Also emit PytestUnknownMarkWarning for unknown marks given in -m

[The-Compiler]

If we use @pytest.mark with a mark that does not exist, pytest warns about it:

import pytest

@pytest.mark.one
def test_one():
    pass

results in:

  .../test_x.py:3: PytestUnknownMarkWarning: Unknown pytest.mark.one - is this a typo?  You can register custom marks to avoid this warning - for details, see https://docs.pytest.org/en/stable/how-to/mark.html
    @pytest.mark.one

This is to avoid silent failures, e.g. a test being marked onne and then never being executed.

---

However, a similar issue exists with -m too! If, with the code above, we use pytest -m onne, that results in:

============================= test session starts ==============================
collected 1 item / 1 deselected / 0 selected                                   

============================ 1 deselected in 0.43s =============================

Thanks to exit code 5 (ExitCode.NO_TESTS_COLLECTED), this can still be spotted, more or less. However, when using e.g. pytest -m "slow or onne", it's not trivially visible anymore that a typo was made (assuming that some tests run due to the slow mark, of course).

Is there a good reason to not validate all marks passed into -m against the list of registered marks as well? Have we had an issue about this before? This gets pointed out every once in a while by participants in my pytest trainings, and I feel like I must have reported it before, but I can't seem to find anything.

[poulami-sau]

Hello! If I'm understanding correctly, I think you reported this same issue in #10514? I am interested in working on this issue, but I'm having a hard time trying to simulate the conditions, and would appreciate some guidance on this!

[The-Compiler]

Hah, so my memory didn't fail me, and I just didn't find it 🙈. Thanks, closing this one as a duplicate then. I'll answer the rest over there.

[The-Compiler]

Duplicate of #10514