>> + if (repo_init(&subrepo, sub_gitdir, sub_worktree)) >> + return; > > Note that in Duy's object-store series he made this function private > (IIRC) so this will result in some clash of the two series. > Yes, that is the case. I wonder if I'd rather revert to v1 where we only use the submodule repo init, or if we revert b2f0eceecf (repository: initialize the_repository in main(), 2018-03-03) partially to have repo_init available. I would think the approach with submodule init is a bit cleaner though has some more lines of code, using just repo_init seems easier, but we really have no use case for a separate repo_init unless they are submodules. And here we ought to check for the repo being a submodule. Not yet sure which path to take.
