issue_comments

But... I've lost enough time to this already, and removing runner.isolated_filesystem() has the tests passing again. So I'm not going to work on this any more.

To clarify: the core problem here is that an error is thrown any time you call os.getcwd() but the directory you are currently in has been deleted.

runner.isolated_filesystem() assumes that the current directory in has not been deleted. But the various temporary directory utilities in pytest work by creating directories and then deleting them.

Maybe there's a larger problem here that I play a bit fast and loose with os.chdir() in both the test suite and in various lines of code in Datasette itself (in particular in the publish commands)?

STILL failing. I'm going to try removing all instances of isolated_filesystem() in favour of a different pattern using pytest temporary files, then see if I can get that to work without the serial hack.

Looking at the source code in Click for isolated_filesystem(): https://github.com/pallets/click/blob/9da166957f5848b641231d485467f6140bca2bc0/src/click/testing.py#L450-L468

python @contextlib.contextmanager def isolated_filesystem( self, temp_dir: t.Optional[t.Union[str, os.PathLike]] = None ) -> t.Iterator[str]: """A context manager that creates a temporary directory and changes the current working directory to it. This isolates tests that affect the contents of the CWD to prevent them from interfering with each other. :param temp_dir: Create the temporary directory under this directory. If given, the created directory is not removed when exiting. .. versionchanged:: 8.0 Added the ``temp_dir`` parameter. """ cwd = os.getcwd() t = tempfile.mkdtemp(dir=temp_dir) os.chdir(t) How about if I pass in that optional temp_dir as a temp directory created using the pytest-xdist aware pytest mechanisms: https://docs.pytest.org/en/6.2.x/tmpdir.html

Tests are still failing in the job that calculates coverage.

That fixed it. My hunch is that Click's runner.isolated_filesystem() mechanism doesn't play well with pytest-xdist.

Still couldn't replicate on my laptop. On a hunch, I'm going to add @pytest.mark.serial to every test that uses runner.isolated_filesystem().

It happens in CI but not on my laptop. I think I need to run the tests on my laptop like this: https://github.com/simonw/datasette/blob/121e10c29c5b412fddf0326939f1fe46c3ad9d4a/.github/workflows/test.yml#L27-L30