 1); } failure: printf("Unable to locate known reference file %s\n", KNOWNREF); printf(" Checked following directories:\n%s\n", tried); #if defined(_WIN32) && !defined(__CYGWIN__) && defined(_DEBUG) DebugBreak(); #endif exit(1); success: free(p); free(pwd); return strdup(buff); } int main(int argc, char **argv) { static const int limit = sizeof(tests) / sizeof(tests[0]); int i = 0, j = 0, start, end, tests_run = 0, tests_failed = 0, option; time_t now; char *refdir_alloc = NULL; const char *progname; char **saved_argv; const char *tmp, *option_arg, *p; char tmpdir[256], *pwd, *testprogdir, *tmp2 = NULL; char tmpdir_timestamp[256]; (void)argc; /* UNUSED */ /* Get the current dir. */ #ifdef PATH_MAX pwd = getcwd(NULL, PATH_MAX);/* Solaris getcwd needs the size. */ #else pwd = getcwd(NULL, 0); #endif while (pwd[strlen(pwd) - 1] == '\n') pwd[strlen(pwd) - 1] = '\0'; #if defined(HAVE__CrtSetReportMode) /* To stop to run the default invalid parameter handler. */ _set_invalid_parameter_handler(invalid_parameter_handler); /* Disable annoying assertion message box. */ _CrtSetReportMode(_CRT_ASSERT, 0); #endif /* * Name of this program, used to build root of our temp directory * tree. */ progname = p = argv[0]; if ((testprogdir = (char *)malloc(strlen(progname) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir, progname); while (*p != '\0') { /* Support \ or / dir separators for Windows compat. */ if (*p == '/' || *p == '\\') { progname = p + 1; i = j; } ++p; j++; } testprogdir[i] = '\0'; if (testprogdir[0] != '/') { /* Fixup path for relative directories. */ if ((testprogdir = (char *)realloc(testprogdir, strlen(pwd) + 1 + strlen(testprogdir) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(testprogdir + strlen(pwd) + 1, testprogdir); strcpy(testprogdir, pwd); testprogdir[strlen(pwd)] = '/'; } #ifdef PROGRAM /* Get the target program from environment, if available. */ testprogfile = getenv(ENVBASE); #endif if (getenv("TMPDIR") != NULL) tmp = getenv("TMPDIR"); else if (getenv("TMP") != NULL) tmp = getenv("TMP"); else if (getenv("TEMP") != NULL) tmp = getenv("TEMP"); else if (getenv("TEMPDIR") != NULL) tmp = getenv("TEMPDIR"); else tmp = "/tmp"; /* Allow -d to be controlled through the environment. */ if (getenv(ENVBASE "_DEBUG") != NULL) dump_on_failure = 1; /* Get the directory holding test files from environment. */ refdir = getenv(ENVBASE "_TEST_FILES"); /* * Parse options, without using getopt(), which isn't available * on all platforms. */ ++argv; /* Skip program name */ while (*argv != NULL) { if (**argv != '-') break; p = *argv++; ++p; /* Skip '-' */ while (*p != '\0') { option = *p++; option_arg = NULL; /* If 'opt' takes an argument, parse that. */ if (option == 'p' || option == 'r') { if (*p != '\0') option_arg = p; else if (*argv == NULL) { fprintf(stderr, "Option -%c requires argument.\n", option); usage(progname); } else option_arg = *argv++; p = ""; /* End of this option word. */ } /* Now, handle the option. */ switch (option) { case 'd': dump_on_failure = 1; break; case 'k': keep_temp_files = 1; break; case 'p': #ifdef PROGRAM testprogfile = option_arg; #else fprintf(stderr, "-p option not permitted\n"); usage(progname); #endif break; case 'q': verbosity--; break; case 'r': refdir = option_arg; break; case 'u': until_failure++; break; case 'v': verbosity++; break; default: fprintf(stderr, "Unrecognized option '%c'\n", option); usage(progname); } } } /* * Sanity-check that our options make sense. */ #ifdef PROGRAM if (testprogfile == NULL) { if ((tmp2 = (char *)malloc(strlen(testprogdir) + 1 + strlen(PROGRAM) + 1)) == NULL) { fprintf(stderr, "ERROR: Out of memory."); exit(1); } strcpy(tmp2, testprogdir); strcat(tmp2, "/"); strcat(tmp2, PROGRAM); testprogfile = tmp2; } { char *testprg; #if defined(_WIN32) && !defined(__CYGWIN__) /* Command.com sometimes rejects '/' separators. */ testprg = strdup(testprogfile); for (i = 