Re: [PATCH 0/3] Unconfuse git clone when two branches at are HEAD.

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 9, 2013, 22:26 UTC
Message-ID: <xmqqk3ip3a36.fsf@gitster.dls.corp.google.com>
In-Reply-To: <CBB3BFF2B74D4A768238512FBF1CD6E7@PhilipOakley>

"Philip Oakley" <philipoakley@iee.org> writes:

Show 9 quoted lines

> These look to work OK. Not sure If I've properly covered all the
> options.
>
> A nice feature is that ls-remote will find the fake ref !
>
> $ git ls-remote /c/gitBundleTest1/RepoHEADnomaster.bundle Test:HEAD
> 41ccb18028d1cb6516251e94cef1cd5cb3f0bcb5        refs/heads/Test:HEAD
>
> It's only the clone that barfs (so far) (which could be 'fixed').

You cannot 'fix' the ones deployed in the wild, but I think saying "funny ref" and not aborting is a good compromise. The updated documentation for bundle can mention that error message and explain that the older version of "clone" will say that in what situation.

I didn't notice it the first time, but I think the above would read better and match what has been discussed on the on-the-wire format to swap the order, i.e. use "HEAD:refs/heads/Test" to mean "HEAD is a symref that points at the Test branch".

> Obviously, if it can be made to work, one check would be that the two
> refs (HEAD and refs/heads/wherever) point to the came commit before
> generating the HEAD symref.

Yes, that is a sensible check.