treewide: skip generating shell completions using $out/bin/… when cross compiling

[jcaesar]

Description of changes

Fixes regressions for cross compiling introduced in #289517 on 61 packages. See #308283 for more discussion.

There are a lot of packages that use the following pattern to generate shell completions:

  postInstall = ''
     installShellCompletion --cmd foo \
       --bash <($out/bin/foo generate-completions bash) \
       …
  '';

The crucial part is that it executes $out/bin/foo, which failed silently on cross builds so far (ignoring binfmt_misc here), but is now raised as a build error from installShellCompletion.

Most of what this PR does is prefix (that part of) those postInstall scripts with:

lib.optionalString (stdenv.buildPlatform.canExecute stdenv.hostPlatform)

Some additional notes:

• This is limited to 61 packages, there are potentially 327 instances of this.
  • I'm not sure this will get merged and fixing ~300 packages manually is too much considering that.
  • They're all buildRustPackage, because I'm familiar with that ecosystem, and the completion generation logic is nearly all the same, so it was easy to investigate problems for me.
• comodoro, attic-client, neverest, himalaya, colmena: Since I already stumbled upon them, it also changes some instances of buildPlatform == hostPlatform to buildPlatform.canExecute hostPlatform.
• trashy: Move from preFixup to postInstall for consistency
• charasay: Fix generation command
• inlyne: Use checked-in completion scripts

Checking

nixpkgs-review isn't super insightful for this, since I took extra care to not introduce any changes, not even whitespace, where unnecessary:

Result of nixpkgs-review run on x86_64-linux 1

3 packages built:

• charasay
• inlyne
• trashy

So, I also obtained a list of changed packages

 git diff-tree --no-commit-id --name-only -r HEAD~2 | each { path dirname | path basename } | str join " "

and tested that all build, with:

env NIXPKGS_ALLOW_UNFREE=1 nix shell --keep-going --impure -vL --expr 'let nixpkgs = builtins.getFlake "git+file:."; in with nixpkgs.legacyPackages.x86_64-linux.pkgsCross.aarch64-multiplatform; mkShell { packages = [comodoro flavours genact inlyne leetcode-cli git-absorb jujutsu youki ab-av1 ast-grep attic-client clipcat codeberg-cli engage himalaya joshuto neverest pace pixi qrtool release-plz sn0int steamguard-cli zola jrsonnet cocogitto diesel-cli espup fnm cargo-show-asm cauwugo typeshare sentry-cli snazy tokio-console volta  dufs colmena charasay dotter fd intermodal miniserve onefetch sheldon starship topgrade trashy twm zellij magic-wormhole-rs fh nix-template bws jwt-cli kbs2 prs rbw  shisho languagetool-rust termbook]; }'

(skipping maa-cli sheesy-cli deno tremor solana-cli elf2nucleus because they fail to compile or link.)

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/)
• 24.11 Release Notes (or backporting 23.11 and 24.05 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
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[pbsds]

diff LGTM sans one error, i also support this approach until someone rather decides to go the QEMU route and puts the work into actually implementing it

[pbsds]

the presets below are not conditioned on canExecute

[jcaesar]

Nice catch. Fixed.

Also rebased on master, and additionally had to fix the formatting of maa-cli now to get a green CI…

[pbsbot]

Result of nixpkgs-review pr 326243 run on x86_64-linux 1

3 packages built:

• charasay
• inlyne
• trashy

[pbsbot]

Result of nixpkgs-review pr 326243 run on x86_64-linux 1

4 packages built:

• charasay
• inlyne
• starship
• trashy