[openstack-dev] [all][tests] Fix it friday! [mock failure in CI]

Robert Collins robertc at robertcollins.net
Fri Jul 10 08:42:32 UTC 2015


On 10 July 2015 at 20:18, Thierry Carrez <thierry at openstack.org> wrote:
> Robert Collins wrote:
>> Good news everybody, mock 1.1.0 is now out. This backports all the
...
>> 1.1.0 makes that an error, for methods with assert prefixes - unless
>> unsafe is specifically requested. So a big chunk of the failing tests
>> are tests that were not testing anything *at all*.
>> One common exampled of that is 'assert_called' - another method that
>> never ever existed. All our tests using that were testing nothing at
>> all.
>
> It's still non-trivial to fix, especially as fixing it may expose a real
> bug in the never-actually-tested thing.
>
> Releasing this on a Friday sounds like bad timing, especially without
> advance notice that we'd have to rush to fix the dozens of new issues it
> would expose.

There would never be a good time to release such things. The whole
point of the constraints system we're rolling out is to make us not
sensitive to the release schedule of upstreams: we can't dictate the
release schedule of the world to be convenient to our cycles.

*Further to that* we've already rolled out the constraints system for
devstack, so this is not a gate wedge. Its per-project in unit tests,
and it can be fixed one project at a time. There's no rush that makes
this worse for anyone than if it came out on Monday. We don't need [at
least so far as I can tell - maybe the neutron AAS's will be
interlocked - but again, that would happen any day] infra intervention
to force-push stuff in anywhere.

The one thing that we may need, for py26 projects, is for this review:
https://review.openstack.org/#/c/200344/ - to get in asap, for any
projects with python2.6 jobs, since 1.1.0 isn't compatible with 2.6.
That will let folk with 26 jobs update their requirements locally and
easily, without tox.ini hacks.

> Now that we have data[1] showing what fails, it's not completely crazy
> to pin it and buy us some time ?

Pinning will be per repo, because its unittests not devstack. So its
doable if we have to, but since we have to land a patch in the
repositories anyway, trying for a fix first would be good IMO.

-Rob

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



More information about the OpenStack-dev mailing list