python3Packages.cffi: fix tests on FreeBSD

[rhelmot]

Description of changes

Followup to #328776, depends on #328943. See both for context.

I don't believe either of these changes are worth submitting upstream. One is modifying the existing patch to do what it claims to do and not just scoped to darwin, and one is just disabling a test. It's kind of a miracle that this is all it takes to pass tests, seeing as there are various comments to the effect of assuming that "anything that's not darwin or win32 must be linux and act exactly like linux". FreeBSD's libdl (the subject of the skipped test) claims to support dlopen, but it actually only has a stub, and if you want the real dlopen you have to link against libc, but you actually need the symbol from the dynamic loader? It's a bit of a mess and I want out of here.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
  • x86_64-freebsd
• 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.

[Ericson2314]

The first part, about the clang issue, should go in the commit message.

The second part, about the disabled test, should be a code comment.

[mweinelt]

Please send this upstream.

[Ericson2314]

I don't believe either of these changes are worth submitting upstream.

What @rhelmot wrote on this subject. Keep in mind we are modifying an existing not-upstreamed patch?

[Ericson2314]

There is still a PR description that should be a commit message, however.

[rhelmot]

omg @Ericson2314 I am so sorry, I rebased the old version of the patch before I addressed your original review comments. They should be fixed now.

[mweinelt]

@reckenrode @tjni @abathur Please upstream the patches that we currently have on cffi.

I'm not at all happy to keep them for years and maintain them downstream.

[reckenrode]

darwin-use-libffi-closures.diff is a workaround for the old SDK on x86_64-darwin. It can be dropped for 25.05 since the default will be newer than 10.15.

clang-pointer-substraction-warning.diff can probably be dropped. Upstream addressed the issue in python-cffi/cffi#78.

[mweinelt]

Okay, then we are back to reporting this issue and a potential patch upstream, before we choose to pick it downstream.

[emilazy]

The patch isn’t actually specific to old SDK versions, it’s specific to using upstream libffi on macOS rather than Apple’s fork. Options include depending on darwin.libffi to just use the real thing, setting the -DFFI_AVAILABLE_APPLE compile flag, or just removing the condition upstream since https://groups.google.com/g/python-cffi/c/xU0Usa8dvhk seems lukewarm on that.

[emilazy]

FWIW the postPatch stuff is obsolete.

[mweinelt]

Eval failures are unrelated.