Open Viech opened 1 month ago
I'm on my phone right now, but you can look at the workflow definition to see how the tests are run there.
Alright, I had a chance to check the workflows myself. We actually do not run any tests, we only build the wheels. I guess, the reason being that GLPK itself is tested and this is a very thin wrapper around it. Additionally, we do use this package for the tests in optlang, so we would notice breaks there.
Not ideal, but also not terrible. If you feel like it, you are welcome to contribute tests.
I was explicitly asked by a repository maintainer to run existing tests within my Arch User Repository package. They probably saw the test_swiglpk.py file or the reference to nosetests
in tox.ini.
So I don't personally need a testbench; this issue is just about the fact that existing tests do not seem to be functioning.
@cdiener any ideas? I can't prioritise fixing the tests at the moment.
Yeah, the tests are ancient and are not used in the CI at the moment. The reason is that we basically test the swigplk API with the optlang and cobrapy tests which are already really comprehensive. Not really clean, but nobody ever updated the tests here.
Running just
nosetests
afterpython setup.py build
fails:So does the following attempt using
PYTHONPATH
:The same goes for
python -m unittest
:Again with
PYTHONPATH
:What is the intended way to run the tests?