Description
I've got a pytest plugin that registers a custom mark in the pytest_configure
hook using config.addinivalue_line
. When I load my plugin by putting pytest_plugins = ("my_plugin",)
in my test file (as described here) and add the mark to one of my tests, I still get a warning PytestUnknownMarkWarning: Unknown pytest.mark.my_mark - is this a typo?
When I load my plugin by putting pytest_plugins = ("my_plugin",)
in the conftest.py
file, I do not get the warning.
Here's my example plugin:
import pytest
# This fixture is just so I know the plugin is working.
# If it's not installed correctly, I'll get an error when the test tries to use the foo fixture
@pytest.fixture
def foo():
return "foo"
def pytest_configure(config):
config.addinivalue_line("markers", "my_mark")
Here's the test file that triggers the warning:
import pytest
@pytest.mark.my_mark
def test_01(foo):
assert foo == "foo"
Here's my environment
Package Version
------------- -------
attrs 21.4.0
iniconfig 1.1.1
my-plugin 1.0.0
packaging 21.3
pip 20.3.4
pkg-resources 0.0.0
pluggy 1.0.0
py 1.11.0
pyparsing 3.0.7
pytest 7.0.1
setuptools 44.1.1
tomli 2.0.1
I'm aware that I can use a setuptools entrypoint to install my plugin, but that isn't an option in the more complicated environment where I encountered the problem. Likewise, the use of conftest.py
to load the plugin is a bit frowned upon - we're trying declare the plugins we're using in the test files themselves so they're more stand-alone readable (among other reasons).
I suspect what's going wrong here is the warning is triggered during assertion re-writing in the test module. If I drop a breakpoint here I see the following near the end of the backtrace
. . .
.../_pytest/assertion/rewrite.py(170)exec_module()
-> exec(co, module.__dict__)
.../my_test.py(4)<module>()
-> @pytest.mark.my_mark
.../_pytest/mark/structures.py(520)__getattr__()
-> warnings.warn(
That happens before my pytest_configure hook is run, so it has no idea that my_mark
is a custom mark yet.
Unfortunately, this is about the limit of my understanding so I can't be more helpful by trying stuff like "suppress this warning during assertion re-writing" to see if that fixes the problem. It's a bummer that this mark triggers a warning.