Use a float for the tolerance in the timer tests

[llucax]

Hopefully this finally fixes the flaky hypothesis tests.

• Remove deprecated uses of pytest-asyncio
• Use a float for the tolerance in the timer tests
• Revert "Use less extreme values for min and max timedelta in tests"

[llucax]

Tests failed again when queuing a PR, so here is another attempt.

[daniel-zullo-frequenz]

Tests failed again when queuing a PR, so here is another attempt.

Hypothesis usually tells you how to reproduce the error by temporarily adding @reproduce_failure({PARAMETERS, FOR, THE, TEST, HERE}) as a decorator on the test case. Have you tried that to be sure the patch solves the issue?

[llucax]

Good tip. I actually validated it manually, but for a previous attempt and forgot to do it with the new approach. I will check with the decorator 👍 💯

[llucax]

FYI, this was the failure:

______________________ test_policy_skip_missed_and_drift _______________________

    @hypothesis.given(
>       tolerance=st.integers(min_value=0, max_value=_max_timedelta_microseconds),
        **_calculate_next_tick_time_args,
    )

tests/test_timer.py:148: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

tolerance = 171726190479152817, now = 171726190479152817
scheduled_tick_time = -1, interval = 1

    @hypothesis.given(
        tolerance=st.integers(min_value=0, max_value=_max_timedelta_microseconds),
        **_calculate_next_tick_time_args,
    )
    def test_policy_skip_missed_and_drift(
        tolerance: int, now: int, scheduled_tick_time: int, interval: int
    ) -> None:
        """Test the SkipMissedAndDrift policy."""
        hypothesis.assume(now >= scheduled_tick_time)
    
        next_tick_time = SkipMissedAndDrift(
            delay_tolerance=timedelta(microseconds=tolerance)
        ).calculate_next_tick_time(
            now=now, interval=interval, scheduled_tick_time=scheduled_tick_time
        )
        if tolerance < interval:
            assert next_tick_time > now
        drift = now - scheduled_tick_time
        if drift > tolerance:
>           assert next_tick_time == now + interval
E           assert 0 == (171726190479152817 + 1)
E           Falsifying example: test_policy_skip_missed_and_drift(
E               tolerance=171_726_190_479_152_817,
E               now=171_726_190_479_152_817,
E               scheduled_tick_time=-1,
E               interval=1,  # or any other generated value
E           )

tests/test_timer.py:166: AssertionError

[daniel-zullo-frequenz]

FYI, this was the failure

I see in this case hypothesis only mentioned how to reproduce the error/falsify the example.

[llucax]

Yeah, I don't know how @reproduce_failure is used exactly, but example() works pretty well. Adding some examples, although I can't test the falsifying example precisely because we switched from int to float for the problematic value.

With this in mind, it seems like it fixes the issue. Pushed some updates, adding the examples so they are always tested just in case.