gcc: only use zerocallusedregs on gcc11+ x86, aarch64

[conroy-cheers]

-fzero-call-used-regs flag isn't supported by all platforms; for example, the following currently fails to build on master:

nix-build -A hello --arg crossSystem '{
      config = "arm-none-eabi";
      libc = "newlib";
      gcc = {
        cpu = "cortex-m0";
        thumb = true;
      };
    }'

This change restricts the -fzero-call-used-regs flag to x86 and aarch64 platforms.

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/)
• 25.05 Release Notes (or backporting 24.11 and 25.05 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.

[tobim]

Not restricting it when it was originally introduced in #274089 was probably an oversight.

[risicle]

It was less an oversight, more that the limited supported platforms is helpfully not documented (https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fzero-call-used-regs) and I struggled to find reliable information on this short of reading the (byzantine) gcc source.

I'd be interested to know what information we're basing the set of supported platforms on - for instance I'm pretty sure this hasn't caused issues on i686 or risc-v since it was enabled by default, both of which this would disable it for.

[conroy-cheers]

For context, I originally ran into this compiling embedded firmware for cortex-m0, which would be the armv6-m arch. Perhaps we can run a minimal test for each of GCC's supported architectures to see which support the flag.

[risicle]

Just going to put my above objections to this on the record lest someone see the above approval and reach for the merge button. This goes too far and disables zerocallusedregs for platforms it seems to work fine on.

Possibly we should explicitly disable it for some targets it's known not to work for, but TBH I don't think people should be using wrapped compilers for embedded targets anyway. cc-wrapper makes a lot of assumptions that it's building something for running on a general-purpose os and it's a bit of an accident that it works at all for embedded targets.

See also the rat's nest of issues around #323869 which exposed a bunch of places in nixpkgs that were using wrapped compilers to build GPU kernels, BPF filters etc.