sageWithDoc: python 3.12 fixes

[collares]

Description of changes

gcc 13.3.0 seems to exploit some undefined behaviour around setjmp/longjmp to do some optimizations that completely break Sage's GAP interface, see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114872 for the upstream investigation and https://trofi.github.io/posts/312-the-sagemath-saga.html for @trofi's analysis. This only affects Python 3.12.

Unfortunately, the minimal upstream patch at sagemath/sage#37951, which was supposed to address the issue, is not enough to fix the issue on Nixpkgs: GCC still emits code in which an always-null pointer is dereferenced with no null checks. Arch is also seeing crashes even with the upstream patch applied. As a stopgap, we compile the relevant file with -O1 (see gap-element-crash.patch)

I also tried other patches similar to sagemath/sage#37951, like having volatile v0, v1 and v2 variables and populating the relevant variables in the 1-arg, 2-arg and 3-arg cases (see here), but that makes crashes more frequent, presumably GCC can then easily infer that __pyx_t_4 (a temporary used in Cython-generated code to hold expressions of the form <GapElement>a[i]) is always non-null when jumping to the error handling cleanup code without longjmp. It then removes the null check, but it also performs other optimizations which notice that __pyx_t_4 is always zero. The case where we jump to the error handling code after longjmp, which is what Sage tests, is undefined behaviour and leads to a crash.

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.

[collares]

This is a big hack, but Sage is missing a lot of volatile variable qualifiers and sagemath/sage#37951 wasn't enough to stop the crashes with Python 3.12 and the newest GCC. I will keep investigating this, but this seems like an acceptable compromise for now. I don't think clang implements the exact same optimizations, so a GCC-only fix seems OK.

[collares]

I think numpy should be propagating this because it is required to use numpy.f2py (see https://numpy.org/doc/stable/f2py/buildtools/distutils-to-meson.html), but in this case I think it should also make the ninja binary accessible. In the meantime, I will add meson-python here and ninja to Sage's runtime path.

(cc @mweinelt)

[7c6f434c]

Looks good on reading through

[collares]

Thanks for the review!