Fix transitive subpackage dependency resolution

[isuruf]

Description

Checklist - did you ...

• Add a file to the news directory (using the template) for the next release's release notes?
• Add / update necessary tests?
• Add / update outdated documentation?

Fixes #3308

[codspeed-hq]

CodSpeed Performance Report

Merging #5603 will not alter performance

_{Comparing isuruf:transitive-subpackage (aa8a2e9) with main (cf35214)}

Summary

✅ 5 untouched benchmarks

[beckermr]

I need to read in more detail, but should we add a specific test for which transitive deps get picked up into the envs?

Right now the tests only ensure that the recipe from the issue builds.

[beckermr]

The logic looks correct, but I am fuzzy on the various m.is_cross branches. We should a test that specifically looks at the resolved envs for the recipe to ensure it has the right packages.

[isuruf]

We should a test that specifically looks at the resolved envs for the recipe to ensure it has the right packages.

I'm not sure what you mean here. The test that I added fails to build with the current main branch.

[beckermr]

What I mean is that in the original issue, the output of conda render was shown. This demonstrated that the various envs made during the build did not have the right packages in them.

I am saying that in addition to ensuring the recipe from the issue builds, we should explicitly verify that the envs made during the build have the correct transitive packages in them.

This will help future maintainers debug any test failures.

[isuruf]

pre-commit.ci autofix

[wolfv]

I am not sure this is relevant, but in rattler-build we actually create variants for each exact=true so that the tree is unfolded correctly :)

[isuruf]

I am not sure this is relevant, but in rattler-build we actually create variants for each exact=true so that the tree is unfolded correctly :)

That's a different issue. Or rather the second comment in the same issue. See #3308 (comment)