Re: [PATCH] git-p4.py: Make submit working on bare repository

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 21, 2016, 07:36 UTC
Message-ID: <xmqqbn7aa522.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <87fuwnd4u7.fsf@freja.aidecoe.name>

Amadeusz Żołnowski <aidecoe@aidecoe.name> writes:

Show 27 quoted lines

> To submit changes from master branch to Perforce, new commits should be
> based on a remote branch p4/master:
>
> (1)
>                   E'---F' master
>                  /
>     A---B---C---D p4/master
>
> Commits from A to D map to Perforce changelists. These commits include
> additional metadata in commit message which most important is changelist
> number.
>
> On submit git-p4 prepares changelists for commits E'-F' and submits
> these to Perforce repository. After this operation it syncs back remote
> branch p4/master. This is the common part for both bare and non-bare
> repository.
>
> (2)
>                   E'---F' master
>                  /
>     A---B---C---D---E---F p4/master
>
> In non-bare repository git rebase is performed and it results in
> following state:
>
> (3)
>     A---B---C---D---E---F p4/master, master

That is because you would want your further work to be built on top of F, not F'.

> In bare repository this operation cannot be performed, so it remains in
> state (2).

And that is what made me wonder.

> With special care state (2) can be transformed to state (3) with manual
> update of refs/heads/master with refs/remotes/p4/master.

What do you mean by transforming with special care?  Do you mean running

    $ git update-ref master remotes/p4/master

or does it involve something else? Is it a fairly involved procedure, I wonder if it makes sense to just silently allowing submit without rebasing, and without telling the user how to clean it up.

And if it is really is the matter of update-ref, wouldn't that procedure apply equally well to a submit made from a non-bare repository?

Show 7 quoted lines

> I understand that implementing rebase for bare repository is unsafe and
> it wouldn't be appreciated. Therefore we have to resort to such a
> barbarity and git-p4 submit shouldn't attempt to perform a rebase
> operation itself. Especially because it performs other operations before
> and we end up in the intermediate state (2) anyway.
>
> Is this explanation satisfactory?

Not quite yet. It is better than no explanation, but it does not quite convince me (a non-git-p4 user) that it makes sense to allow submitting in a bare repository in the first place.

This is not an informed objection; I want to first understand why doing this is safe and does not hurt the end users.