Fix or silence pylint warnings in test_lint_free_pylint test

[juliangilbey]

Building 1.7.0 for Debian, two tests failed: test/plugins/test_pylint_lint.py::test_lint_free_pylint and test_per_file_caching, first with the output

>       assert not pylint_lint.pylsp_lint(
            config, ws, Document(uris.from_fs_path(__file__), ws), True)
E       assert not [{'code': 'C0103', 'message': '[invalid-name] Variable name "ws" doesn\'t conform to snake_case naming style', 'range'...t class', 'range': {'end': {'character': 83, 'line': 94}, 'start': {'character': 62, 'line': 94}}, 'severity': 2, ...}]

and once I'd fixed that, then with the output

>       assert not pylint_lint.pylsp_lint(
            config, wksp, Document(uris.from_fs_path(__file__), wksp), True)
E       AssertionError: assert not [{'code': 'W0212', 'message': '[protected-access] Access to a protected member _endpoint of a client class', 'range': {'end': {'character': 85, 'line': 94}, 'start': {'character': 64, 'line': 94}}, 'severity': 2, ...}]

This patch addresses both of these two failures; the first by changing ws to wksp and the second by disabling this pylint warning (as there's no (public) accessor method for the _endpoint member.

The thing I really don't understand, though, is how this test passes on the GitHub CI system! I'm using clean chroot and lxc containers to run the tests with a completely vanilla pylint 2.15.9 installation, just as GitHub CI is doing.

[juliangilbey]

OK, that PR failed the CI tests, but I haven't got a clue what's gone wrong - the diagnostics don't give enough information to debug it. So I'm going to push a further commit to this PR to get some information...

[juliangilbey]

OK, I've submitted #334 to allow this test to not crash-and-burn, but I'm a little stumped now. On my clean chroot, the test crashes if I don't include the disable-next: protected-access directive, but on the GitHub CI setup, it crashes if I do include it. But the pylint versions are identical, which is really weird. I guess we could try ignoring both, but that seems a bit ugly.

Thoughts would be appreciated!

[juliangilbey]

OK, I think I've found the solution: Debian's pylint has slightly different defaults to the PyPI version of pylint (as installed via pip). I have no idea why. But the key differences, regarding this PR, are that Debian disables useless-suppression and enables invalid-name, whereas the PyPI version is the opposite. That explains why this patch is necessary for Debian but fails on GitHub CI. But this exploration did pick up on a genuine bug (#334), so all is not wasted!

So I suggest closing this PR now, as it is not necessary for GitHub CI (and it seems hard to resolve cleanly without introducing a local pylintrc file).

[ccordoba12]

But this exploration did pick up on a genuine bug (#334), so all is not wasted!

Yep, thanks for your help with that.

So I suggest closing this PR now, as it is not necessary for GitHub CI (and it seems hard to resolve cleanly without introducing a local pylintrc file).

Ok, closing then.