How to access the pytest.mark.xfail reason keyword argument?

[adehad]

I've just migrated a project from Py27 using pytest-html==1.22.1 and am now using pytest-html==3.1.1, I have a plugin that was previously able to get the reason keyword argument using the wasxfail attribute of report, but now this seems to be a boolean, which is understandable.

Is there a suggestion on how to access this reason in a clean way?

This should be a minimal working example, (stripping the other logic my plugin does hopefully):

@pytest.hookimpl(optionalhook=True)
def pytest_html_results_table_header(cells):
    """Change the report table headers."""
    # By default the headers stored in cells are RESULT, TEST, DURATION, LINKS
    result, test, duration, _ = cells
    cells[:] = [
        test,
        result,
        html.th("Manual Tests:"),
        duration,
    ]


def get_manual_tests(item):
    return "\n".join(
        [i.kwargs.get("reason", "") for i in item.own_markers if i.name == "xfail"]
    )


@pytest.hookimpl(optionalhook=True)
def pytest_html_results_table_row(report, cells):
    """Change the report table data."""
    result, test, duration, _ = cells
    cells[:] = [
        test,
        result,
        html.td(report.manual_tests),
        duration,
    ]


@pytest.hookimpl(hookwrapper=True)
def pytest_runtest_makereport(item, call):
    outcome = yield
    report = outcome.get_result()
    report.manual_tests = get_manual_tests(item)

and say have a test file like this :

manual_test = partial(pytest.mark.xfail, strict=False)

class TestCoreFunctions(object):
    @manual_test(reason="T-1, T-2"))
    def test_ui(self):
        pass

[jeffwright13]

I am running Pytest 7.1.1 with Python 3.9.9.

wasxail is a string that gives the reason. It's not a boolean. You can see this as it gets populated by looking at the Pytest source code, file skipping.py, in function pytest_runtest_makereport. However, wasxffail is not visible in every TestReport (which is what your report variable is). wasxfail only gets populated when a test is actually marked xfail, and is only available during the call phase.

How do you develop your plugin? Do you have the ability to set breakpoints in your IDE as your test runs proceed? If so, perhaps you can use that to examine Pytest's internals while running your tests.

That said I'm pretty sure the reason variable should be available for failed tests, but when I searched the Pytest source code I came up with >1000 matches, too many to examine right now!