3 ms·
The mentioned git-in-git problem always annoyed me, at least it should be possible to store .git/ subtrees anywhere else but on the toplevel (it very likely is
by zeotroph 7y ago
The mentioned git-in-git problem always annoyed me, at least it should be possible to store .git/ subtrees anywhere else but on the toplevel (it very likely is already supported in the data model). Then git can scream at me for not using submodules, and I can then silence that warning because I obviously know what I am doing!
legit might be an esoteric use case for that, but storing raw testdata for git hooks (not a zip or the commands to create the repo) are actual usecases.
- Phrodo_00 7y agoFor now I guess you could use a different dvcs like mercurial
- codetrotter 7y agoOr you could serialize the original git tree into another format and commit that. Then have a build step that deserializes it back into a git tree. And if you make a change to your legit program then you serialize the new tree and make a new commit with that. I think git format-patch would be "appropriate" for this. git format-patch will generate a patch-file per commit. It might not be very pleasant and I wouldn't suggest doing it this way for most cases, but I think in the context of "legit" it would be the most obvious way and it would be in line with the spirit of "legit" itself. And in fact if Mercurial Queues are what you were referring to then I seem to recall that it works somewhat in the same way that I described above, doesn't it? At least in terms of being about working with diffs of diffs I mean.
- bayareanative 7y agoI wonder if you can use gitmodules configured out-of-tree in .git/config. In my experience, git in git ignores other git trees like they're not there IIRC.