Re: [PATCH 2/2] submodule.c: warn about missing submodule git directories

From: Stefan Beller <sbeller@google.com>
Date: Aug 28, 2018, 21:49 UTC
Message-ID: <CAGZ79kYApAdFzC82LZNjYDzL5uWxtXHbvqUyjZ9OF+b=pGNbUA@mail.gmail.com>
In-Reply-To: <xmqqtvnefhgx.fsf@gitster-ct.c.googlers.com>

On Tue, Aug 28, 2018 at 11:56 AM Junio C Hamano <gitster@pobox.com> wrote:

Show 13 quoted lines

>
> Stefan Beller <sbeller@google.com> writes:
>
> > This is the continuation of f2d48994dc1 (submodule.c: submodule_move_head
> > works with broken submodules, 2017-04-18), which tones down the case of
> > "broken submodule" in case of a missing git directory of the submodule to
> > be only a warning.
>
> After seeing this warning, as we do not do any remedial action in
> this codepath, the user with a repository in this state will keep
> seeing the 'missing' message.  Wouldn't we want to give a hint in
> addition (e.g. 'you can run "git submodule update $name" to
> recover', or something like that)?

Not quite, as this is only triggered in the case of 'old_head = NULL', which is when you have a superproject that is missing the submodule in the working tree before and wants to have it afterwards.

Looking at the test in the previous patch, I would think a reasonable workflow in the test is

    git clone --recurse-submodules super1 super1_clone && cd super1_clone
    git checkout --recurse-submodules historic_state
    # see warning, but checkout proceeds

    git fetch --recurse-submodules
    # clones the submodule as it was configured active via the clone

    git checkout --recurse-submodules historic_state
    # this checkout will put the submodule in place
    #  not sure if -f is needed here, though.

I am currently working on the cloning of submodules that are not currently in the working tree while fetching in the superproject, which would address the latter part.

> The MOVE_HEAD_FORCE codepath that follows this hunk is, eh, already
> forcing to correct the situation, so there is no need to touch that,
> which makes sense, if I understand correctly.

No, that is not executed for now as it depends on 'old_head'.

In case of FORCE we might want to die instead of just warn about that submodule.

Stefan