Remove Lockfile

[chrahunt]

Following up on #6954, our HTTP cache now uses write-then-move instead of explicit file locking. This was the last place that Lockfile was used, so it has been removed.

Closes #4766.

[chrahunt]

Test failure looks unrelated.

[chrahunt]

@cjerdonek, any ideas why TestCandidateEvaluator.test_sort_best_candidate__best_yanked_but_not_all (from here) would be failing here but not in other PRs/master?

Windows Python 2.7 x86 (log) and Python 3.7 x64 (log) both fail in this test. It appears that some output from pep425tags which traces at DEBUG level is getting captured. caplog should not be changing the default level (WARNING) unless explicitly asked so I'm pretty sure it's unrelated to the test itself. Maybe PIP_VERBOSE is set somewhere?

[cjerdonek]

@cjerdonek, any ideas why TestCandidateEvaluator.test_sort_best_candidate__best_yanked_but_not_all would be failing here but not in other PRs/master?

Yes, I noticed this on another PR, too. I believe this is due to the test ordering being different (probably due e.g. to the introduction of different test modules combined with platform-dependent skipped tests, etc) combined with the tests not being completely isolated. A possible quick fix is to set the appropriate log level on caplog at the beginning of the errant tests (probably caplog.set_level(logging.INFO)). And the longer term fix is to see if pytest has a way to automatically reset caplog's level to a constant after (or before) each test. I'm sure pytest has some capability like this, but I didn't look into it yet.

In the absence of a pytest feature like that, we would need to reset caplog at the end of each test that changes it (e.g. using the context manager form).

[pradyunsg]

In the absence of a pytest feature like that, we would need to reset caplog at the end of each test that changes it (e.g. using the context manager form).

I think it'd be better to override the fixture, the same way we override tmpdir.

[cjerdonek]

Hmm, the pytest docs say it already resets the caplog level at the end of every test: https://docs.pytest.org/en/latest/logging.html#caplog-fixture
And yet manually setting the level at the beginning of the affected test worked for me when I noticed this. So I'm not sure why pytest doesn't appear to be behaving as documented.

[chrahunt]

I put that workaround in place for the failing test and created #7026 to track the investigation and fix for the underlying cause.

[chrahunt]

Thanks for reviewing @pradyunsg!