Re: Regression: submodule worktrees can clobber core.worktree config

From: Stefan Beller <sbeller@google.com>
Date: Jan 10, 2019, 20:07 UTC
Message-ID: <CAGZ79kZ9ibM4eDyK=M6YWEDsjt+JfqJH-Gm56+092VATGuZDaw@mail.gmail.com>
In-Reply-To: <CAG0vfyR3KnDDBrpyG-n-RFbu-xgCLFUa6HUXQ+dk8E4HutR+ow@mail.gmail.com>

Show 9 quoted lines

> I had a look at https://gitlab.com/pclouds/git/commits/submodules-in-worktrees,
> and it doesn't seem to be quite all okay.
>
> The submodule update step of the repro (that breaks the config on 2.20) emits
> an error message instead, and leaves the config unchanged:
>    git -C b2 submodule update
>    fatal: could not set 'core.worktree' to '../../../../../../b2/a'
> It looks a bit like it's still trying to do the wrong thing, but errors out
> during the attempt (repo_config_set_worktree_gently returns false).

There is more than just that. After adding the worktrees, (and after the first status call)

    $ cat b2/.git
gitdir: /u/git/t/trash directory.t7419-submodule-worktrees/b/.git/worktrees/b2
    $ cat b2/a/.git
gitdir: /u/git/t/trash
directory.t7419-submodule-worktrees/b/.git/modules/a/worktrees/a

Are worktrees using absolute path for their gitlinks? Submodules themselves try really hard to use relative path:

    $ cat b/a/.git
gitdir: ../.git/modules/a

> Curiously, even though it says "fatal", it will then perform the actual
> submodule update if it's required.

Oh. :/ I think we should solve that by either warning (but that gives bad UX) or actually aborting, by adding a "|| exit 1" in git-submodule.sh in cmd_update where we call "git submodule--helper ensure-core-worktree".

When we run "git -C b2 submodule update", it calls "git submodule--helper ensure-core-worktree a" which currently would make sure that b2/a/.git points to b2/.git/modules/a, but that is not the case as b2 and b2/a are worktrees, whose git directories are housed in b/.git/worktrees.

So maybe we need to be a bit more careful and check if b2/a/.git resolves to a worktree and if so we'd not touch it at all (and warn about it?).

Show 9 quoted lines

>
> Same behavior on master with a subset of that branch cherry-picked, that is:
> https://gitlab.com/pclouds/git/commit/94751ada7c32eb6fb2c67dd7723161d1955a5683
> along with two others it needed to build:
> https://gitlab.com/pclouds/git/commit/d26ab4c5013f6117814161be3e87c8d2b73561a4
> https://gitlab.com/pclouds/git/commit/b2e21eece6b35e00707ed3a8377a84a95da6b778
>
> --
> Tomasz Śniatowski