Re: "warning: Updating the currently checked out branch may cause confusion" on bare repositories

From: Jeff King <peff@peff.net>
Date: Jan 19, 2010, 22:43 UTC
Message-ID: <20100119224343.GA11758@coredump.intra.peff.net>
In-Reply-To: <20100119185906.GF23466@spearce.org>

On Tue, Jan 19, 2010 at 10:59:06AM -0800, Shawn O. Pearce wrote:

Show 12 quoted lines

> Adam Megacz <adam@megacz.com> wrote:
> > "Shawn O. Pearce" <spearce@spearce.org> writes:
> > > That should already be the case.  Did you create the bare repository
> > > by taking a copy of a non-bare's .git directory?  Check your bare
> > > repository's config file and see if core.bare = false, if so set
> > > it to true to signal it really is bare.
> > 
> > Hrm, is there a reason why this is explicitly configured rather than
> > detected?
> 
> I don't know why we do this.  I think its because the guessing
> logic can guess wrong sometimes.

Yup, although I am having trouble finding any actual discussion of when it can go wrong. There is some mention here:

  http://article.gmane.org/gmane.comp.version-control.git/35993

and a little bit here in the 1.5.0 release notes:

  http://article.gmane.org/gmane.comp.version-control.git/39612

But the latter talks about the heuristic being "does it end with .git", which I don't think we actually use anymore. So who knows if the heuristic failures are even trigger-able anymore. I tried looking at the setup code for an answer, but my eyes started bleeding from the horrible spaghetti code and I couldn't make out any of the text.

> Its rare to see it fail.  But I think it can fail if you create
> a file called HEAD in the top level of your source code tree.
> (For example.)

I don't think that would do it. When looking for a git dir we check for a well-formed HEAD, as well as objects and refs directories. The is_bare check seems to be more about "did we actually find a workdir" these days.

-Peff