[pytest-dev] Testing fixture teardown / finalisation

James Cooke me at jamescooke.info
Wed Jan 6 12:36:28 EST 2021


Hi all,

It's been a while since I've had to write a complicated and robust pytest fixture and I'm struggling with testing the teardown / finalisation code.

Instead of boring you with the [my-work-project] requirements of cleaning up GCS after tests, I'll refer to the fixture in the docs https://docs.pytest.org/en/stable/fixture.html#fixture-finalization-executing-teardown-code :

    @pytest.fixture(scope="module")
    def smtp_connection():
        smtp_connection = smtplib.SMTP("smtp.gmail.com", 587, timeout=5)
        yield smtp_connection  # provide the fixture value
        print("teardown smtp")
        smtp_connection.close()

My (very old) usual strategy for testing teardown of a fixture that uses `yield` would be:

* Start test

    - Instantiate the fixture.

    - Manipulate the instance with `next()` to trigger the teardown.

    - Ensure that teardown was successful - in this case assert that the SMTP connection returned when the fixture was instantiated was closed successfully.

Given that since v4.1 (https://github.com/pytest-dev/pytest/issues/4545), the arrangement step of "Instantiate the fixture" does not work, could someone point me at the recommended method of testing fixture teardown?

I would like to ensure that teardown is resilient and can resolve multiple conditions that can happen in [my-work-project]'s test suite. I can see that there are multiple tests on fixtures in https://github.com/pytest-dev/pytest/blob/48c9a96a03261e7cfa5aad0367a9186d9032904a/testing/python/fixtures.py , are there any preferred / recommended / efficient methods in this file? Alternatively, is there a way to by-pass the fixture wrapping that happens that prevents it being callable? Should I be using `smtp_connection.__pytest_wrapped__.obj()` to instantiate? (this seems bad)

Any pointers / suggestions would be great.

Thanks,

James




More information about the pytest-dev mailing list