feat: support if(...) conditional package dependencies

[Hofer-Julian]

Description

Add support for conditional dependencies in the [package] section using an if(<expression>) key inside the dependency tables, e.g. [package.build-dependencies."if(host_platform != build_platform)"]. The expression is passed through to rattler-build. The old [package.target.*] syntax keeps working but emits a deprecation warning.

Fixes #4625

How Has This Been Tested?

pixi run test

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.

Tools: {e.g., Claude, Codex, GitHub Copilot, ChatGPT, etc.}

Checklist:

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have added sufficient tests to cover my changes.
• I have verified that changes that would impact the JSON schema have been made in schema/model.py.

[Hofer-Julian]

@baszalmstra I've adapted a lot of tests to use the pass-through backend instead and removed the example satisfiability tests as we discussed.

However, I am a bit unsure about your "simple_test", maybe you can check this one out

[GuillaumeQuenneville]

Will this support all the parameters / virtual packages from the specificiers "rich" platforms pr? #4458

I am particularly interested in using if(__cuda)

at the moment the way to swap between cuda and non cuda dependencies is through the pixi environments. This works for workspaces but not for building packages. There is no syntax at the package level to define "these deps if CPU, these deps if GPU" as far as i know without going to rattler-build.

in the example you added in the latest commit. how do you specify to a cuda and non-cuda variant for windows and linux? This example would break for all workspace users on a machine without an nvidia gpu since they dont have the __cuda virtual package.

# --- Conditional dependencies: the feature this example shows off ---------
[package.run-dependencies."if(linux or win)"]
cuda-version = "12.*"
cuda-nvcc = "12.*"

[Hofer-Julian]

TODO for myself:

• make sure that invalid syntax is correctly rejected by rattler-build
• check whether we should bump api-version one more than v3 of it is fine to just put both changes in same bump

[GuillaumeQuenneville]

cuda toolkit is installable on systems without cuda drivers and the CMakeLists only checks for presence of cuda toolkit. From my understanding, users on win and linux would build unuseable cuda binaries that crash due to missing drivers instead of falling back to the CPU build.

it would be useful if the example showed how to use build variants to have a CPU and CUDA version. And how switch between them on Windows and Linux depending on the presence of the __cuda virtual package.