Bump pytest and pytest-asyncio

[PGijsbers]

SubRequest was removed, but it was really used to denote a FixtureRequest, so I updated that part and now everything works on the latest versions again.

edit: with pytest 8+, test ordering changed which exposed abuse of global state in tests, so I had to remove that too.

[PGijsbers]

Whoops, I accidentally branched off the wrong branch.. rebasing.

[PGijsbers]

Otherwise Pytest will complain it cannot collect from this class since it has an (implicit) __init__ method.

[PGijsbers]

I am a bit confused why some tests failed.. they pass locally. Trying to look into it. Have you experienced this before? Are there flaky tests (on CI)?

[PGijsbers]

Closing and reopening to see if that triggers CI.

[PGijsbers]

The test fails when run in the entire suite, not when run in isolation. Have narrowed down to the specific version change of pytest 7.4.4->8.0.0 with pytest-asyncio fixed at 2.3.8.

[PGijsbers]

I'm still not entirely sure what's happening, it looks like a different responses mock is used than intended in the authentication tests... and the tests results depend on execution order. I started down this path since it was a breaking change of pytest 8.0.0, verified by renaming test_authorization.py to auth_test.py so it is executed earlier (alphabetical order..), we verify the tests do work then. So the tests somehow violate independence, but I am not sure yet how since the auth tests seem OK (all work is within the scope of a context manager).

[PGijsbers]

The problem was practically every other test modifying global state.
I patched it for now to where they at least revert the change after the test is ran, but perhaps we should redesign the use of global variables. That would be a separate issue.

[Taniya-Das]

Looks correct to me.