Nightly action failing for Unpinned releases.

[MridulS]

Looks like there were updates to scipy optimisers in the latest 1.15 release and this messes up the assertions in one of the tests https://github.com/scipp/esssans/actions/workflows/unpinned.yml.

E       AssertionError: 
E       Arrays are not equal
E       when comparing values
E       Mismatched elements: 11 / 100 (11%)
E       Max absolute difference among violations: 4.7683716e-07
E       Max relative difference among violations: 1.1717832e-07
E        ACTUAL: array([4.216311, 4.616472, 4.709669, 4.601788, 4.390278, 4.089487,
E              3.743527, 3.373496, 3.044424, 2.733558, 2.482377, 2.240015,
E              2.070589, 1.893589, 1.749089, 1.633279, 1.532102, 1.421638,...
E        DESIRED: array([4.216311, 4.616472, 4.709668, 4.601788, 4.390278, 4.089487,
E              3.743527, 3.373496, 3.044424, 2.733558, 2.482377, 2.240014,
E              2.070589, 1.893589, 1.749089, 1.633279, 1.532102, 1.421638,...

[jokasimr]

Looks like there were updates to scipy optimisers in the latest 1.15 release and this messes up the assertions in one of the tests

This sounds very likely, last two failing runs use scipy 1.15 and the latest passing run uses scipy 1.14.

But I don't see how it could happen. As far as I can tell scipy.optimize is only used in beam_center_from_iofq() but the failing tests use the other beam center finding routine beam_center_from_center_of_mass().

[jokasimr]

Okay the issue is the new scipy release, but not the changes to the optimizers, it was the changes to the physical constants that made the assertion fail:

In version 1.14.1:

In [1]: import scipy as sp
In [2]: sp.version
Out[2]: '1.14.1'

In [3]: sp.constants.neutron_mass
Out[3]: 1.67492749804e-27

In version 1.15.1:

In [1]: import scipy as sp
In [2]: sp.version
Out[2]: '1.15.1'

In [3]: sp.constants.neutron_mass
Out[3]: 1.67492750056e-27

I've verified that replacing the new neutron mass with the old in scipp.constants makes the failing test pass 👍