Skipped tests have different phases depending on if skip is implemented with pytest.mark.skip decorator, or skipped in the test with pytest.skip

[jeffwright13]

I have two different tests that implement skipping in different ways:

def test_d_skip():
    pytest.skip("skipping this test")

@pytest.mark.skip
def test_5_marked_SKIP(log_testname):
    assert 1

What I notice is that only the 'setup' and 'teardown' phases exist if a test is skipped with the pytest.mark.skip decorator; but all three phases exist ('setup', 'call' and 'teardown') if it is skipped with the pytest.skip statement inside the test.

Is there any specific reason for this difference in behavior?

[symonk]

I guess in order to hit the skip the test item is actually executed so having a 'call' phase seems alright for that use case, a test could fail before the skip for example and only skip under some logic.

[jeffwright13]

Yeah but wouldn't the result be the opposite from what I see? The decorator could be used by Pytest as a way to know that it doesn't even need to execute the test, so the call phase could be skipped. Whereas the test case which calls pytest.skip inside the test definition would need to be executed (and hence have a call phase) before it knew to skip.

[RonnyPfannschmidt]

It's behaviour as expected, the call phase only happens when the setup passes

[jeffwright13]

Thanks for the response. Still don't quite get it but the good news is, this forced me to rethink how tests are classified in my own code, and I just realized that instead of trying to recreate the complex logic of Pytest in my own after-test results processing, I can just leverage the -v option and categorize test results on how Pytest marks them in the verbose output. And that way if anything changes in the future I won't have to redo my code.