rustc: 1.64.0 -> 1.65.0

[tjni]

https://github.com/rust-lang/rust/releases/tag/1.65.0

I'm planning to run as many builds as I can to test this.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandbox = true set in nix.conf? (See Nix manual)
• 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/)
• 22.11 Release Notes (or backporting 22.05 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
  • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
• Fits CONTRIBUTING.md.

[lblasc]

Upstream is using llvm15, we should wait for #194634

[tjni]

I'm temporarily pulling in unmerged LLVM update into this PR for testing.

[Fuuzetsu]

Might want to take #199300 at the same time as it causes full rust crate rebuilds too (though ofborg seems to not think so?)

[tjni]

If yours is ready, I'd put it on staging right away. I'm not yet sure when this will be ready, although I hope it's soon :)

[Fuuzetsu]

If yours is ready, I'd put it on staging right away. I'm not yet sure when this will be ready, although I hope it's soon :)

I can at best send a PR as I don't have commits rights. Feel free to do it yourself if you'd like.

[tjni]

I don't have commit rights either, 🤞 your PR is reviewed soon.

[tjni]

1. Created a separate PR for humility: update libusb alias and vendor hash #199794.
2. Wait for deno 1.27.1 fails to build with rust 1.65.0 denoland/deno#16540.
3. Follow Failed Compilation shuttle-hq/synth#337.

[oxalica]

@lblasc

Upstream is using llvm15, we should wait for #194634

I found this in section "Compatibility Notes" of the release note,

• Update the minimum external LLVM to 13

I guess LLVM 15 is not necessary required?

[lblasc]

hey @oxalica , off topic, thank you for all the work on rust in nix, on NixCon I heard about nocargo project which is awesome!

About the llvm compatibility, rust 1.65 should work on >=LLVM13, if I'm not mistaken rust upstream is not doing extensive testing on older LLVM versions, in past mismatched LLVM versions between nix and upstream caused some strange build and runtime issues in our monorepo which contains workspace with 80+ crates. That's the only argument I can provide, but sure LLVM14 should work just fine, maybe we can update afterwards.

[kevincox]

I was trying to build 763c18e and it seemed mostly successful. Unfortunately it didn't finish as a few packages appear to be hung:

4/11111/11124 built (179 failed)

However trying to build 137017341b809b8a0f01576849331cbafe756283 I am getting an eval error.

% nixpkgs-review pr --no-shell --post-result -c commit $(c) --build-args '-j4 --cores 32 --max-silent-time 3600'
$ git -c fetch.prune=false fetch --no-tags --force https://github.com/NixOS/nixpkgs staging:refs/nixpkgs-review/0 pull/199664/head:refs/nixpkgs-review/1
$ git worktree add /home/kevincox/.cache/nixpkgs-review/pr-199664-11/nixpkgs 137017341b809b8a0f01576849331cbafe756283
Preparing worktree (detached HEAD 137017341b8)
Updating files: 100% (32616/32616), done.
HEAD is now at 137017341b8 rustc: 1.64.0 -> 1.65.0
error: Function called without required argument "libusb" at /home/kevincox/.cache/nixpkgs-review/pr-199664-11/nixpkgs/pkgs/development/tools/rust/humility/default.nix:4, did you mean "librsb", "libusb1" or "ibus"?
(use '--show-trace' to show detailed location information)
nix --experimental-features nix-command --system x86_64-linux eval --json --impure --expr (import /nix/store/nci1s3y693yifx3kqg0cd81p8alybnqk-nixpkgs-review-2.7.0/lib/python3.10/site-packages/nixpkgs_review/nix/evalAttrs.nix { allowAliases = false; attr-json = /run/user/1000/tmpht_x_zp3; }) failed to run, /run/user/1000/tmpht_x_zp3 was stored inspection
https://github.com/NixOS/nixpkgs/pull/199664 failed to build

[kevincox]

Ignore the previous version of this message, it was triggered by a NIx bug NixOS/nix#6572

error: derivation '/nix/store/j817n3h7wqz7xa4i7i5ljda3wwd5pzdk-shards-0.17.1.drv' requires non-existent output 'lib' from input derivation '/nix/store/v67yvwfsm6s9kp6pfr4vzhfa5yr12srn-crystal-1.2.2.drv'

[tjni]

I hope that error is resolved for you soon. It sounds painful from reading the issue.

[kevincox]

It seems to be triggering frequently on that version of this PR for some reason so I'm having trouble testing. Is there anything holding this back though? May make sense to dump it into stshung if we don't have any known issues.

[tjni]

The main question I would need an owner to resolve is whether we want to track upstream's LLVM exactly. This is requested in a comment that, right now, is in 1_64.nix, which was originally committed in #76100 by @edolstra. @cstrahan @globin and @Havvy are maintainers of rustc whose opinions I think should matter.

If so, we need to wait for #194634 to land. If not, I can update this PR to do without that and create one to follow up when that's available. Without any owner opinion, I plan to wait for the LLVM update to land.

I could not run nixpkgs-review, because it required too much disk space and time. However, I did build a large sample of the applications listed on https://github.com/rust-unofficial/awesome-rust in my aarch64-darwin machine and on an x86_64-Linux machine from GitHub Actions.

If it aids you in testing, on my Cachix cache tjni, I have LLVM 15 built from a previous iteration of #194634. I haven't yet, but I'm going to spin up a GitHub Actions runner to build firefox and thunderbird.

Edit: I built thunderbird successfully. My GitHub Actions runner terminated before firefox could finish, but it ran for quite awhile without any issues.

[oxalica]

@lblasc

About the llvm compatibility, rust 1.65 should work on >=LLVM13, if I'm not mistaken rust upstream is not doing extensive testing on older LLVM versions, in past mismatched LLVM versions between nix and upstream caused some strange build and runtime issues in our monorepo which contains workspace with 80+ crates. That's the only argument I can provide, but sure LLVM14 should work just fine, maybe we can update afterwards.

I'd prefer using the current LLVM14 for now if #194634 cannot be merged in the near future. While rust 1.65.0, which is a big update, will be required for more and more crates targeting latest stable compiler, this PR would block all of them.

[kevincox]

I've been building the latest commit and it doesn't appear that there are any major breakages. I think it makes sense to move this into staging.

[tjni]

Thanks, everyone. I've updated the commit to use LLVM 14 and extracted the deno fix into #201122.

[Mic92]

x86_64-linux: firefox built
aarch64-macos: fd built
aarch64-linux: still building.