cpython: provide libuuid for the _uuid module

[mweinelt]

Relying on libuuid offers synchronization primitives, so that "no two
processes can obtain the same UUID"¹.

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.

[booxter]

I think I misled you in Matrix chat. What's actually happening is that Popen() is called once per process, so it's not a big deal performance-wise. Actually, the in-python implementation using clock source is more performant than calling to _uuid -> libuuid. See:

$ cat test.py
import uuid

for i in range(1000000):
    _ = uuid.uuid1()

$ time result/bin/python test.py # <- python with _uuid
result/bin/python test.py  5.03s user 6.76s system 96% cpu 12.160 total

$ time python test.py # <- python without _uuid
python test.py  3.84s user 0.30s system 99% cpu 4.172 total

But: it looks like unless _uuid module is used to generate UUID, it will be marked as is_safe=unknown, as per https://docs.python.org/3/library/uuid.html#uuid.SafeUUID which may be important to users of the module if they want to guarantee uniqueness for uuid across processes (which makes sense, since clock is used as the source of "randomness" otherwise). So maybe for the sake of safety - and to align with other platforms like RHEL or even Darwin nixpkgs builds? - we may want to still switch to libuuid. Something to think about.

(UPD: Looks like while Darwin nixpkgs build ships _uuid, the platform does not support _safe variant of the uuid_generate_time function, so it's not an example to emulate.)

[booxter]

Yes, let's do it (on staging, obviously). 🙏

[mweinelt]

Doing some test builds on my hydra and using master as a base for comparisons.

Not sure if the util-linuxMinimal changes are fine.