Updates the sdss mwmStar loader to fix bug with file datasums

[havok2063]

This PR updates the SDSS-V mwm data loaders to account for a bug in a subset of the DR19 SDSS-V mwmStar files. A subset of files have extensions with empty data arrays and non-zero DATASUM keywords. This was due to some data needing to be cleaned out but the header wasn't fully scrubbed/corrected.

The previous loader would attempt to load first extension found with non-zero DATASUM and would crash on the empty data. This fix allows the loader to correctly skip over it and load the first real data extension it finds.

It would be great if this can be backported to specutils 1.x, so it can be used currently in Jdaviz.

[rosteen]

Thanks for the contribution @havok2063, unfortunately it looks like there's a real test failure. Could you address that, please? Once we get this in I can backport and do a bugfix release, although with the next Jdaviz release we'll be using specutils 2 there so it won't be needed for too long.

[havok2063]

@rosteen Are you referring to the Python 3.12 Windows run? Unfortunately, I don't have a windows machine to test on. I'll see if I can run the action locally. I just pushed fixes for the Code style checks. Do I need to worry about the rtd or release action failures? (looks like they are passing now)

[rosteen]

@rosteen Are you referring to the Python 3.12 Windows run? Unfortunately, I don't have a windows machine to test on. I'll see if I can run the action locally. I just pushed fixes for the Code style checks. Do I need to worry about the rtd or release action failures? (looks like they are passing now)

Yup, I was talking about the Windows run, looks like the others were transient problems with remote data. I have a Windows machine at home I can try debugging on tomorrow or Friday, unfortunately I just went into the office for the rest of the day today.

[havok2063]

Thanks for trying to look into this. I couldn't run the action locally, but I can try to spin up a Virtual Box of windows or something. I do have an old windows machine lying around that I can try to spin up git+conda on. Let me know if you end up testing this though.

[rosteen]

I was able to reproduce the test failure locally, it's due to arr.itemsize being 0 at https://github.com/astropy/astropy/blob/412db1e64cca4e4d05cd373dc3cbe12ff2d6a257/astropy/io/fits/util.py#L589. That code is working around a bug but is 11 years old, so I wonder if its still relevant. You also may be able to change the test to work around it but I don't 100% understand the datasum issue to change it myself. I'm trying to figure out on my Windows machine if that fits code is even needed anymore...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.83%. Comparing base (702f0ce) to head (66bc8c3).
⚠️ Report is 9 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1253   +/-   ##
=======================================
  Coverage   86.83%   86.83%           
=======================================
  Files          63       63           
  Lines        4732     4732           
=======================================
  Hits         4109     4109           
  Misses        623      623           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.