Repository navigation
Revert Gaussian test change from #14718 - #14771
Conversation
|
Thank you for your contribution to Astropy! 🌌 This checklist is meant to remind the package maintainers who will review this pull request of some common things to look for.
|
|
👋 Thank you for your draft pull request! Do you know that you can use |
mhvk
left a comment
There was a problem hiding this comment.
Bit odd that Gaussian now is OK, but the tests pass, so hopefully it remains OK. The other changes also look good.
pllim
left a comment
There was a problem hiding this comment.
These numpy changes are painful for me. Hard to keep track what is what. At least the remaining failures are not modeling, so LGTM. Thanks!
@mhvk see #14718 (comment) for my analysis of what specifically is going on. Its either a small numerical change (maybe numpy/numpy#23721?), a thresholding fix (maybe numpy/numpy#23695?), or something else. I don't precisely know when the new numpy dev wheels were made and what changes were in each version of the wheels to track it down further... I don't think its critical for me to find the changes which caused then fixed the Gaussian issue from numpy either. |
|
Going to kick off backports again so it picks up all the diff. @meeseeksdev backport to v5.2.x @meeseeksdev backport to v5.3.x |
|
Could not push to auto-backport-of-pr-14771-on-v5.2.x due to error, aborting. |
|
Could not push to auto-backport-of-pr-14771-on-v5.3.x due to error, aborting. |
|
C'mone. |
|
@meeseeksdev backport to v5.3.x |
|
Could not push to auto-backport-of-pr-14771-on-v5.3.x due to error, aborting. |
Revert Gaussian test change from astropy#14718
Revert Gaussian test change from astropy#14718
|
Thanks for the explanation, @WilliamJamieson! |
from numpy cos/sin changes that have since been reverted in 1.25.dev
from numpy cos/sin changes that have since been reverted in 1.25.dev
from numpy cos/sin changes that have since been reverted in 1.25.dev
from numpy cos/sin changes that have since been reverted in 1.25.dev
I can no longer reproduce the error skipped by #14718 for the Gaussian locally, this PR demonstrates that skipping this was unnecessary. However, there are new issues... This PR goes on to correct the remaining modeling issues as those were only related to slight numerical changes that required slightly changing tolerances.
I believe this change maybe due to numpy/numpy#23721 or numpy/numpy#23695 updating some underlying things.
Fixes #14766