Wheel builds for aarch64 and others

[kafonek]

Closes #114

https://github.com/PyO3/maturin-action has couple of links to CI files for other projects that build wheels with maturin. I thought https://github.com/milesgranger/pyrus-cramjam/blob/master/.github/workflows/CI.yml and https://github.com/messense/py-dissimilar/blob/main/.github/workflows/CI.yml looked really clean in particular. The py-dissimilar CI looks like it builds all the linux and linux-cross python versions in one run but when I tried that I got an error saying maturin couldn't find an interpreter (https://github.com/kafonek/ypy/actions/runs/4093394652/jobs/7058804065 vs one of their working runs https://github.com/messense/py-dissimilar/actions/runs/3968643400/jobs/6802077215)

I think running tests on the built wheels is a good idea, but leaving that for a follow-up PR. I also think adding y-py to the maturin-actions readme would be cool. I can put in that PR if you all are good with it.

Lastly, I want to say python-version: ["3.7", "3.11"] actually skips building wheels for 3.8, 3.9, 3.10? Or is there some magic in the github action matrix syntax that makes this work?

[kafonek]

Link to an action with the aarch64 builds https://github.com/kafonek/ypy/actions/runs/4093450537

[davidbrochart]

Don't we need architecture: ${{ matrix.target }} here and there?

[kafonek]

Yep you're right on the linux build. Doesn't look like there's a macos arch matrix, I see the same for py-dissimilar CI

[davidbrochart]

Lastly, I want to say python-version: ["3.7", "3.11"] actually skips building wheels for 3.8, 3.9, 3.10?

But wheels are on PyPI for these versions.
@Waidhoferj is publishing to PyPI really done from the CI? Also, could you add me as a maintainer on PyPI so that I can release?

[davidbrochart]

@Waidhoferj Can we make a release with aarch64 wheels?

[davidbrochart]

@Waidhoferj Could you add me as a maintainer on PyPI so that I can make a release?

[Waidhoferj]

I don't think I have authorization to add you. @dmonad would you be able to do this?

[davidbrochart]

Thanks, in the meantime could you make a release with this PR @Waidhoferj ? People not being able to install JupyterLab 3.6 is becoming critical.

[Waidhoferj]

Sure sounds good

[Waidhoferj]

@davidbrochart, it looks like some of these new changes are causing the publishing pipeline to fail

[kafonek]

Come on CI, why are you so difficult. That step ran fine when I tested the actions on my branch https://github.com/kafonek/ypy/actions/runs/4093450537/jobs/7058899187

[kafonek]

Hmmmm, or the action on my branch is not using the architecture I expected it to

I must have run that action before the architecture fix @davidbrochart pointed out.

[kafonek]

I'm debugging this now, I'll put up a new PR. If you want to open a new issue for the failing CI, I can post updates over there.