Check build stability

[timkpaine]

Opening because it seems the build went unstable between when I opened the last PR and its CI ran, and when I merged it. Will debug on this PR.

Checklist

• Used a personal fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-webservices]

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) and found it was in an excellent condition.

[mattip]

Upstream excludes grpcio==1.56.0 for darwin. We may need that exclusion after #123

[timkpaine]

I still see the != pin here: https://github.com/ray-project/ray/blob/ray-2.8.0/python/setup.py#L264

[mattip]

I am not sure what is going on.

Half of the builds are failing to create a virtural environment with

KeyError: Dist(channel='conda-forge', dist_name='wcwidth-0.2.12-pyhd8ed1ab_0', \
   name='wcwidth', fmt='.conda', version='0.2.12', build_string='pyhd8ed1ab_0', \
   build_number=0, base_url=None, platform=None)

It seems wcwidth was updated recently. Perhaps the conda locking mechanism is getting confused here?

I will restart the failing jobs.

[timkpaine]

The problem with restarting the failing jobs is we can't do this after merging, so on main it's all or nothing apparently and each time it rebuilds it republishes half the packages which is bad. Maybe @h-vetinari or someone with access to azure directly can restart just the failed jobs, but this isn't sustainable.

[mattip]

There should be a way of ensuring conda install gets a consistent state, even when packages are updating. @isuruf is this something you have seen elsewhere, where a CI job fails because of new packages?

[mattip]

here is an example of the failure from a merge.

[timkpaine]

Yeah it's another key error, I usually see this with mamba. I wonder if it's a general conda bug that just happens to coincide with our merges

[timkpaine]

conda-forge/conda-smithy#1805

[timkpaine]

Maybe @conda-forge/core knows more

[jaimergp]

Sorry about this! It's essentially conda/conda-libmamba-solver#386. We are tracking down the bug and I think Marcel found the root cause in conda/conda-libmamba-solver#393. Reverting temporarily to conda_built_tool: mambabuild in conda-forge.yml should fix it.

[jaimergp]

@conda-forge-admin, please rerender

[github-actions]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

This message was generated by GitHub actions workflow run https://github.com/conda-forge/ray-packages-feedstock/actions/runs/6960908910.

[jaimergp]

@conda-forge-admin, please rerender

[mbargull]

We have a new conda-smithy release that has this a the default for now.

[mattip]

Thanks for tracking this down

[jaimergp]

Yeah it's another key error, I usually see this with mamba.

Btw @timkpaine, by "mamba" there, do you mean mambabuild or conda-libmamba-solver? AFAIK this KeyError issue is a conda-libmamba-solver only thing, but maybe you had seen it with mambabuild too? Just double checking while we figure this out.

[timkpaine]

I have frequently seen this when installing mamba and boa and then trying to render a recipe with conda mambabuild, for ~4-5 months now. I had assumed it was due to incompatibilities between released versions of mamba, Conda-build, and boa (and boa in particular has been due for a new release for quite some time).

[timkpaine]

@conda-forge-admin please rerender

[github-actions]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

This message was generated by GitHub actions workflow run https://github.com/conda-forge/ray-packages-feedstock/actions/runs/6998757196.

[mbargull]

I have frequently seen this when installing mamba and boa and then trying to render a recipe with conda mambabuild, for ~4-5 months now. I had assumed it was due to incompatibilities between released versions of mamba, Conda-build, and boa (and boa in particular has been due for a new release for quite some time).

Interesting. If you happen to stumble on such a case again, please let us know (with aconda info/conda list outputs and recipe, ideally)!

[isuruf]

The KeyError used to happen (not frequently) and went away after some time. For eg: conda-forge/rust-feedstock#75 (comment)

[timkpaine]

The Mac 3.8 build failed on NPM dos prevention which sometimes happen on CI services. If the rest of the builds work I think we can merge and see if a core maintainer can redo any failing builds.