Fix bug 200942 - Don't squash fts_children errors

[cse_cem_gmail_com]

Summary

Preserve errors from fts_* via fts_children by saving errno over
fchdir(2) and close(2).

[cse_cem_gmail_com]

Don't preserve errno from close(2)

We don't return NULL to the caller on close(2) errors, so they have no reason
to check errno anyway.

[darius-dons.net.au]

LGTM

[jilles]

lib/libc/gen/fts.c

574–577

This should probably be

serrno = sp->fts_child == NULL ? errno : 0;

since fts_build() may set errno to a non-zero value even if it does not fail (e.g. if symlinks are being followed and there is a dangling symlink (FTS_SLNONE)).

[cse_cem_gmail_com]

Per Jilles' feedback, only set serrno value if fts_build actually fails.

[cse_cem_gmail_com]

Thanks!

[cse_cem_gmail_com]

lib/libc/gen/fts.c

574–581

This is maybe now bogus. In particular, the sequence:

1. fts_build() succeeds but sets errno (symlink case);
2. fchdir() succeeds and doesn't clear errno
3. serrno gets the symlink error that we were trying to avoid

[cse_cem_gmail_com]

Address the concern I brought up re: using stale errno value after fchdir(2).

[jilles]

Looks good to me.

[cse_cem_gmail_com]

Thanks. Can you commit this for me? I don't have a src bit yet.

[jilles]

Committed to head.