Thanks! I think we're close to getting this merged :) Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> writes: > = Changes since v2: > > * Fixes for test_when_finished() in test setup, and got rid of > redundant test_config_global. > > * There's a new 2/9, which passes along get_super_prefix() as a > parameter. This allows us to gradually replace it, and drop the > *_sp() variants of functions that the previous version introduced, > and it adds "super_prefix" to the absorb_git_dir_into_superproject() > call in submodule_move_head(), which as Glen noticed I'd missed > > * Squashed the "deinit" change into that 2/9. > > * Explain why we keep the "fsmonitor" test bits that we do. > > * Dropped the new "git branch" output tests, turns out I was just > wrong, and was conflating it with the subsequent read-tree > invocation... > > So, this should address all outstanding feedbakc, unless I've missed > something. As noted in [1], I think we might have missed some sites where we should pass super_prefix. [1] https://lore.kernel.org/git/kl6lcz9eep9k.fsf@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx, > The one loose end here is that I still have no idea if you can invoke > get "read-tree" to invoke that submodule_move_head() in such a way as > to have the "super_prefix" used, I've failed to come up with a test > case for that. > > But for the purposes of this topic it doesn't really matter. In 8/10 > we'll start passing the new "--super-prefix" that "read-tree" gets > down to that function. At worst we're handing it to it redundantly, > but that was the case before too. > > So we can leave potentially turning that into a "NULL" for some other > time, for now providing the "super_prefix" is harmless, and guarantees > that we avoid any regression in that area from not providing it, in > case I've missed a way to have it matter in that case. Yeah, I tried to do the same too, and it turned out to be harder than expected. The "absorbgitdir during read-tree" code is definitely in use (e.g. by t1013), so it's certainly possible, but as you noted, I don't really think it matters as long as we avoid a regression for the time being.
