[dev][infra][qa][tact-sig] Tox 4.0.0 breaking changes

Ghanshyam Mann gmann at ghanshyammann.com
Thu Dec 8 21:16:04 UTC 2022


 ---- On Thu, 08 Dec 2022 13:01:12 -0800  Jeremy Stanley  wrote --- 
 > On 2022-12-08 12:44:17 -0800 (-0800), Ghanshyam Mann wrote:
 > [...]
 > > It seems master jobs using tox<4[1] but all stable jobs using
 > > tox>=4.0[2] and failing. And both using ensure-tox
 > > 
 > > [1] https://zuul.opendev.org/t/openstack/build/8a4585f2961a4854ad96c7d2a188b557/log/controller/logs/pip3-freeze.txt#235
 > > [2] https://zuul.opendev.org/t/openstack/build/78cddc1180ff40109bbe17df884d23d8/log/controller/logs/pip3-freeze.txt#227
 > [...]
 > 
 > Those are Tempest jobs, which do some of their own installing of tox
 > directly rather than just relying on the version provided by
 > ensure-tox. Check with the QA team, since they've been working on
 > fixes for those.

I was debugging that bug only.

I do not think it is Tempest's own installation. It is a devstack installation of tox (which I think uses ensure-tox
on stable also unless we have any hidden installation overriding it) which is same on master as well
as on stable but the master installation uses old tox which I am hoping due to ensure-tox capping tox
but on stable it is not. I am confused by this different behaviour. I do not think devstack does any
separate things for master and stable/zed for tox installation. Something from ensure-tox? 


 > 
 > 
 > > Does it mean all the stable jobs also will start using the latest
 > > tox (>4.0) which is breaking the jobs and we need to fix them in
 > > stable branches too?
 > [...]
 > 
 > Unless we change those jobs to pin to an old version of tox for
 > stable branches, yes.

Ok, that means ensure-tox capping and uncapping are not useful for
stable branch jobs. We should cap it in constraints also? or any specific installation
of tox. Because we should not force stable branch jobs to use latest tox and break them
and we fix them all.

=gmann

 > -- 
 > Jeremy Stanley
 > 



More information about the openstack-discuss mailing list