You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We shouldn't disable the image size comparison, as this should definitely count as a failure. The current output is currently wrong, either due to a bug in Matplotlib or a change in API in Matplotlib, so we need to address the underlying cause either way.
For example, test_patches_distortion has:
Before:
After:
I'm currently bisecting Matplotlib to understand why this changed (and is now wrong)
Bisecting points to matplotlib/matplotlib#14134 as the culprit. I'll try and understand if this is an actual bug in MPL or whether we need to fix anything in WCSAxes.
Thanks, @astrofrog ! #8722 did indeed fix sty not found error, but we still have the xfail-ed ones to deal with whenever you can get to it (no rush currently unless they are released that version soon).
Activity
pllim commentedon May 9, 2019
I also posted at matplotlib/matplotlib#14180 . Maybe disabling image size comparison in the test (if possible) is the best route to go.
astrofrog commentedon May 11, 2019
We shouldn't disable the image size comparison, as this should definitely count as a failure. The current output is currently wrong, either due to a bug in Matplotlib or a change in API in Matplotlib, so we need to address the underlying cause either way.
For example, test_patches_distortion has:
Before:
After:
I'm currently bisecting Matplotlib to understand why this changed (and is now wrong)
astrofrog commentedon May 11, 2019
Bisecting points to matplotlib/matplotlib#14134 as the culprit. I'll try and understand if this is an actual bug in MPL or whether we need to fix anything in WCSAxes.
astrofrog commentedon May 11, 2019
A minimal example that demonstrates the issue is:
this seems to happen only when
aspect='equal'. I'll investigate further.astrofrog commentedon May 11, 2019
I've posted a comment here in case there are any insights on the Matplotlib side.
astrofrog commentedon May 11, 2019
And I've now narrowed this down to a minimal example in matplotlib/matplotlib#14134 (comment), so let's see what they say.
pllim commentedon May 15, 2019
mpldevjob is failing now with a different error:https://circleci.com/gh/astropy/astropy/31251
astrofrog commentedon May 15, 2019
Whaaaaaat
astrofrog commentedon May 17, 2019
See matplotlib/matplotlib#14151 - PR forthcoming
pllim commentedon May 17, 2019
Thanks, @astrofrog ! #8722 did indeed fix
sty not founderror, but we still have the xfail-ed ones to deal with whenever you can get to it (no rush currently unless they are released that version soon).astrofrog commentedon May 17, 2019
The xfailed ones will be fixed by matplotlib/matplotlib#14216 :)