BUG fix zipping for c stdlib

[conda-forge-admin]

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

I've started rerendering the recipe as instructed in #5591.

If I find any needed changes to the recipe, I'll push them to this PR shortly. Thank you for waiting!

Here's a checklist to do before merging.

• Bump the build number if needed.

Fixes #5591

[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.

[hmaarrfk]

how is this proposed logic any different than before? they seem equivalent.

[beckermr]

Two changes

• it covers platforms like s390x etc. these not getting covered caused rendering issues.
• it zips these keys with the cdt ones since both need to match IIUIC

[h-vetinari]

I was hoping we wouldn't have to have the compiler and stdlib in the same zip... Definitely the cdt needs to be zipped with the stdlib, but perhaps we can split that group out of the mega-zip, at least once we've dropped cos6 this summer.

[beckermr]

Possibly? I don't see the issue here really except that it makes matching things up somewhat error prone.

[h-vetinari]

Possibly? I don't see the issue here really except that it makes matching things up somewhat error prone.

It's not a big issue per se, but it gets hard to change anything non-trivial about such a large zip in a local CBC without running into breakages. Also conceptually, compiler and stdlib shouldn't have to be coupled. But again, not an issue for now. Thanks for fixing it!

[jakirkham]

If we do this, we need to update the CUDA migrators too as they touch this key. Currently migration is broken on any CUDA feedstock ( for example: conda-forge/ucx-split-feedstock#158 (comment) )

That said, this kind of smells like a bug in conda-smithy

cc @mbargull (in case you have any ideas 🙂)

[jakirkham]

Also for CUDA this needs to be 2.17 (even on x86_64)

[h-vetinari]

If we do this, we need to update the CUDA migrators too as they touch this key.

Yeah, this is the kind of thing why I would have liked to avoid adding the stdlib into that mega-zip_keys... 😑

I think we'll have to duplicate much of the os.environ.get("CF_CUDA_ENABLED", "False") etc. stuff also for the stdlib version.

[beckermr]

Oh no sorry! We have to zip it with cdt so we're stuck. How can I help fix things? I'm not sure what is needed.