Pin numpy to 2.0.0rc1

[astrofrog]

Some packages are forcing numpy<2 to be installed, so pinning numpy to 2.0.0rc1 to see what works and what doesn't. Not for merging at this point!

Where the wild things are

• asdf-astropy: Update copy usage in Quantity converter to deal with astropy 6.1 changes asdf-astropy#224
• astropy-healpix: BUG: fix compatibility with numpy 2.0.0 (copy semantics) astropy-healpix#215
• ccdproc: Error with Numpy 2.0.0rc1 ccdproc#818
• photutils: BLD: build with numpy 2.0.0rc1 (or newer) photutils#1717 and Update Numpy build-time requirement (time sensitive) pydata/bottleneck#445
• regions: Build with numpy 2.0.0rc1 or newer regions#545
• reproject: Build against Numpy 2.0.0rc1 or later reproject#436 and astropy-healpix fix above
• specreduce: Might be automatically fixed by photutils fix 🤷‍♀️
• specutils: MNT: Compat with numpy 2 and astropy 6.1 specutils#1130
• sunpy: Build and test against numpy 2.0rc1 sunpy/sunpy#7555

[astrofrog]

I'm trying to get reproject to work and have released new wheels but that job is still failing as it requires a new release of astropy-healpix. This is in preparation in astropy/astropy-healpix#214 but running into issues related to copy= which @neutrinoceros is going to investigate.

[pllim]

asdf-astropy -- I think @braingram is aware of this.

But just in case:

• Copy semantics change in numpy 2.0 caused incompatibility with astropy v6.1rc asdf-astropy#226

[pllim]

For completeness, I cross-linked related issues in the original post above. Hope you don't mind. Thanks!

[astrofrog]

That's fantastic, thank you!

[astrofrog]

I've restarted the CI to check where things stand right now.

[astrofrog]

Restarted again

[neutrinoceros]

Thanks ! I see ccdproc is still not caught up, I'll get back to it !

[neutrinoceros]

@astrofrog if you'd like to keep the PR description updated, here's the latest (and last) patch needed for ccdproc astropy/ccdproc#824

Also worth noting that it depends on astroscrappy, which Simon in working on in astropy/astroscrappy#85

[neutrinoceros]

Also, I wanted to double check on specreduce and it doesn't just need photutils: the test suite also requires a new release for synphot. @pllim already opened a PR for it as spacetelescope/synphot_refactor#387, and I confirm that it's sufficent to get specreduce's test suite green

[pllim]

Thanks for the update! I guess I have to release synphot sooner than later...

Update: Released 1.4.0

[pllim]

Some downstream packages like ginga and acstools need a new photutils release to get their RC testing job to pass. cc @larrybradley

• ginga passed but it couldn't run stuff that needs photutils.
• acstools failed because photutils is required -- https://github.com/spacetelescope/acstools/actions/runs/8653450052/job/23728621513

[larrybradley]

I should have time to do a photutils and regions release tomorrow afternoon.

[pllim]

I think we're almost there. Should we omit bottleneck to get photutils green or wait it out?

[pllim]

I merged #17 . Please rebase to ensure sunpy is now green. Thanks!

[astrofrog]

Rebased!

[astrofrog]

@Cadair @pllim - sunpy is still red due to opencv errors.

[nabobalis]

sunpy release has not got the numpy fixes yet. We need to get 5.1.3 out.

sunpy-dev still is not skipping the figure tests. I wonder if there is a problem in the pytest config?

[nabobalis]

Actually I think those sunpy dev issues might be on us but I am investigating.

[nabobalis]

For sunpy dev I opened: #18 it points to my branch that I hope to merge but it shows sunpy-dev passing under numpy2.0rc now (that last commit needs to go before a merge)

[astrofrog]

Ok so the remaining issues for the CI to turn green are:

• bottleneck still needs to be released - not a big deal because it's only an optional dependency for astropy and photutils anyway
• astroscrappy needs to be released for ccdproc - that one is a required dependency but we can do the release ourselves.
• sunpy needs to be released

None of the issues appear to be related to astropy 6.1 so we should be safe to go ahead and do a release I think.

[pllim]

Any of these diff actually need to be merged for next time? If not, we can close the PR without merge. Thanks!

[astrofrog]

Looks like bottleneck have an rc out with numpy 2.0 support, so trying to re-run

[neutrinoceros]

Seems like I missed a spot when I worked on ccdproc, however, the one remaining failure doesn't affect users, and its fixed in astropy/ccdproc#829

[astrofrog]

Everything is green except for ccdproc stable which makes sense since it requires astropy/ccdproc#829 which hasn't been released yet (and is not urgent since doesn't affect users). I think we can close this now!

[pllim]

Thanks!