unittest.mock's patch.multiple being interpreted as fixtures

[jamesbraza]

From pytest's docs on unittest.TestCase support:

pytest supports running Python unittest-based tests out of the box.

However, I think I have a case where TestCase is not properly supported: unittest.mock's patch.multiple.

From the unittest.mock's docs on patch.multiple:

@patch.multiple('__main__', thing=DEFAULT, other=DEFAULT)
def test_function(thing, other):

You can break out the patches into args for a test function. However, this doesn't seem to work with pytest, as it interprets thing and other as fixtures.

Minimal Repro

from unittest import TestCase
from unittest.mock import DEFAULT, MagicMock, patch

class Foo:
    def bar(self) -> None:
        pass

class TestFoo1(TestCase):
    @patch.multiple(Foo, bar=DEFAULT)
    def test_patch_multiple(self, bar: MagicMock) -> None:
        """Works when subclassing TestCase."""

class TestFoo2:
    @patch.multiple(Foo, bar=DEFAULT)
    def test_patch_multiple(self, bar: MagicMock) -> None:
        """Fails when not subclassing TestCase."""

Here is the output when running pytest

patch_multiple.py::TestFoo2::test_patch_multiple ERROR                   [100%]
test setup failed
file /path/to/patch_multiple.py, line 14
      @patch.multiple(Foo, bar=DEFAULT)
      def test_patch_multiple(self, bar: MagicMock) -> None:
E       fixture 'bar' not found

By subclassing TestCase, pytest doesn't interpret bar as a fixture, and the test runs. Are there any known workarounds to stop pytest from interpreting bar as a fixture?

Versions

I am using macOS Catalina version 10.15.7.

Python==3.7.7
pytest==6.2.4

[nicoddemus]

Thanks @jamesbraza for the report. 👍

The problem is in num_mock_patch_args, which properly handles functions patched by @patch.object but not @patch.multiple.