python3Packages.warp-lang: 1.9.0 -> 1.9.1, fix build

[yzx9]

• fix build (switching to LLVM 19)
• dynamic linking clang / LLVM (~200MB -> 686KB)
• 1.9.0 -> 1.9.1

Hydra: https://hydra.nixos.org/build/310728547

ZHF: #457852

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.

[yzx9]

@ConnorBaker Hi! Could you please take a look at why I’m running into this error?

> NIXPKGS_ALLOW_UNFREE=1 nix build --impure .#python3Packages.warp-lang.passthru.tests.cuda.cudaOnly

error: Package ‘cuda12.8-cuda_nvcc-12.8.93’ in /nix/store/r5pa0cg7psiqm1whgwwwswzspsg82zcl-source/pkgs/development/cuda-modules/packages/cuda_nvcc.nix:174 is marked as broken, refusing to evaluate.

I suspect it might be caused by the following code, but I’m not sure why pkgsLLVM is being used on an x86_64-linux machine:

nixpkgs/pkgs/development/cuda-modules/packages/cuda_nvcc.nix

Lines 156 to 171 in 038f97c

	brokenAssertions = [
	# TODO(@connorbaker): Build fails on x86 when using pkgsLLVM.
	# .../include/crt/host_defines.h:67:2:
	# error: "libc++ is not supported on x86 system"
	#
	# 67 | #error "libc++ is not supported on x86 system"
	# | ^
	#
	# 1 error generated.
	#
	# # --error 0x1 --
	{
	message = "cannot use libc++ on x86_64-linux";
	assertion = backendStdenv.hostNixSystem == "x86_64-linux" -> backendStdenv.cc.libcxx == null;
	}
	];

[ConnorBaker]

Use --show-trace --trace-verbose since that will print the failed assertions.

EDIT: make sure you’re setting config.cudaSupport as well when you instantiate Nixpkgs

[yzx9]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 456471
Commit: bde133813053445a0485231aaf61354f3b18a6be

---

x86_64-linux

✅ 4 packages built:

• python312Packages.warp-lang
• python312Packages.warp-lang.dist
• python313Packages.warp-lang
• python313Packages.warp-lang.dist

---

aarch64-darwin

✅ 4 packages built:

• python312Packages.warp-lang
• python312Packages.warp-lang.dist
• python313Packages.warp-lang
• python313Packages.warp-lang.dist

[yzx9]

@ConnorBaker Thanks! You’re right. It was caused by the fact that I hadn’t enabled cudaSupport yet. It works now, although I still think the message is a bit confusing.

I’ll run a nixpkgs-review for the unit tests later. It’s a bit tricky to do right now since Torch just got upgraded and the cache is missing.

Update: about message #456908

[yzx9]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 456471 --extra-nixpkgs-config '{ allowUnfree = true; cudaSupport = true; }' --package python3Packages.warp-lang.passthru.tests
Commit: bde133813053445a0485231aaf61354f3b18a6be

---

x86_64-linux

✅ 3 packages built:

• python3Packages.warp-lang.passthru.tests.cpu
• python3Packages.warp-lang.passthru.tests.cuda.cudaOnly
• python3Packages.warp-lang.passthru.tests.cuda.cudaWithLibmathDx

[yzx9]

I noticed that warp-lang v1.10.0 has been released. However, I still hope this PR can be merged and backported to 25.11 to fix the build issue during the ZHF period.
Since v1.10.0 announced the removal of the warp.sim module, it’s not suitable for backporting. I’ll open a new PR to bump the version instead.

[ConnorBaker]

Apologies for the delay getting this approved and merged! Thank you for maintaining this package, it's one of the few I use :)

[yzx9]

@ConnorBaker Thanks for the review! Great to see you’re care about CUDA support again.

[ConnorBaker]

Oh I’m still absolutely burnt out :)