[ovirt-devel] Re: [OST] Testing master suite on using python 2.7?!
On Wed, Nov 18, 2020 at 6:40 PM Nir Soffer <nsoffer(a)redhat.com> wrote:
I'm trying to add a test module for image transfer:
https://gerrit.ovirt.org/c/112274/
The test use concurrent.futures module form the standard library.
This module is not available in python 2.7 but we don't support 2.7 in master
and it was EOL since Jan 2020.
The test fail when starting the suite:
[2020-11-18T16:17:10.526Z] ============================= test session
starts ==============================
[2020-11-18T16:17:10.526Z] platform linux2 -- Python 2.7.5,
pytest-4.6.9, py-1.9.0, pluggy-0.13.1 -- /usr/bin/python2
...
[2020-11-18T16:17:11.105Z] ==================================== ERRORS
====================================
[2020-11-18T16:17:11.105Z] ___ ERROR collecting
basic-suite-master/test-scenarios/008_image_transfer.py ___
[2020-11-18T16:17:11.105Z] ImportError while importing test module
'/home/jenkins/agent/workspace/ovirt-system-tests_standard-check-patch/ovirt-system-tests/basic-suite-master/test-scenarios/008_image_transfer.py'
[2020-11-18T16:17:11.105Z] Hint: make sure your test modules/packages
have valid Python names.
[2020-11-18T16:17:11.105Z] Traceback:
[2020-11-18T16:17:11.105Z]
../basic-suite-master/test-scenarios/008_image_transfer.py:22: in
<module>
[2020-11-18T16:17:11.105Z] import concurrent.futures
[2020-11-18T16:17:11.105Z] E ImportError: No module named concurrent.futures
Should we use pytest.skip() to skip this test when running on python 2?
Or just remove the python 2 build, I don't have any idea why we run master code
python 2.
I added skip for python 2, but even with python 3 job, we need the
ovirt-imageio-client
package, available only on python 3.
We need to change ci to use different builds for python 2 and 3, or
drop the python 2
builds from the master suite.
In the current state we cannot test image transfer in the CI, only
locally. Using local
OST is easy and reliable. even with nested setup.
I think we can solve this with markers - instead of collecting all the tests,
we can use:
@pytest.mark.ci
def test_that_works_in_ci():
...
def test_that_does_not_work_in_ci():
...
The ci job can run:
pytest -m "ci"
So it picks only tests that can run in the ci environment.
What do you think?