feat: allow uploading on main channel for v1 recipes

[nichmor]

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.

now that we have implemented all rattler-build lints checks and we have a working feedstock: https://github.com/conda-forge/jolt-physics-feedstock
I would like to uplift restrictions of uploading rattler-build to main channel.

Should be merged after release of: conda-forge/conda-smithy#2037

[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/meta.yaml) and found it was in an excellent condition.

[beckermr]

We need to wait a bit imho for the dust to settle on the linter. There are a few PRs left. Also the bot is not yet finished, which might break things.

[nichmor]

We need to wait a bit imho for the dust to settle on the linter. There are a few PRs left. Also the bot is not yet finished, which might break things.

Hey ! Thanks for the review!

Can we make an actionable plan for linter work?
I know that there is conda-forge/conda-smithy#2028 work that needs to be done.
Do you have anything else on your radar?

Regarding the bot work- can it block us from allowing us to upload on the main channel? I think that even we break things we do it on bot side, uploaded packages are not affected? Do you have something on your mind how we can achieve this?

[beckermr]

It depends on how the bot breaks. If people convert feedstocks that need ABI migrations or version migrations, we'll break more than just a single feedstock. Things that affect the whole ecosystem like this have to move slowly enough for folks to adapt.

[baszalmstra]

@beckermr I completely understand where you are coming from but I feel the issue of packages not being uploaded is a bit of a chicken-and-egg problem. We have already had people (in the chat) wanting to convert their feedstock but holding off because the packages would not get uploaded. That makes it hard for the dust to settle. I know @jaimergp also had different ideas about this.

From my perspective, these are the issues that are left open:

conda-smithy:

• fix: pass --extra-meta for rattler-build conda-smithy#2037
• [ ] add source linting for recipe v2 conda-smithy#2028
• fix: remove example recipe.yaml lint conda-smithy#2036
• rerender with rattler-build removing CODEOWNERS conda-smithy#2047

bot:

• misc: disable recipe.yaml for migrators regro/cf-scripts#2918
• [ ] add v2 recipe editing capabilities for bumping version & build number regro/cf-scripts#2920
• [ ] Enable jpeg turbo for new recipes regro/cf-scripts#2924

Am I missing issues or pull-requests?

edit: These items have either been deemed not needed or done in other places.

[beckermr]

The ABI migrations run mini migrators on top of the recipes. We'll need to merge the prs to skip those or make them work on the new format.

[baszalmstra]

The ABI migrations run mini migrators on top of the recipes. We'll need to merge the prs to skip those or make them work on the new format.

Is there already an issue for this so we can keep track of it?

[beckermr]

On the bot repo it is issues 2918 2920 and 2924

[baszalmstra]

Thanks I updated the overview. 👍

[beckermr]

Another issue is that the solver checks repo does not currently work for v1 recipes: regro/conda-forge-feedstock-check-solvable#54

Do we have a tracking issue on the website repo for this? If not we should.

[wolfv]

Some smaller progress was made wrt to the CODEOWNERS file - a change in rattler-build-conda-compat should have made it work.

[wolfv]

I bumped the version to 4.9.5 - would love to see this merged :)

[beckermr]

@conda-forge/core shall we go ahead with this? I am a +1.

[jakirkham]

Am in favor of trying this

Would suggest we do this on a non-weekend adjacent day (so not Monday or Friday). For example we could do this tomorrow (Thursday morning) or some time next week (Tuesday or later). Mention this as most people likely don't want to spend their weekends debugging and fixing issues. So picking some time where people will be around and ready to work on issues would be best

[beckermr]

I'm around for the next four hours so I'll merge this now and watch things on the heroku server.

[jakirkham]

Sounds good. Thanks Matt and thanks everyone working on this upgrade! 🙏

[jakirkham]

Noticed the conda-forge-ci-setup cross-compiled architectures are failing (for example)

__main__.py: error: argument --sysroot: expected one argument

Do we need to bootstrap this change somehow (wait for native builds and restart)? Or is there something missing in those cases?

[beckermr]

That appears to be a cross-python issue unrelated to this package, right?

cc @conda-forge/core @conda-forge/cross-python

[jakirkham]

Yeah came to the same conclusion

That package has this logic in its activation script:

    $BUILD_PREFIX/bin/python -m crossenv $PREFIX/bin/python \
        --sysroot $CONDA_BUILD_SYSROOT \
        --without-pip $BUILD_PREFIX/venv \
        --sysconfigdata-file "$sysconfigdata_fn" \
	--machine ${machine} \
        --cc ${CC:-@CC@} \
        --cxx ${CXX:-@CXX@}

Am guessing $CONDA_BUILD_SYSROOT is undefined when building with rattler-build. Though maybe there is another change or explanation that I'm missing

[beckermr]

We don't build with rattler build in ci-setup, so it is something else.

[jakirkham]

AFAIK we don't set CONDA_BUILD_SYSROOT here (with the exception of macOS)

Note this error happens for the CPU-only ARM/PPC builds here (not CUDA ones). In case that provides any clues

Separately: Maybe we should switch conda-forge-ci-setup to use rattler-build? It would be good to have a check here to confirm that we are able to build with rattler-build before publishing new conda-forge-ci-setup packages

[beckermr]

I think this is conda-build 24.9

I get the same error with 24.7.1

[beckermr]

Something shipped today broke everything. The CI ran fine this morning.

[jakirkham]

Could this be related ( conda-forge/binutils-feedstock#76 )?

[jakirkham]

Discussion continues in PR: #349

[jakirkham]

In any event, think we agree this is not related to the changes in this PR. Just external changes that are now affecting conda-forge-ci-setup

Given that, decided to kick the tires on this change. Recently, while I was out, my colleagues added a brand new feedstock numba-cuda to conda-forge, which uses the new recipe format. Of course no packages built from that at the time. Started a fresh CI build to see if it would upload packages. It appears to have worked! 🎉 Will work with them on testing these packages out

Also have started a fresh build of jolt-physics. Let's keep an eye on how that goes

[jakirkham]

Also have started a fresh build of jolt-physics. Let's keep an eye on how that goes

This unfortunately failed. Though for reasons that appear to have to do with that feedstock itself

Am guessing it just needs PR ( conda-forge/jolt-physics-feedstock#2 ) to fix it