Background

I'm working towards introducing pants. Eventually I would like to use pants to run tests to take advantage of the fine-grained per-file caching of results that accounts for dependencies by python files (pants infers the deps by reading the python files). But, pants uses pytest to run tests not nosetest.

This is another of several PRs that improves our compatibility with pytest. Others include: #5686, #5689, #5690

Note: I do not have pytest support completely ironed out, but this is a step in that direction. Don't expect running all our tests with pytest to work yet.

Overview

DBTestCase was relying on some unittest/nosetest implmentation details that are not available when running under pytest.

The purpose of those adjustments was: Keep the DB after a test failure (do not drop it) to facilitate debugging.

However, pytest does not make results availble to fixtures. There are some complex workarounds, but there are issues with blindly copying the previous logic. For example, pytest can continue running tests after a failure or error unlike nosetest which we have configured to stop on failures. In order to safely skip dropping the DB, we need to make the DB available in a clean state after the test(s) using it. So, re-implementing this test feature requires something that renames a DB to preserve data for debugging, so that subsequent tests have a clean DB to work with.

I left a TODO note about this for whoever ends up trying to implement this. For now, this just avoids accessing attributes on an TestResult object that pytest does not provide.