Container action is only supported on Linux

[pllim]

Hello.

I ran into this when I attempted to upload wheels generated in Windows and OSX for a package with C-extension. Do you have plans to re-implement this action in Typescript or Javascript, or is that even possible?

2021-01-15T17:48:26.9532380Z ##[group]Run pypa/gh-action-pypi-publish@master
2021-01-15T17:48:26.9533060Z with:
2021-01-15T17:48:26.9533460Z   packages_dir: wheelhouse/
2021-01-15T17:48:26.9534130Z   repository_url: https://test.pypi.org/legacy/
2021-01-15T17:48:26.9534690Z   verbose: true
2021-01-15T17:48:26.9535230Z   verify_metadata: true
2021-01-15T17:48:26.9535660Z   skip_existing: false
2021-01-15T17:48:26.9536020Z env:
2021-01-15T17:48:26.9536590Z   pythonLocation: /Users/runner/hostedtoolcache/Python/3.8.7/x64
2021-01-15T17:48:26.9537160Z ##[endgroup]
2021-01-15T17:48:26.9586610Z ##[error]Container action is only supported on Linux
2021-01-15T17:48:26.9688420Z Post job cleanup.

Not sure if this would be addressed by #4 or not.

Thank you!

[reynoldsnlp]

Same problem here. Is there a workaround for non-Linux OSes, or do I need to find a different solution?

[webknjaz]

Linking an old answer to this: #48 (comment). Though, this is essentially the same as I wrote below...

[reynoldsnlp]

It would be helpful to post this limitation prominently in the README.

[webknjaz]

If you can send a PR, please go for it :)

[webknjaz]

It would be helpful to post this limitation prominently in the README.

You are right. Feel free to send a PR updating the README.

Same problem here. Is there a workaround for non-Linux OSes, or do I need to find a different solution?

Not exactly, you should just publish your artifacts from the build jobs and then collect them in a separate publish job. This is explained in #15 (comment) and a few other discussions.

I ran into this when I attempted to upload wheels generated in Windows and OSX for a package with C-extension. Do you have plans to re-implement this action in Typescript or Javascript, or is that even possible?

I've explained this elsewhere but no, we'll keep docker because we need to control an isolated environment as opposed to mutating what the users have in their VMs (which may break something for those users implicitly).

Besides, publishing from parallel uncoordinated workers is a bad idea™ — you wouldn't be happy when two job (say Win+Mac) publishes the wheels but Linux fails and you'll end up with a half-published release, would you?

It's always best to upload all the dists after you have them all and after you have all tests passing. use a job with needs in GHA.

Not sure if this would be addressed by #4 or not.

No. But there's a tracking issue for improving the tutorial to accommodate this use-case: #29.