Re: [PATCH 3/3] http-push: update tests

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jan 17, 2009, 19:37 UTC
Message-ID: <alpine.DEB.1.00.0901172035000.3586@pacific.mpi-cbg.de>
In-Reply-To: <7vsknh7og5.fsf@gitster.siamese.dyndns.org>

Hi,

On Sat, 17 Jan 2009, Junio C Hamano wrote:

Show 9 quoted lines

> The original seems to want the push to succeed, and also it wants the 
> file refs/heads/master to be present after the push (presumably because 
> there should be that ref when the push succeeds).  If you fixed "push" 
> that used to fail to succeed, that is great, and s/failure/success/ is a 
> good thing.
> 
> But you are removing something else without explanation.  Why do you 
> need to remove the part of the test that checks if refs/heads/master is 
> present? Is it looking for a file in a wrong place?

As I mentioned with two other patches, the push does not succeed, and that is the reason for the "failure" in test_expect_failure.

It does not succeed for two reasons:

- due to an off-by-path_len bug, xmalloc() tries to allocate ~4GB of 
  memory, which is a bit much, so http-push die()s with an OOM.

- even with that fix, the push fails because it cannot find any common 
  refs.  It cannot find them because it does not download info/refs as it 
  is supposed to do, but it looks through refs/, missing the fact that the 
  refs are packed (which it cannot handle).

Ciao, Dscho