mbrola-voices: extract from mbrola, allow language selection

[mweinelt]

Allow filtering voices based on any number of extglob patterns.

Description of changes

Restricting mbrola-voices to only one voice per language saves 387M, and is what I recommend we should do on the ISO, because it likely doesn't compromise accessbility too much.

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.

[vcunat]

kept as a fixed-output derivation to limit "duplicates"

I wonder if it's worth the effort to make the default one a FOD again.

For the files that often remain in people's closures. I think it could make a significant difference for some people, unless they have nix.settings.auto-optimise-store = true; or filesystem-level deduplication.

[mweinelt]

I wonder if it's worth the effort to make the default one a FOD again.

Defaulting to the derivation source now for mbrola, so the default is effectively that again.

[vcunat]

I think that piece of code doesn't work as intended? Maybe because of the self-reference. I still see this in the default closure

/nix/store/8ja11xcr6dbvh1bfc0p3ny79jdwwgvmg-mbrola-voices-0-unstable-2020-03-30

[vcunat]

Maybe we could write mbrola-voices like this?

if languages == [ ] then src // { inherit meta; } else stdenv.mkDerivation { inherit src meta; /*...*/ }

[vcunat]

🤔 the overlay approach won't be a good longer-term solution, as it really needs lots of rebuild work to get the ISOs (every time a rebuild happens). Users would be slightly impacted on installation from that media, as hashes of those packages would differ, so they couldn't be pulled from the ISO.

Using the reduced set of voices by default... maybe, I'm no good with accessibility. Decoupling the data and plugging it by some environment variable is probably the best shot for longer-term solutions, though it's "less pure". Maybe this WIP PR: #329658

[vcunat]

For the immediate unblocking I increased the limit on Hydra a bit, so that we have some time to improve the closure sizes.

EDIT: here https://hydra.nixos.org/build/267286447#tabs-buildsteps

[vcunat]

Superseded by #329658

[vcunat]

At least I think we're not missing anything now.