[staging-next] python311Packages.levenshtein: fetch submodules

[bryango]

Description of changes

Fix currently broken build @vcunat @fabaff

• staging-next 2024-01-18 #281828 (comment)
• https://hydra.nixos.org/build/247364986/nixlog/2/tail

This allows access to the vendored dependency rapidfuzz-cpp, which is older (2.0.0) than the nixpkgs version (3.0.0).

The nixpkgs rapidfuzz-cpp is kept in buildInputs, which enables us to seamlessly switch back to it when levenshtein catches up in the future.

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/)
• 24.05 Release Notes (or backporting 23.05 and 23.11 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
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[vcunat]

The nixpkgs rapidfuzz-cpp is kept in buildInputs, which enables us to seamlessly switch back to it when levenshtein catches up in the future.

I probably wouldn't do that, at least not this way. But let's hear feedback from others.

[bryango]

The nixpkgs rapidfuzz-cpp is kept in buildInputs, which enables us to seamlessly switch back to it when levenshtein catches up in the future.

I probably wouldn't do that, at least not this way. But let's hear feedback from others.

Indeed, this seems a bit flaky (no pun intended), but upstream has a nice logic to detect system supplied dependencies and only uses vendored ones if they are not found:

https://github.com/rapidfuzz/Levenshtein/blob/main/CMakeLists.txt

find_package(rapidfuzz 2.0.0 QUIET)
if (rapidfuzz_FOUND)
    message(STATUS "Using system supplied version of rapidfuzz-cpp")
else()
    message(STATUS "Using packaged version of rapidfuzz-cpp")
    add_subdirectory(extern/rapidfuzz-cpp)
endif()

So I guess it would be okay, this time? I'll also wait for feedbacks and change accordingly.

[vcunat]

Let me merge this PR as is, so that the build is fixed at least. Further improvements can be done later.

[dotlambda]

Please add python3.pkgs.levenshtein to rapidfuzz-cpp's passthru.tests .