Fixing git’s wagon, part 2
Even after making git STFU about using repositories in my traditional manner, I was still having difficulties making things work. IN PARTICULAR, I ended up with copies of discount that had a master and and origin/master branch, and after doing any work on the machine with git 1.7, I’d end up with things getting pushed into that repository that were inaccessable except by pushing to the (git 1.5.2.2) upstream, then deleting the repository and cloning it back.
Needles to say, this started getting very tiring. So when I realized that git 1.7 was contaminating my repositories with imaginary origin/master branches (and shoving those imaginary branches to my archive @ github,) I went back and revisited why my configure.sh scripts ended up generating unbuildable configurations on FreeBSD 7.1 (answer: because sockaddr_storage wasn’t in netdb.h anymore, but in sys/socket.h instead, and I wasn’t checking there before telling git to emulate it itself) and, once I fixed that, I blew away git 1.7 and replaced it with the older and less doctrinaire 1.5.2.2 release.
I realize that this is merely an invitation for trouble when the repository format for git is completely rewritten for their 2.0 release, but at that point I’ll merely revert back to only publishing release tarballs instead of exposing my daily development commits.