0; testprg[i] != '\0'; i++) { if (testprg[i] == '/') testprg[i] = '\\'; } testprogfile = testprg; #endif /* Quote the name that gets put into shell command lines. */ testprg = malloc(strlen(testprogfile) + 3); strcpy(testprg, "\""); strcat(testprg, testprogfile); strcat(testprg, "\""); testprog = testprg; } #endif #if !defined(_WIN32) && defined(SIGPIPE) { /* Ignore SIGPIPE signals */ struct sigaction sa; sa.sa_handler = SIG_IGN; sigemptyset(&sa.sa_mask); sa.sa_flags = 0; sigaction(SIGPIPE, &sa, NULL); } #endif /* * Create a temp directory for the following tests. * Include the time the tests started as part of the name, * to make it easier to track the results of multiple tests. */ now = time(NULL); for (i = 0; ; i++) { strftime(tmpdir_timestamp, sizeof(tmpdir_timestamp), "%Y-%m-%dT%H.%M.%S", localtime(&now)); sprintf(tmpdir, "%s/%s.%s-%03d", tmp, progname, tmpdir_timestamp, i); if (assertMakeDir(tmpdir,0755)) break; if (i >= 999) { fprintf(stderr, "ERROR: Unable to create temp directory %s\n", tmpdir); exit(1); } } /* * If the user didn't specify a directory for locating * reference files, try to find the reference files in * the "usual places." */ refdir = refdir_alloc = get_refdir(refdir); /* * Banner with basic information. */ printf("\n"); printf("If tests fail or crash, details will be in:\n"); printf(" %s\n", tmpdir); printf("\n"); if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("Reference files will be read from: %s\n", refdir); #ifdef PROGRAM printf("Running tests on: %s\n", testprog); #endif printf("Exercising: "); fflush(stdout); printf("%s\n", EXTRA_VERSION); } else { printf("Running "); fflush(stdout); } /* * Run some or all of the individual tests. */ saved_argv = argv; do { argv = saved_argv; if (*argv == NULL) { /* Default: Run all tests. */ for (i = 0; i < limit; i++) { tests_run++; if (test_run(i, tmpdir)) { tests_failed++; if (until_failure) goto finish; } } } else { while (*(argv) != NULL) { if (**argv >= '0' && **argv <= '9') { char *vp = *argv; start = 0; while (*vp >= '0' && *vp <= '9') { start *= 10; start += *vp - '0'; ++vp; } if (*vp == '\0') { end = start; } else if (*vp == '-') { ++vp; if (*vp == '\0') { end = limit - 1; } else { end = 0; while (*vp >= '0' && *vp <= '9') { end *= 10; end += *vp - '0'; ++vp; } } } else { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } if (start < 0 || end >= limit || start > end) { printf("*** INVALID Test %s\n", *argv); free(refdir_alloc); usage(progname); return (1); } } else { for (start = 0; start < limit; ++start) { if (strcmp(*argv, tests[start].name) == 0) break; } end = start; if (start >= limit) { printf("*** INVALID Test ``%s''\n", *argv); free(refdir_alloc); usage(progname); /* usage() never returns */ } } while (start <= end) { tests_run++; if (test_run(start, tmpdir)) { tests_failed++; if (until_failure) goto finish; } ++start; } argv++; } } } while (until_failure); finish: /* Must be freed after all tests run */ free(tmp2); free(testprogdir); free(pwd); /* * Report summary statistics. */ if (verbosity > VERBOSITY_SUMMARY_ONLY) { printf("\n"); printf("Totals:\n"); printf(" Tests run: %8d\n", tests_run); printf(" Tests failed: %8d\n", tests_failed); printf(" Assertions checked:%8d\n", assertions); printf(" Assertions failed: %8d\n", failures); printf(" Skips reported: %8d\n", skips); } if (failures) { printf("\n"); printf("Failing tests:\n"); for (i = 0; i < limit; ++i) { if (tests[i].failures) printf(" %d: %s (%d failures)\n", i, tests[i].name, tests[i].failures); } printf("\n"); printf("Details for failing tests: %s\n", tmpdir); printf("\n"); } else { if (verbosity == VERBOSITY_SUMMARY_ONLY) printf("\n"); printf("%d tests passed, no failures\n", tests_run); } free(refdir_alloc); /* If the final tmpdir is empty, we can remove it. */ /* This should be the usual case when all tests succeed. */ assertChdir(".."); rmdir(tmpdir); return (tests_failed ? 1 : 0); }