On Tue, May 24, 2011 at 05:57:59PM -0400, Jeff King wrote: > On Tue, May 24, 2011 at 05:52:02PM -0400, Jeff King wrote: > > > On Tue, May 24, 2011 at 05:46:18PM -0400, Jeff King wrote: > > > > > I think the right fix is simply to drop the "don't re-check the > > > environment after the first time" logic. It's not expensive to parse > > > compared to parsing config files, which is when we would do it. We can > > > just drop the existing list and reparse. You can even get rid of the > > > whole list and drop a bunch of code, I think, like: > > > > Ack, wrong patch. That one doesn't even come close to compiling. > > > > Try this (still not well tested, though). > > Ugh, broken. That will teach me to just paste any random junk into my > MUA. Hopefully you got the gist of what I was trying to say, but let me > come up with a more readable and tested series. OK, for real this time. This is how I would do the whole fix on top of master, including your 1/2. I'll let you handle the apply-to-maint-and-merge as you would have with your original series. The first two are refactoring to make 3/4 a little easier to read. The third one is my fix, and the fourth is your original patch (together with my 3/4 it passes the test; btw, I ended up writing a slightly simpler test. Feel free to throw it out if you prefer yours). [1/4]: config: make environment parsing routines static [2/4]: git_config: don't peek at global config_parameters [3/4]: config: always parse GIT_CONFIG_PARAMETERS during git_config [4/4]: handle_options(): do not miscount how many arguments were used -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