gforth: override swig, not swig3

[bcdarwin]

Description of changes

Removes one of two remaining usages of the top-level swig3 expression. (The other comes from older LLDB versions.)

Also removes use of overrideDerivation in favour of preferred overrideAttrs.

Unfortunately gforth still needs a custom fork of swig 3.0.9, but if absolutely necessary we could perhaps remove the swig dependency altogether, at the cost of some C interface support (I don't exactly understand what is lost by removing support from the compiler vs removing access to a standalone Forth-compatible swig, though).

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.

[emilazy]

I’m wondering if the better approach wouldn’t be to get rid of the old LLDBs (cc @RossComputerGuy), and then just move pkgs/development/tools/misc/swig/3.x.nix into the Gforth directory and specialize it to this version. Though if we think the SWIG derivation will continue being compatible with this old version going forward this could be fine too, especially since the 3.x and 4.x derivations are pretty duplicated currently.

[emilazy]

Okay actually apparently the patch to fix old LLDB versions is really simple: swig/swig#2377 (comment). So let’s just do that.

[emilazy]

And actually the patch was upstreamed as of LLVM 16: llvm/llvm-project@f0a25fe. So only LLDB 12–15 would need patching or dropping.

[emilazy]

LLDB 12 and 13 are already broken (on aarch64-linux at least); 14–17 work with SWIG 4 with no code changes. I don’t entirely understand why given the upstream patch, but it seems like we can just remove the conditional entirely and be left with no SWIG 3 users other than Gforth.

(Sorry for the only tangentially‐related noise on this PR. If you think the override approach here is the most maintainable thing going forward, let’s do that.)

[bcdarwin]

I’m wondering if the better approach wouldn’t be to get rid of the old LLDBs (cc @RossComputerGuy), and then just move pkgs/development/tools/misc/swig/3.x.nix into the Gforth directory and specialize it to this version. Though if we think the SWIG derivation will continue being compatible with this old version going forward this could be fine too, especially since the 3.x and 4.x derivations are pretty duplicated currently.

I don't mind the current overriding - swig seems fairly stable and this seems a bit cleaner than copying it over, and this one package does not really deserve a swig refactoring. (The weird pcre2 = pcre can just be solved by renaming the variable in swig4 if we care.)

[emilazy]

Alright, let’s just merge this. The whole thing is kind of gross anyway :)

I’ll put up a PR for LLDB.