Re: bug?: git reset --mixed ignores deinitialized submodules

From: Stefan Beller <sbeller@google.com>
Date: Mar 13, 2017, 17:51 UTC
Message-ID: <CAGZ79kZmWaobW9e4iPY05y0N6PLcFphGnZmDHtrGKeV0Up70vg@mail.gmail.com>
In-Reply-To: <1489180018.10192.45.camel@novalis.org>

On Fri, Mar 10, 2017 at 1:06 PM, David Turner <novalis@novalis.org> wrote:

Show 25 quoted lines

> Git reset --mixed ignores submodules which are not initialized.  I've
> attached a demo script.
>
> On one hand, this matches the documentation ("Resets the index but not
> the working tree").  But on the other hand, it kind of doesn't: "(i.e.,
> the changed files are preserved but not marked for commit)".
>
> It's hard to figure out what a mixed reset should do.  It would be
> weird for it to initialize the submodule.  Maybe it should just refuse
> to run?  Maybe there should be an option for it to initialize the
> submodule for you?  Maybe it should drop a special-purpose file that
> git understands to be a submodule change?  For instance (and this is
> insane, but, like, maybe worth considering) it could use extended
> filesystem attributes, where available.
>
> #!/bin/bash
> mkdir demo
> cd demo
>
> git init main
>
> (
>         git init sub1 &&
>         cd sub1 &&
>         dd if=/dev/urandom of=f bs=40 count=1 &&

We prefer reproducability in tests, so if we take this into
a (regression) test later we may want to
    s/dd.../echo "determinism!" >f/

Show 8 quoted lines

> # commit that change on main,  deinit the submodule and do a mixed
> reset
> (
>         cd main &&
>         git add sub1 &&
>         git commit -m 'update sub1' &&
>         git submodule deinit sub1 &&
>         git reset --mixed HEAD^ &&

As of now most commands (including reset) are unaware of submodules to begin with. They are ignored in most cases, i.e. git-status has some tack-on to (pseudo-)report changes in submodules, but it's not really builtin.

A submodule that is not initialized ( i.e. submodule.<name>.url is unset) ought to not be touched at all. This is spelled out in the man page for "submodule update" only at this point.

>         git status # change to sub1 is lost

The change is not really lost, as you can get it via

    git checkout HEAD@{1}
    git submodule update --init

This works most of the time, but it is unreliable as the submodule may have had some gc inbetween which threw away important objects.

Steping back a bit, rereading the subject line, what do you think is the bug here?

* git-status not reporting about uninitialized submodules?
* git reset --mixed not touching the submodule worktree
* lack of --recurse-submodules in git-reset? (and that not
  being default, see prior point)
* submodules being in detached HEAD all the time?

Thanks, Stefan