Hello Marten, Thank you for taking a look at this! Looks like the general opinion I've seen around the NumPy community (especially those that have worked on the code for a while) is to avoid the usage of fixtures, so thank you for stating your preference as well.
For your question, for most tests that use setup_method, they do not explicitly define a teardown method. Instead, it looks like pytest automatically calls some sort of implicit teardown procedure to undo anything created during the setup_method. To prevent that from happening would involve that complex change to pytest-run-parallel that I mentioned in the post. And yeah haha, it is a lot of changes, but nothing too crazy! I've managed to make PRs for all the setup methods that were causing problems, so that step is more or less done. Hopefully this style of writing tests shouldn't be too hard to learn once we create documentation for it. Thank you for your support of these changes! Best, Britney _______________________________________________ NumPy-Discussion mailing list -- [email protected] To unsubscribe send an email to [email protected] https://mail.python.org/mailman3//lists/numpy-discussion.python.org Member address: [email protected]
