Teach bsdgrep(1) how to handle the special case of a pattern being terminated early with a NULL byte. We ignore these patterns, in line with how gnugrep(1) handles the same case.
PR: 202022
Details
Details
- Reviewers
emaste ngie - Commits
- rS316542: bsdgrep: Handle special case of single-byte NUL pattern
Test kcwu@csie.org's original example, ensure no segfault and no matches.
Run kyua tests to check that we've not caused any further regressions.
Diff Detail
Diff Detail
- Lint
Lint Passed - Unit
No Test Coverage - Build Status
Buildable 8491 Build 8794: arc lint + arc unit
Event Timeline
Comment Actions
Practically, when does this happen? Malicious/test inputs only?
Also, what happens with a nul mid-line? Just the first part of the line is used, the second part discarded?
| usr.bin/grep/grep.c | ||
|---|---|---|
| 318 | style(9): space between "if" and "(". | |
Comment Actions
Yes, or it could be under the class of rm -rf /$empty problems -- the following produces a perfectly valid pattern file that does exactly what you expect, because getline(3) does the right thing:
printf "BSD\0\nLicensing\0\n" > /tmp/pat grep -f /tmp/pat /COPYRIGHT
It's a quick step from there to, in some case, printf "$empty\0\n" >> /tmp/pat in some loop to generate a series of expressions to test by. The only problem here is we hadn't taken into consideration what getline(3) does when it comes across a "\0\n" line, which is produce the tested-for above.
Also, what happens with a nul mid-line? Just the first part of the line is used, the second part discarded?
Correct.
| usr.bin/grep/grep.c | ||
|---|---|---|
| 318 | Could you please check for '\0' instead of 0 for reader/static analysis tool clarity? | |