On Wed, 2014-05-07 at 10:42 -0700, Junio C Hamano wrote: > David Turner <dturner@xxxxxxxxxxxxxxxx> writes: > > > This causes my test to pass and generally seems correct to me. > > Yes, this approach is very sensible, and I'll queue. > > But watchman support _should_ be prepared for a program that does > not do this. Developing your support in on a codebase with this > patch may be sweeping a bug in your code under the rug. I agree that good defensive coding practice would be to not depend on the cwd. That's just what everything in environment.c presently does. I don't want to change the rest of the get_*_file functions, because I don't know what their callers expect of them. Here is my patch to do change just the fs_cache bit: https://github.com/dturner-tw/git/commit/3fe93aeaee9719ee171a253c49af5126a057c513.patch (I went ahead and made it part of the watchman branch on https://github.com/dturner-tw/git.git ) I wanted to just do the fixup where the path is used, but I couldn't see a function for that. If there's no function, that indicates to me that it's probably a bad idea. But maybe I'm just missing it. -- 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