Migrate CI to GitHub Actions #27
Conversation
|
GHA CI is failing on my fork with NaN vs 0 assertion errors: Will investigate. |
jrbourbeau
left a comment
There was a problem hiding this comment.
@phobson do we know if those assertions are due to a change in crick results across Python versions, or a difference in what scipy is returning?
|
@jrbourbeau the failures are related to changes in what scipy returns. Since the expected values of the tests generated by scipy, I marked those tests as xfails for now. I also removed the |
|
Tests are passing: https://github.com/phobson/crick/actions/runs/3501524992/jobs/5865250644 |
| @@ -1,33 +1,34 @@ | |||
| import sys | |||
| import os | |||
| import sys | |||
There was a problem hiding this comment.
The changes in this file look like they're coming from black / isort / ... Is that the case? Just double checking the changes here are formatting-only
There was a problem hiding this comment.
yes -- formatting only. I can revert those changes in try to keep them in #29 if you think that's more appropriate
| RTOL = 0 | ||
| ATOL = 1e-8 | ||
|
|
||
| scipy_xfail_mark = pytest.mark.xfail(reason="modern scipy return NaN") |
There was a problem hiding this comment.
I'm still a little uncertain here. Did scipy change its output and crick no longer matches the more recent scipy behavior? Or do we just need to update the tests?
There was a problem hiding this comment.
scipy changed it's behavior, so we'll have to update crick to match the new behavior. Alternatively, we could hardcode the expected values instead of using scipy to generate them, but that's not a very attractive solution to me. The assumption there is that scipy had good reason to change the behavior of these edge cases.
Co-authored-by: James Bourbeau <[email protected]>
Co-authored-by: James Bourbeau <[email protected]>
Co-authored-by: James Bourbeau <[email protected]>
jrbourbeau
left a comment
There was a problem hiding this comment.
Thanks @phobson! Could you open up a separate issue for updating crick so we can remove scipy_xfail_mark?
fixes #28
ref #26