On Thu, Feb 21, 2019 at 11:53:12AM +0100, Mateusz Loskot wrote: > On Thu, 21 Feb 2019 at 05:17, Jeff King <peff@xxxxxxxx> wrote: > > On Thu, Feb 21, 2019 at 10:47:45AM +0700, Duy Nguyen wrote: > > > > > > > I managed to identify where that garbage injections come from: > > > > > from git aliases I've got configured [1] > > > > > > > > > > Could anyone explain what is happening here? > > > > > Is there anything wrong with my ~/.gitconfig [1] ? > > > > > > > > I see the right behavior on my system (Debian amd64/sid) with > > > > 2.21.0.rc0.258.g878e2cd30e and with master. I've built with ASan and > > > > don't see any memory warnings on my system. > > > > > > I ran valgrind yesterday (gentoo amd64) and didn't find the problem > > > either even though it clearly looked like some dangling pointers. I > > > also audited this code and couldn't find anything obviously wrong. My > > > only suspicion is maybe some strange getenv() behavior on Windows > > > (_if_ the super prefix is used, but I can't see how...) and that's > > > just a dead end for me. > > > > Certainly we fixed a bunch of getenv() problems in the upcoming 2.21 > > release, including 8aac69038f (get_super_prefix(): copy getenv() result, > > 2019-01-11). > > > > Mateusz, can you try with the one of the v2.21.0 release candidates (or > > the current tip of "master")? > > Jeff, I have just tried git version 2.21.0.rc2.windows.1 and > I no longer see the problem. > It seems like a bug in earlier version indeed, that has been fixed now. Great, thanks for confirming. Junio, Johannes: I don't know if there were particular plans to pick up those getenv() fixes for "maint", but if we're going to do a v2.20.2, it might be worth it. It's also possible that it's just due to ca1b411648 (mingw: safe-guard a bit more against getenv() problems, 2019-02-15), but I'd consider that to be maint-worthy as well. It may be that we just don't do a v2.20.2, though, since v2.21 is so close to release. -Peff
