Open Stack

On 9 April 2013 08:51, Clark Boylan <clark.boylan at gmail.com> wrote:
> On Mon, Apr 8, 2013 at 1:45 PM, Kevin L. Mitchell
> <kevin.mitchell at rackspace.com> wrote:
>> On Tue, 2013-04-09 at 08:36 +1200, Robert Collins wrote:
>>> Indeed, and sorry about that. The cause was a new release of subunit,
>>> which added the new version 2 wire protocol. I had tested the whole
>>> ecosystem for compatibility forward and backwards, and the specific
>>> case of 'new subunit old testrepository' was known to break *unless* a
>>> subunit-2to1 filter was put in the pipeline; when I discussed that
>>> with the -infra folk they didn't think we'd have any issues, because
>>> the new testrepository would be brought down with everything else due
>>> to the PyPI mirror : sadly that global install mentioned in the review
>>> wasn't pulling from the PyPI mirror, and so it sucked down a new
>>> subunit on it's own.
>>
>> That explains it.  However, I have to say that the problem still
>> persists…
>
> Assuming the failures for
> http://logs.openstack.org/26310/2/check/gate-python-novaclient-python27/1404/
> are the same failures you are seeing I don't think this is related to
> the testrepository and python-subunit changes. That looks like a real
> test failure to me and the html report is being generated. Also,
> looking at the output from pip freeze we seem to still be using the
> old version of python-subunit and testrepository (our pypi mirror has
> not pulled in the new goodness yet).
>
> Clark

That is definitely not the subunit change. Not sure what would cause
the header to be added when it wasn't previously.

-Rob

-- 
Robert Collins <rbtcollins at hp.com>
Distinguished Technologist
HP Cloud Services