Re: git rebase bug?

From: Jakub Narebski <jnareb@gmail.com>
Date: Jul 7, 2010, 21:44 UTC
Message-ID: <m3bpajm0gw.fsf@localhost.localdomain>
In-Reply-To: <20100707205126.GA11240@glandium.org>

Mike Hommey <mh@glandium.org> writes:

> On Wed, Jul 07, 2010 at 08:00:04PM +0200, Björn Steinbrink <B.Steinbrink@gmx.de> wrote:
> > On 2010.07.07 17:05:45 +0200, Mike Hommey wrote:

Show 14 quoted lines

> > > See how the security/manager/ssl/public/nsIBadCertListener.idl file that
> > > was created by the original patch is created as
> > > xulrunner/examples/simple/content/contents.rdf.
> > 
> > The "problem" is that nsIBadCertListener.idl wasn't actually created by
> > the cherry-picked commit, but was modified. It was an empty file before,
> > created in 4292283190983fa91b875e22664a79a3aa9ea45d.
> > 
> > And as nsIBadCertListener.idl is missing from the xulrunner/2.0 branch,
> > git does the usual rename detection, finding another empty file and ends
> > up patching that one instead.
> 
> Oh, makes sense. Thanks. So that's a quite troubling corner case...
> I wonder if empty files shouldn't be special cased...

Well, similarity score (of contents and of filename) is weighted by contents length, but perhaps empty files / zero length somehow fall out as an edge case...

I agree that empty files should be special cased... unless filename is _very_ similar.

-- 
Jakub Narebski
Poland
ShadeHawk on #git