From: Shawn O. Pearce <spearce@spearce.org>
Date: Sat, 14 Mar 2009 00:13:45 GMT
Subject: Re: [JGIT PATCH 0/6] Add tests for RevWalk and its supporting code
Message-ID: <20090314001345.GI22920@spearce.org>
In-Reply-To: <20090313223933.GH22920@spearce.org>

"Shawn O. Pearce" <spearce@spearce.org> wrote:
> These new tests cover some of the common cases we see with using a
> RevWalk, and increase our code coverage in this critical area of
> the JGit library.
> 
> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
> ---
>  Robin Rosenberg <robin.rosenberg.lists@dewire.com> wrote:
>  > fredag 13 mars 2009 03:07:37 skrev "Shawn O. Pearce" <spearce@spearce.org>:
>  > > Today I uncovered some ugly cases with "jgit rev-list B ^A", where
>  > > some commits reachable from A were still being output, even though
>  > > we asked that they be excluded.
>  > 
>  > How about a test suite to prove this is better than before?
> 
>  Its better.  :-)

Its actually not better.  Although the tests I'm replying to are
right, there's another bug introduced by this series that they
don't detect, but that shows up on the Linux kernel repository.

I have yet to track down the full thing... but I know its busted.
 
-- 
Shawn.