Different 2.0.0.1 binaries for aarch64 darwin vs aarch64 ubuntu

[brandonchinn178]

I've downloaded the two aarch64 bundles from the 2.0.0.1 releases page and see that they support different versions:

• aarch64-darwin
  • 8.10.7
  • 9.2.7
  • 9.2.8
  • 9.4.5
  • 9.6.2
• aarch64-ubuntu20
  • 8.10.7
  • 9.0.2
  • 9.2.7
  • 9.2.8
  • 9.6.2

Is there a reason the bundles contain different GHC versions?

[hasufell]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

[joyfulmantis]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

Could we declare the list of versions as a constant and use that throughout the file to prevent this issue happening again? Perhaps with yaml's merge key?

[hasufell]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

Could we declare the list of versions as a constant and use that throughout the file to prevent this issue happening again? Perhaps with yaml's merge key?

Afaik this is not valid GH action syntax.

actions/runner#1182

[joyfulmantis]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

Could we declare the list of versions as a constant and use that throughout the file to prevent this issue happening again? Perhaps with yaml's merge key?

Or I guess yaml aliases would probably be more appropriate

[joyfulmantis]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

Could we declare the list of versions as a constant and use that throughout the file to prevent this issue happening again? Perhaps with yaml's merge key?

Afaik this is not valid GH action syntax.

actions/runner#1182

That's too bad, it would have made future releases more resilient.

[hasufell]

The CI was rewritten and there are still some rough edges.

You can use 2.0.0.0 with 9.4.5 still.

Could we declare the list of versions as a constant and use that throughout the file to prevent this issue happening again? Perhaps with yaml's merge key?

Afaik this is not valid GH action syntax.

actions/runner#1182

That's too bad, it would have made future releases more resilient.

What we can do instead is examine the generated json: https://github.com/haskell/ghcup-metadata/blob/develop/hls-metadata-0.0.1.json

And then execute some sanity checks on it. However... some architectures may diverge on purpose (e.g. FreeBSD). So you'll have to keep that knowledge somewhere after all.

[michaelpj]

I have some hope to generate the matrix. I have something like this for the test job, but I need to find some time to apply it to the release CI job.

[michaelpj]

Old, new releases are hopefully fixed