libidn2: hack to avoid referencing bootstrap tools

[vcunat]

Due to bootstrap tools getting purged from closure of libidn2.dev,
a very large rebuild is caused.

---

Fixes #175693

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.

[vcunat]

Eh, well, the conflict is trivial to resolve, now I rather want this to be testable without a stdenv rebuild.

This approach is the best I could come with, though perhaps someone will have a better idea or some nits to improve in this PR.

[roberth]

This leads to unexpected results when requesting libidn2 from the CLI, which generally works with drvPath and its libstore-level outputs, rather than the outputs attribute. I'm proposing to redefine "package" in Nix here NixOS/nix#6507.

[roberth]

Presumably it works fine when used in expressions though.

It is only a bit of a problem when doing nix-build -A libidn2.

[vcunat]

What are you getting from that command? I'm getting the right -bin output, i.e. the one without boostrap tools in closure. (also with nix build)

[vcunat]

libidn2.all is a bit confusing, though I don't think it has so much usage. I also tried to make that not too bad... so you're missing info and devdoc, and out only differs by adding a symlink layer.

[roberth]

Right, it's more of a problem with the new CLI's output selector syntax, which operates at the store level (via drvPath) rather than the expression level.

[roberth]

$ nix-build -A libidn2.info
/nix/store/klbv14mif5q34nqj04v2bgv9d7qv2pk6-libidn2-2.3.2-info

$ nix build .#libidn2^info --print-out-paths 
warning: Git tree '/home/user/h/nixpkgs' is dirty
error: derivation '/nix/store/zy3r4mqqb8wsljg1cfdnw02mqplaqw8n-libidn2-2.3.2.drv' does not have wanted outputs 'info'

[vcunat]

Oh I forgot to mention that stuff like libidn2.info.bin is also broken by such overrides, and I don't know a nice way of fixing that. (But again, I think it's rarely used.)

[vcunat]

I'll assume that we take this until a better proposal appears.