On Fri, Jan 20, 2017 at 1:58 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote:
> Junio C Hamano <gitster@xxxxxxxxx> writes:
>
>> Jeff King <peff@xxxxxxxx> writes:
>>
>>>> And in my current understanding of submodules the check in
>>>> .gitmodules ought to be enough, too.
>>>
>>> Yeah, that probably makes sense. You can have a gitlink without a
>>> .gitmodules file, but I don't quite know what that would mean in terms
>>> of submodules (I guess it's not a submodule but "something else").
>>
>> That may be a lot better than reading the index unconditionally, but
>> I'd rather not to see "git rev-parse" read ".gitmodules" at all. It
>> would discourage scripted use of Git for no good reason.
>
> Thinking about this more, I suspect that
>
> cd sub && git anything
>
> when the index of the top-level thinks "sub" must be a submodule and
> the user is not interested in "sub" (hence it hasn't gone through
> "git submodule init" or "update") should get the same error as you
> would get if you did
>
> cd /var/tmp/ && git anything
>
> when none of /, /var, /var/tmp/ is controlled by any Git repository.
> I.e. "fatal: Not a git repository".
I agree. The idea with a tombstone sounds great from a
performance perspective as you do not need to do extra work
in the superproject at all, because any gitlink is detected early
in the discovery.
The big BUT is however the following:
How do current users know if a submodule is e.g. populated?
(From say a third party script). Most likely they use something like
test -e ${sub}/.git
as that just worked. So if we go with the tombstone idea, we
may break things. (i.e. the fictional third party script confirms any
submodule to be there, but it is not)
I do really like the idea though, so maybe we also need to provide
some submodule plumbing that we opine to be the "correct" way
to see the submodules state[1] to make the transition easier for the
script writers?
[1] c.f. submodule states in
https://github.com/gitster/git/commit/e2b51b9df618ceeff7c4ec044e20f5ce9a87241e
>
> Perhaps we can update two things and make it cheap.
>
> - checking out the top-level working tree without populating the
> working tree of a submodule learns to do a bit more than just
> creating an empty directory. Instead, it creates the third kind
> of ".git" (we currently support two kinds of ".git", one that is
> a repository itself, and another that is points at a repository),
> that tells us that there is not (yet) a repository there.
>
> - the "discovering the root of the working tree" logic learns to
> notice the third kind of ".git" and stop with "Not a git
> repository".