rustc: 1.65.0 -> 1.66.0

[figsoda]

Description of changes

https://blog.rust-lang.org/2022/12/15/Rust-1.66.0.html

built and tested fd on master (x86_64-linux)

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/)
• 23.05 Release Notes (or backporting 22.11 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
  • (Release notes changes) Ran nixos/doc/manual/md-to-db.sh to update generated release notes
• Fits CONTRIBUTING.md.

[figsoda]

@ofborg build fd

[tjni]

Thank you for putting this out so quickly.

/cc @winterqt per #190093

[figsoda]

I don't have a darwin machine so I can't test the stuff mentioned in #190093, feel free to directly push to this pr if you have any changes you want to make

[winterqt]

I will do so by the end of the day (EST), thank you both.

@figsoda (and @tjni, if you want more Darwin power) https://github.com/winterqt/darwin-build-box exists as of recently, feel free to request access. :)

[winterqt]

Turns out building 11k packages wasn't needed to figure out the issue.

I've described the strategy for dealing with this in 70d39d1679e38e018a540de0a582f97e6a89cff7, as well as added a note that I believe is correct in ebce964b12f5c99d9f30b2bbe2e0f43e3c4e4713. Let me know if anything is unclear.

cc @zowoq

[tjni]

I opened vectordotdev/vector#15627 to track fixing vector when compiled using Rust 1.66.

[figsoda]

Thanks @tjni! we should be able to just remove the #![deny(warnings)] if upstream doesn't fix the issue before this gets into master

[figsoda]

zowoq hasn't had any github acitivity for 10 days, perhaps we can just merge this and fix the issues that might pop up later

[tjni]

I'm in favor of merging it.

[winterqt]

Just to confirm: does my approach to the libiconv situation make sense to you both, are there any issues? This should cause no breakages right now, but it just makes things easier for us to eventually start only adding it when needed by not propagating it alongside rustc (only adding it to the inputs of any Rust package using buildRustPackage, as we did previously).

[tjni]

It makes sense to me at a high level (I'm not as in tune with the details). To move is in that direction, what needs to happen so that:

1. Rust 1.66 is used to bootstrap.
2. Packages bump their libc versions.

[winterqt]

I think there may be a typo in your message, or is that a question?

Either way, assuming I understand your question(?) correctly:

1. libiconv being a dependency to build rustc doesn't matter too much here, and will go away when 1.66 is used to bootstrap; it doesn't propagate anymore.
2. Once enough packages in the repo bump their libc version, we can go through and find which ones fail without it, and manually add those. I'd say we can maybe start doing this around March? (Just pulled a date out of thin air.)

Does that clarify things?

[figsoda]

Makes sense to me as well, we just need to keep an eye on the packages that don't use buildRustPackage as they might still depend on an older version of libc

IIUC removing libiconv from buildRustPackage should also work, would just require a lot more work and we might as well wait a little longer for packages to update their dependencies

[winterqt]

IIUC removing libiconv from buildRustPackage should also work, would just require a lot more work and we might as well wait a little longer for packages to update their dependencies

Exactly, that'll be the next step, per my last comment.

Give me one sec to adjust the commit message, please.

[winterqt]

Let me know how those messages sound, adjusted some wording and formatting. Then we should be good to go!

[figsoda]

messages & changes both lgtm, thanks!

[tjni]

Thanks for the context. This LGTM (though might need to be rebased against latest staging, which should include the cargo auditable changes). I created #207451 to fix vector pending upstream fix.