On Wed, Mar 02, 2011 at 10:48:00PM -0500, Chad Joan wrote:
> It seems to be working! I've tried it with 'git rm' and when pulling
> deletions.
Great.
> I imagine that race condition can happen if files in the directory are
> being modified while git does an rmdir? If that's the case then I'm
> not too worried. There is only one other programmer that might be
> working with me at the same time on an infrequently used directory.
The race condition I mentioned earlier was for a different workaround.
Basically there are two strategies, each with a difference race:
1. Don't rmdir on non-empty directories. This means we have to opendir
the directory and look for entries before rmdir(). If there is file
activity in the directory while we are looking we may think it is
empty when it's not and rmdir(), screwing up the permissions.
2. Before any rmdir, check permissions. Do the rmdir, and then restore
the permissions if rmdir fails. The race here is if somebody is
modifying the permissions on a non-empty directory, we may
overwrite their changes.
Obviously the patch does (2), so there is still that race.
> diff -crB git-1.7.3.4/dir.c git-1.7.3.4-new/dir.c
Context diff? Eww. There is this awesome tool called "git" that can help
you with managing versions of software. :)
-Peff
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html