Continuious Integration platform for Mercurial

[Mads Kiilerich]

On 03/15/2010 09:52 AM, Pierre-Yves David wrote:
> On Fri, Mar 12, 2010 at 11:32:21AM -0700, Bill Barry wrote:
>> is the failure on crew/standard/2.4 normal?
>>
>> http://apycot.hg-scm.org/testexecution/91397?tab=hgtest
>>
>> How would that even happen? Something runs in different order when
>> executing standard vs pure? Or is the test turned off in pure?

It is a new test. Perhaps the test is too verbose and thus relies on 
something that is undefined. A > /dev/null could fix the failure - but I 
don't know if it would reduce the test coverage.

> This Failure is not "normal". This test is not "stable" and randomly fails.
> They must be a race condition that triggered in specific case.
>
> Two test have stability issue :
>      - test-clone-pull-corruption (failing here).
>      - test http.

Could that be because tests are being run in parallel and tries to use 
the same port numbers?

http://www.selenic.com/pipermail/mercurial-devel/2010-February/018513.html 
is an attempt to mitigate that.

But perhaps it would be better if tests always choose ports 
automatically as introduced in d3f27d15c9cb and the following cleanups.

/Mads