Re: Strangeness with git-add and nested repositories

From: Stefan Beller <sbeller@google.com>
Date: Apr 28, 2016, 16:54 UTC
Message-ID: <CAGZ79ka0WB58HyHYXAkaMYKccWD-NwSxhRUGHvq1TrA_ix+Hxw@mail.gmail.com>
In-Reply-To: <xmqq1t5p7kmp.fsf@gitster.mtv.corp.google.com>

On Thu, Apr 28, 2016 at 9:39 AM, Junio C Hamano <gitster@pobox.com> wrote:

Show 20 quoted lines

> Stefan Beller <sbeller@google.com> writes:
>
>> I think (pure speculation), that it the error is in the context
>> (repository) switching logic.
>> What happens if you alter the order, i.e. give testfile first and then
>> the files in the nested
>> repos?
>>
>>     git add -- file path/to/subdir/file
>>
>> should do internally IMHO:
>>
>>     git add file
>>     git -C path-to-subdir add file
>
> My undertanding of what _should_ happen in the world order as
> currently defined (not necessarily implemented) is:
>
>  * "git add -- A B" must work the same way as "git add -- B A" and
>    "git add -- A; git add -- B"

I agree.

>
>  * "git add -- path/to/subdir/file", when any of path/, path/to/,
>    path/to/subdir/ is a Git repository that is different from the
>    current Git repository, must fail.

I agree that this is the current expectation for the world order. However I would like to propose to change that eventually. (Once the submodule groups are there and we can treat submodules as a special form of narrow checkout, we want to have the feature of adding across submodules and even committing across submodules/repos, I would think)

Show 29 quoted lines

>
> IOW, if 'path' is a repository (whether it is known as a submodule
> to the repository whose working tree contains it, or it is an
> untracked directory from the containing repository's point of view),
> the index of the containing repository cannot get path/$anything in
> it.  If you managed to do so, you found a bug [*1*].
>
> path/.git/index can of course have "to/subdir/file" in it, and from
> that point of view, "git -C path/to/subdir add file" may one day
> become an improved world order.  It is just we haven't discussed
> that possibility or reached concensus that it is a good idea.
>
>
> [Footnote]
>
> *1* Of course, some of the bugs in this class may fundamentally be
>     unfixable and would fall into the same category as "doctor, it
>     hurts when I do this--don't do it then".  For example, you may
>     treat path/ as the top of the working tree of another repository
>     whose git-dir is not at path/.git by arranging GIT_WORK_TREE and
>     GIT_DIR environment variables, but you may do so only when you
>     actually are accessing the contents of path/ as its own project.
>     And when you are using the enclosing project (whose .git/ would
>     sit next to path/), there is no way for "git add path/to/file"
>     to know that everything under "path/" does not belong to the
>     current repository and instead it is part of the project rooted
>     at path/, which is an obvious example of "fundamentally
>     unfixable" case.
>