llvmPackages: 17.0.6 -> 18.1.5 on Linux

[alyssais]

Description of changes

The next staging cycle will likely have Rust using LLVM 18 (#309580), so to avoid ending up with more LLVMs in closures than necessary, it's time to start working on upgrading the default in general.

I've built all my systems with LLVM 18 with no build regressions (which is better than 16 -> 17, where there was 1), so hopefully it'll just be a few leaf packages that need to have their LLVM versions pinned. Or maybe we'll be lucky, and the pinning from last round will mean there aren't any?

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.05 Release Notes (or backporting 23.05 and 23.11 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[RossComputerGuy]

I see we use else 17, is there a reason why we don't change it to 18 or are we only sticking to updating Linux to 18?

[RossComputerGuy]

Tried a nixpkgs-review (big mistake with all the nix log errors) and a lot of packages are borked on aarch64-linux.

900 packages marked as broken and skipped
23748 packages failed to build

[alyssais]

Tried a nixpkgs-review (big mistake with all the nix log errors) and a lot of packages are borked on aarch64-linux.

900 packages marked as broken and skipped
23748 packages failed to build

That's a very surprisingly high number, especially considering most of my test builds were on aarch64-linux. I often find nixpkgs-review struggles to handle and isn't very useful for builds that big, because it eats memory, causes huge CPU contention leading to test timeouts, etc. Can you test a sample of the packages it reported broken and see if they actually are?

[RossComputerGuy]

Can you test a sample of the packages it reported broken and see if they actually are?

One of them was every package in the expidus set. I can't provide any useful info since I'm having ISP issues (thanks Cox for your wonderful service which never has issues /s).

[RossComputerGuy]

I tried on the aarch64 builder and it appears that works better. Must be Apple Silicon related weirdness.

[alyssais]

Builds fine for me on Apple Silicon, fwiw.

[RossComputerGuy]

Strange, I know I'm on an older version of things (23.11 with 6.6.0-asahi) but it at least has better results on the aarch64 builder box.

[alyssais]

If you only tried building with nixpkgs-review, it was very probably the CPU or memory contention issues I mentioned before.