From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Tue, 08 Nov 2005 11:54:11 GMT
Subject: Re: Errors cloning over http -- git-clone and cg-clone fail to fetch a reachable object...
Message-ID: <Pine.LNX.4.63.0511081252070.2649@wbgn013.biozentrum.uni-wuerzburg.de>
In-Reply-To: <46a038f90511071837g474bdc44vf60dd0758511f24c@mail.gmail.com>

Hi,

On Tue, 8 Nov 2005, Martin Langhoff wrote:

> On 11/8/05, Nick Hengeveld <nickh@reactrix.com> wrote:
> 
> > You might try this to see exactly what request/response headers
> > curl thinks are passing back and forth.
> 
> It definitely looks like it's doing a few requests in parallel and
> getting them mixed up. BTW, this repo is public and sitting on a box
> that doubles up as kernel mirror -- feel free to hit it ;-)

It could be a similar problem to what I experienced in http-push: 
Uninintialized memory due to the code path. Can you run

	valgrind --leak-check=yes --show-reachable=yes git-http-fetch...

and check for jump depending on uninitialized memory?

Hth,
Dscho