From: Jeff King <peff@peff.net>
Date: Thu, 16 Apr 2009 07:06:26 GMT
Subject: Re: [PATCH 4/5] archive: do not read .gitattributes in working directory
Message-ID: <20090416070626.GB20071@coredump.intra.peff.net>
In-Reply-To: <7vhc0pnpn3.fsf@gitster.siamese.dyndns.org>

On Wed, Apr 15, 2009 at 10:17:36PM -0700, Junio C Hamano wrote:

> > +	/* tar-tree defaults to fix-attributes as before */
> > +	nargv[nargc++] = "--fix-attributes";
> > +
> >  	switch (argc) {
> >  	default:
> >  		usage(tar_tree_usage);
> 
> I screwed up this part; nargv[] array needs to be enlarged by one element
> because of this change.  It resulted in a funny breakage in tests that
> triggered only when run from the toplevel of the tree but did not surface
> when the individual test was done from t/ directory, which my final
> testing on the k.org machine caught, and that is why I still haven't
> managed to push the result out for tonight.

FWIW, running t5000 with --valgrind does catch it. I'll try to run the
full test suite with valgrind on the last -rc before 1.6.3 to catch any
hidden issues which have cropped up during this cycle (I can do it
earlier, too, but it is a real pain to run the whole thing, so I want to
just wait until the last minute).

-Peff