Schedule atexit failure at import time

[giampaolo]

Hello.
In psutil tests I have the following idiom:

import psutil, pytest, traceback

try:
    HAS_CPU_FREQ = hasattr(psutil, "cpu_freq") and bool(psutil.cpu_freq())
except Exception:
    traceback.print_exc()
    HAS_CPU_FREQ = False

class TestCase:
    @pytest.mark.skipif(not HAS_CPU_FREQ, reason="not supported")
    def test_cpu_freq(self):
        ...

Assuming this makes sense in the first place, would it make sense to give the ability to schedule a failure occurring at import time, so that it gets reported when the test run has finished? Something like:

pytest.fail_atexit(reason)  # runs at the end of the cleaning run.

For the record, I later switched to:

try:
    HAS_CPU_FREQ = hasattr(psutil, "cpu_freq") and bool(psutil.cpu_fre())
except Exception:  # noqa: BLE001
    atexit.register(functools.partial(print, traceback.format_exc()))
    HAS_CPU_FREQ = False

This at least prints a traceback on exit [1], so I will generally notice, but what would be needed here is an actual failure .

[1] unless I run tests in parallel, in which case the stdout is suppressed

[The-Compiler]

For more complex things like this, I'd use a fixture that uses pytest.skip instead:

@pytest.fixture
def ensure_cpu_freq():
    if not (hasattr(psutil, "cpu_freq") and bool(psutil.cpu_freq())):
        pytest.skip("cpu_freq not supported")

and then possibly

@pytest.mark.usefixtures("ensure_cpu_freq")
def test_cpu_freq():
    ...

which will now skip the test if it's not available, and fail the test if there is an exception.

Also I'm converting this to a discussion, as I don't think there is anything actionable for pytest here.