Re: What's cooking in git.git (Jan 2012, #08; Tue, 31)

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 10, 2012, 04:09 UTC
Message-ID: <7vhayzs51s.fsf@alter.siamese.dyndns.org>
In-Reply-To: <7vsjijs9rq.fsf@alter.siamese.dyndns.org>

Junio C Hamano <gitster@pobox.com> writes:

Show 9 quoted lines

> It turns out that this series, as a side effect, fixes a long-standing bug
> that the --branch option cannot be used with the --mirror option.
>
> I am tempted to merge it also to 1.7.9.1 maintenance track. The issue is
> minor (you can always clone with --mirror and then switch the branch with
> "checkout") and it does not look like it warrants further backpointing to
> 1.7.8 and older releases, though.
>
> Comments and/or objections?

Replying to myself.

Regardless of "'refs/heads/$branch' is not found because we do not look for in the right place" issue, shouldn't the use of the --mirror without the --bare option forbidden or at least warned about?

Cloning as --mirror, with a working tree with a current branch, would mean the next fetch would directly try to update it by overwriting. We do pass the --update-head-ok from pull, so "git pull" in such a repository should be "sort of" safe, as long as the repository is never used to create its own commits on whatever local branches set to be overwritten by --mirror, but still I have this gut feeling that such an arrangement should at least be discouraged.

Comments and/or opinions?