Ci differences

[liamhuber]

In pyiron_workflow I'm running into a problem where the tests hang, and when cancelled I ultimately get pympipool complaints that the test files were not importable modules.

I found these lines where "." is getting added to the pythonpath (similar for mpiexec), and notice that pympipool workflows follow a pattern of cd tests; python -m unittest discover ., while the centralized workflows skip the CI and simply invoke things in the pattern python -m unittest discover tests. From my experiments in pyiron_workflow it looks like this is what's making the difference.

Here I want to double check that hypothesis and test two-work arounds. If the hypothesis is correct, I'll open a corresponding issue.

[liamhuber]

Unittests_by_path fails as anticipated:

Traceback (most recent call last):
  File "/home/runner/work/pympipool/pympipool/pympipool/backend/mpiexec.py", line 90, in <module>
    main()
  File "/home/runner/work/pympipool/pympipool/pympipool/backend/mpiexec.py", line 46, in main
    input_dict = interface_receive(socket=socket)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/share/miniconda3/envs/test/lib/python3.11/site-packages/pympipool/shared/communication.py", line 147, in interface_receive
    return cloudpickle.loads(socket.recv())
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'test_meta'
Error: The action has timed out.

Here it sensibly has a timeout set and fails cleanly instead of hanging forever, which is a minor improvement I can learn for the pyiron_workflow tests.

Very nicely, both versions of my "patch" to the environment passed and facilitate discovering the tests by a path.

This can be closed, but I'll leave it open until the corresponding issue is resolved in case that has any impact on viewing the test logs.