SDL2: point alias to sdl2-compat

[LordGrimmauld]

This is a mass rebuild mostly because of ffmpeg. The only component in ffmpeg depending on sdl is ffplay, which is easy enough to test.
Most other rebuilds caused by this are leaf packages (or close to being leaf packages) and thus can be fixed wither once this hits staging-nexet or retroactively on master once this passes through hydra.

sdl2-compat no longer propagates libX11, libGL or libGLU. movit has somewhat used dependencies like Krita, so i fixed it. I also fixed the-powder-toy with similar breaks as an example. Switching to sdl2-compat automagically made it use wayland-native, which is neat.

I do expect more apps to break due to losing libX11/libGL/libGLU and will fix those in follow-up(s). This change is important imo, because SDL2 has no maintainer in nixpkgs and is basically unmaintained upstream too.

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.

[LordGrimmauld]

Alright, eval should now be fixed - the SDL2 specific overrides obviously fail on current sdl2-compat, so those were switched back to SDL2_legacy

[LordGrimmauld]

This was cleanup for the commit history, makes review easier and there now is no single commit that doesn't eval

[marcin-serwin]

I've run into some issues when moving some packages manually from SDL2 to sdl2-compat, mostly minor things, like fullscreen not being on by default on wayland, but also hedgewars crashing on aarch64-linux: libsdl-org/SDL#12560. Some other apps may also be affected.

[LordGrimmauld]

I've run into some issues when moving some packages manually from SDL2 to sdl2-compat, mostly minor things, like fullscreen not being on by default on wayland, but also hedgewars crashing on aarch64-linux

apart from ffmpeg/ffplay (which i tested), none of the affected packages would be mass rebuilds. Yes, i expect things to break - but those can trivially be fixed by pinning SDL2_legacy/.SDL2_classic. Maybe even before 25.05 hits.

[LordGrimmauld]

Related: #395499

[trofi]

Proposed love_0_10 fix as well:

• love_0_10: add missing xorg.libX11 depend #395871

[trofi]

Bisect says d96d13a SDL2: point alias to sdl2-compat broke python3Packages.pygame ion staging and staging-next on x86_64-linux as:

$ nix build --no-link -f. python3Packages.pygame
...
FAILED (failures=7, errors=1, skipped=6)
====================================================================
ERROR: test_num_fingers (pygame.tests.touch_test.TouchTest.test_num_fingers)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/nix/store/ng89gwad6r45z0pl18n443izc8j15qa8-python3.12-pygame-2.6.1/lib/python3.12/site-packages/pygame/tests/touch_test.py", line 33, in test_num_fingers
    touch.get_num_fingers(touch.get_device(0))
pygame.error: Unknown touch device index -1234

[LordGrimmauld]

I had built pygame after this PR to ensure the libX11 drops didn't affect it - pygame built successfully for me.
However, to just unbreak it it might be worth trying SDL2_classic for pygame, just to see what happens.

[trofi]

https://hydra.nixos.org/job/nixpkgs/staging-next/python312Packages.pygame.x86_64-linux shows similar failures on staging-next.

[7c6f434c]

Hmm, looks like RenPy triggers vague segfaults with sdl2-compat but works fine with SDL2_classic, so builds passing is not all…

[LordGrimmauld]

Hmm, looks like RenPy triggers vague segfaults with sdl2-compat but works fine with SDL2_classic, so builds passing is not all…

feel free to pin it to SDL2_classic then! It was known sdl2-compat might not always be an adequate replacement. While upstream is trying to fix all the compatibility issues, we can pin. