numpy/numpy-user-dtypes

No sdist on pypi

Closed this issue · 5 comments

Currently there's no sdist available on pypi, so pip install numpy-quaddtype fails for Python 3.13 because there aren't Python 3.13 wheels either.

We should set up wheels for 3.13 and 3.14 and make sure the sdist gets uploaded by the release automation.

I'd be very happy if we actually released a new version now. While it doesn't contain everything that I want yet, it would be good to test all the things that we've worked on irl

Unfortunately I'm pretty swamped right now so I can't commit time to work on this for at least a week or two.

If you have time and interest then I should have time to at least review fixes for the release automation. It'd also be nice to add Python 3.13 and 3.14 builds too. I don't know what the SLEEF thread safety story is, but maybe free-threaded builds as well?

Currently there's no sdist available on pypi, so pip install numpy-quaddtype fails for Python 3.13 because there aren't Python 3.13 wheels either.

We should set up wheels for 3.13 and 3.14 and make sure the sdist gets uploaded by the release automation.

Sorry it seems I wasn't get notified somehow of the new opened issues, the conda-forge team also pinged me to upload the sdist so that both branches stay updated. I replied to do that with v1.0 release.
On a dev branch I can expand the Python in CI and test the builds for 3.13 and 3.14, given this I am also occupied in my "job's" work so this upcoming week might be tight for me to work.
But I'll try do keep fixing small things on the way

On a dev branch I can expand the Python in CI and test the builds for 3.13 and 3.14, given this I am also occupied in my "job's" work so this upcoming week might be tight for me to work.

Awesome. Obviously your day job takes precedence. I really appreciate the time you put into this. I may also try to poke at the release automation if I have some spare cycles in the next few weeks. There's no time pressure on this.

Fixed by #135