Locate actual wheels in WHEELS_DIR

[kalikiana]

• The wheels need to be stored in a way that they can be modified easily rather than pool which is volatile.
• By default the pool is used as always.
• When cloning the path is explicitly logged.

Fixes: https://progress.opensuse.org/issues/134390

[perlpunk]

t/08-autotest.t and t/14-isotovideo.t failing
https://github.com/os-autoinst/os-autoinst/actions/runs/6022113292/job/16336207154?pr=2356#step:3:930

[perlpunk]

https://github.com/os-autoinst/os-autoinst/actions/runs/6037760508/job/16382651406?pr=2356#step:3:894

Cannot read t/data/wheels_dir/writer/lib/Copy/Writer/Content.pm: No such file or directory at /usr/lib/perl5/site_perl/5.38.0/Devel/Cover/Report/Codecovbash.pm line 76.

I think t/data/wheels_dir needs to be added to this ignore:
https://github.com/os-autoinst/os-autoinst/blob/master/tools/invoke-tests#L47

[perlpunk]

I think we need to test the case where the wheel would be located in the pool dir as well. As I wrote in my comment, this would fail with the PR in its current form.

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (d8ad077) 95.07% compared to head (b3ebf1e) 95.08%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2356   +/-   ##
=======================================
  Coverage   95.07%   95.08%           
=======================================
  Files         155      155           
  Lines       15338    15367   +29     
=======================================
+ Hits        14582    14611   +29     
  Misses        756      756           

Files Changed	Coverage Δ
OpenQA/Isotovideo/Runner.pm	100.00% <100.00%> (ø)
OpenQA/Isotovideo/Utils.pm	100.00% <100.00%> (ø)
autotest.pm	80.68% <100.00%> (+0.16%)	⬆️
t/14-isotovideo.t	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kalikiana]

Ah! I forgot to add the docs update. Will push shortly.

[okurz]

https://github.com/os-autoinst/os-autoinst/actions/runs/6015946938/job/16318885617#step:3:647 shows that previously the test took 62s. Now in https://github.com/os-autoinst/os-autoinst/actions/runs/6147257051/job/16678324796?pr=2356#step:3:647 75s. That is a runtime increase that I consider too big to be acceptable.

I created
https://progress.opensuse.org/issues/135521 for t/14-isotovideo.t running into timeouts already in local executions.

[kalikiana]

https://github.com/os-autoinst/os-autoinst/actions/runs/6015946938/job/16318885617#step:3:647 shows that previously the test took 62s. Now in https://github.com/os-autoinst/os-autoinst/actions/runs/6147257051/job/16678324796?pr=2356#step:3:647 75s. That is a runtime increase that I consider too big to be acceptable.

I created https://progress.opensuse.org/issues/135521 for t/14-isotovideo.t running into timeouts already in local executions.

It's not much slower given for what it does, but fair enough. I checked again if we can avoid calling isotovideo here - and eventually realized that we can't test that libs from a wheel are loaded any other way.
So I'm going the route of splitting the test out into another file.

[perlpunk]

I checked again if we can avoid calling isotovideo here - and eventually realized that we can't test that libs from a wheel are loaded any other way.

I'm wondering about that. Maybe with the right mocking it can be done?

[kalikiana]

I checked again if we can avoid calling isotovideo here - and eventually realized that we can't test that libs from a wheel are loaded any other way.

I'm wondering about that. Maybe with the right mocking it can be done?

My concern was more to do with being able to reproduce the loading of the libraries. Good news is I was able to reproduce it via load_test_schedule. Bad news is it doesn't in fact add the inc path.

[perlpunk]

I played with it and got this working: perlpunk@cffabc6

• I had to replace glob with File::Glob::bsd_glob and don't really know why. The normal glob had problems on the pool directory.
• Instead of mocking checkout_git_repo_and_branch, which does the actual necessary push to @INC, I only mock clone_git
• I created two different sets of wheels because once a certain Package::Name is loaded, perl won't load it again (unless you cheat and remove it from %INC). I decided to make it distinguishable with the added numeric suffix.

edit: and faster 7: [22:33:11] ./t/14-isotovideo.t ..................................... ok 58049 ms than before :)

[okurz]

7: [12:06:39] ./t/14-isotovideo.t ..................................... ok    59521 ms ( 0.00 usr  0.00 sys + 39.51 cusr  2.45 csys = 41.96 CPU)