Fix #72: Add java artifacts

[yhori991]

Checklist

• Used a personal fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)

I've tested the build only for Linux-64 Python3.12 environment so far.
Let me see how it goes on all the other platforms.

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

I do have some suggestions for making it better though...

For recipe/meta.yaml:

• ℹ️ The recipe is not parsable by parser conda-souschef (grayskull). This parser is not currently used by conda-forge, but may be in the future. We are collecting information to see which recipes are compatible with grayskull.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/15041903025. Examine the logs at this URL for more detail.}

[mattip]

Do the linux builds take too much memory and trigger an OOM error? They are crashing after ~5 1/2 hours with no error message.

[yhori991]

I think so too. The way Linux builds failed looked to me like the one from OOM.
For all cases, the log stream seemingly got cut off without any signs of logical errors.
I hope that we can tweak conda-forge CI/CD pipeline settings for this...

The windows build on the other hand looks to me simpler.
We don't have unzip package for windows platform and we can exclude windows on this.

However, the bezel build recipe is using unzip internally.
https://github.com/ray-project/ray/blob/ea987b5cfc6464ee74ace6ce81bbfed1792d3d93/java/BUILD.bazel#L286
So, in the end, it might be necessary to support unzip on windows to solve this.

[yhori991]

FWIW, it seems that the upstream does not support building/testing the Java artifacts on the Windows platform.
Ref: https://github.com/ray-project/ray/blob/a90a8184f990a18b031cf95cc2ecf44e0474f6fe/ci/ray_ci/windows/tests.env.Dockerfile#L14
Setting RAY_INSTALL_JAVA=0 in this file indicates that the Windows build does not include Java bindings.

[mattip]

What is the motivation for this PR?

I hope that we can tweak conda-forge CI/CD pipeline settings for this

I don't know how to use a larger machine in conda-forge/azure, maybe someone else in the @conda-forge/ray-packages team knows. I don't understand why it takes 5 hours to build ray. Locally the build takes ~20 minutes.

[yhori991]

Well... my main motivation is to run RayDP purely on conda environments.
It's already possible by mixing PyPi and installing ray and raydp from there but it's so hard to do that without breaking something.

I don't understand why it takes 5 hours to build ray. Locally the build takes ~20 minutes.

It takes roughly the same amount of time (~20mins) on my local machine too.

[apmorton]

It takes forever because the azure devops runners have 2 cores.

I don't know how to use a larger machine in conda-forge/azure, maybe someone else in the @conda-forge/ray-packages team knows.

I've put a request in for larger runners for this feedstock - conda-forge/admin-requests#1509

[mattip]

Close/reopen to use the larger linux runnuers

[mattip]

The security policy on the larger machines requires that the commit come from someone approved to use them, so I pushed an empty commit in my name.

[mattip]

Empty commit only ran windows and macOS CI jobs, maybe rebasing is needed to get the linux ones to run?

[mattip]

@jaimergp any ideas why this is not running the linux CI after merging #210? Do I need to submit it from my fork?

[jaimergp]

Not sure, maybe it has to do with the conflicts? IIRC, GHA won't run in pull_request triggers if there are conflicts. I'd try fixing those and if that doesn't fix the runs, re-open the PR from your fork? (But I really think it should work once conflicts are fixed).

[yhori991]

I've fixed the conflicts. Let's see 🤞

[yhori991]

We can see all the targets included in the CI/CD now.
The run is not kicked off since I'm not in the whitelist, but this looks promising.

[yhori991]

My bad. It's already run and green. Thanks you for updating @mattip