ghc: hadrian patch fix for fully_static

[NorfairKing]

Description of changes

Fixes #208959 and #316850

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.

[NorfairKing]

This will need a backport to nixos-24.05 as well.

[maralorn]

We could theoretically only apply this patch for the exact configuration which we want to use it for. i.e. enableShared = false;. Not sure. @sternenseemann, what do you think?

[NorfairKing]

We can also only turn on this patch when enableShared is false, to make it less intrusive.

Edit: Started typing that before the previous comment was posted.

[sternenseemann]

Thanks for tracking this patch down.

The commit message should use haskell.compiler.ghc96 as prefix since it is more specific.

[sternenseemann]

Please use lib.optionals for consistency with the rest of the logic.

[sternenseemann]

Please use the name argument to fetchpatch to give this patch a descriptive name which helps when debugging.

[sternenseemann]

We could theoretically only apply this patch for the exact configuration which we want to use it for. i.e. enableShared = false;. Not sure. @sternenseemann, what do you think?

No, patches should be unconditional if possible to reduce complexity.

[NorfairKing]

@sternenseemann Thanks for the quick review. I've fixed what you mentioned, I think.

[sternenseemann]

Ah, one more thing: You need to remove the broken flag from meta in the expression since that is based on the problem we are fixing here.

[NorfairKing]

@sternenseemann In order! What's the rest of the process? Is there any info about how to do a backport?

[sternenseemann]

I'll merge this for the next haskell-updates iteration so we don't block fixes in the current iteration much longer.

[sternenseemann]

Is there any info about how to do a backport?

The backport will happen semi-automatically via github actions. The process is also described in CONTRIBUTING.md

[github-actions]

Successfully created backport PR for staging-24.05:

• [Backport staging-24.05] ghc: hadrian patch fix for fully_static #318356

[nh2]

Note that e.g. on nixos-24.05 there's still this code that makes pkgsStatic use GHC 9.4 instead of the default 9.6:

  haskellPackages = dontRecurseIntoAttrs
    # Prefer native-bignum to avoid linking issues with gmp
    # GHC 9.6 rts can't be built statically with hadrian, so we need to use 9.4
    # until 9.8 is ready
    (if stdenv.hostPlatform.isStatic then haskell.packages.native-bignum.ghc94
    # JS backend can't use gmp
    else if stdenv.hostPlatform.isGhcjs then haskell.packages.native-bignum.ghc96
    else haskell.packages.ghc96)

https://github.com/NixOS/nixpkgs/blob/3a0e59b841eff01ba5535b1cc807a60eff543c66/pkgs/top-level/all-packages.nix#L15900-L15907

This PR fixes at least part of that; we should see if that can be changed now.

There's still this problem:

• GHC 9.8 TemplateHaskell doesn't work in pkgsStatic #275304