The documentation for pytest.mark.xfail() does not match reality.
reason=None: If no conditions are passed, reason’s default is "" (code). Adding @pytest.mark.xfail(reason=None, strict=True) to a passing test results in “TypeError: can only concatenate str (not "NoneType") to str”. If a condition is passed, passing no reason behaves like passing None to it (code). I’m unsure how this subtlety could be captured in the function signature. pytest.mark.skipif() has the same issue. pytest.mark.skip()’s default value for reason is in fact "unconditional skip" (code).
raises: The type (according to the type annotations) is Union[Type[BaseException], Tuple[Type[BaseException], ...]] instead of Type[Exception]. Neither captures the fact that None can be passed to it (which is also the documented default value).
strict=False: The default value of strict is actually determined by the xfail_strict config (code).
The documentation for
pytest.mark.xfail()does not match reality.reason=None: If no conditions are passed,reason’s default is""(code). Adding@pytest.mark.xfail(reason=None, strict=True)to a passing test results in “TypeError: can only concatenate str (not "NoneType") to str”. If a condition is passed, passing noreasonbehaves like passingNoneto it (code). I’m unsure how this subtlety could be captured in the function signature.pytest.mark.skipif()has the same issue.pytest.mark.skip()’s default value forreasonis in fact "unconditional skip" (code).raises: The type (according to the type annotations) isUnion[Type[BaseException], Tuple[Type[BaseException], ...]]instead ofType[Exception]. Neither captures the fact thatNonecan be passed to it (which is also the documented default value).strict=False: The default value ofstrictis actually determined by thexfail_strictconfig (code).