cudaPackages.cuda_nvcc: wrap binary so NVCC can always find compiler

[ConnorBaker]

• Wrap NVCC so it can always find a compatible compiler on the path.
• Set compiler-bindir in nvcc.profile instead of in setupCudaHook so it is always set, even when NVCC is run in ad-hoc fashion in installed environments rather than inside mkDerivation.

These changes needed to be made to allow the tests for pycuda (#465047) to work.

I was able to build python3Packages.onnxruntime and firefox with:

{pkgs}: {
  allowUnfreePredicate = pkgs._cuda.lib.allowUnfreeCudaPredicate;
  cudaCapabilities = [ "8.9" ];
  cudaForwardCompat = false;
  cudaSupport = true;
}

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

[SomeoneSerge]

Probably want to break up the hook's dependency on the cc too 🤔 Because now we have 3 references all from different places

[SomeoneSerge]

Ah damn this is off by 1 again: nvcc here is hostTarget, and cc - buildHost

[SomeoneSerge]

OTOH, we never made cross actually work, might as well postpone the larping

[ConnorBaker]

cc is unspliced but backendStdenv is spliced and from pkgsHostTarget (the default package set drawn from by callPackage) so I would hope cc is also from hostTarget.

[SomeoneSerge]

Nooo that's the thing about stdenvs, they're always offset by -1:

$ nix eval -f "<nixpkgs>" --argstr crossSystem aarch64-linux stdenv.cc.name
"aarch64-unknown-linux-gnu-gcc-wrapper-14.3.0"

But maybe we should merge regardless and fix whenever

[SomeoneSerge]

Yeah screw it