Repository navigation
test_icrs_consistency and test_against_pyephem failing with pyerfa 1.7.2 #11305
Description
This test is failing following the release of pyerfa 1.7.2 (it passes locally with 1.7.1 and fails with 1.7.2):
https://github.com/astropy/astropy/runs/1850026053?check_suite_focus=true#step:6:933
E AssertionError:
E Not equal to tolerance rtol=1e-07, atol=1e-06
E
E Mismatched elements: 1 / 1 (100%)
E Max absolute difference: 3.15489777e-05
E Max relative difference: inf
E x: array(3.154898e-05)
E y: array(0.)
Activity
mhvk commentedon Feb 8, 2021
The new SOFA version wrapped in 1.7.2 has changes to the polar motion, making them more accurate. I noticed that quite a few of the SOFA tests themselves were changed (and I know I copied some of those for
EarthLocation, etc., to astropy). But then, it would seem this should not affect the consistency tests!I guess the first question now is what to do:
One lesson learned is that I really should have checked astropy with the new erfa...
pllim commentedon Feb 8, 2021
To clarify: There are 201 tests failing because of this -- partial log at #11307
[-]test_icrs_consistency failing with pyerfa 1.7.2[/-][+]test_icrs_consistency and test_against_pyephem failing with pyerfa 1.7.2[/+]mhvk commentedon Feb 8, 2021
Shall we just use
pyerfa < 1.7.2for now? I can look into the failures in detail this only at the end of the day...olebole commentedon Feb 11, 2021
On Debian, we now also see a failure -- however only on
test_against_pyephem. Do I understand it right that this is mainly a problem of the test, so that I could just disable it?pllim commentedon Feb 11, 2021
🤷
mhvk commentedon Feb 11, 2021
@olebole - yes, it is a problem with the test, which checks that results do not change relative to the older version of
erfa- but the newer version does polar motion more accurately.mkbrewer commentedon Feb 26, 2021
@mhvk This should be an easy fix. You just need to update ErfaAstrom.apio() to use the new polar motion algorithm.
mhvk commentedon Feb 27, 2021
Yes, agreed the fix is obvious. It is a matter of getting around to it...
pllim commentedon Apr 2, 2021
Looks like this was unresolved for 4.2.1. I am tentatively milestoning this issue to 4.3 so we don't forget about it before next major release.
mhvk commentedon Apr 6, 2021
Finally got around to this. See #11485 for a fix that gets rid of all but one of the errors. Am not yet seeing the solution for the last one... Help appreciated!
pllim commentedon Apr 12, 2021
Resolved by #11485. Thanks, everyone!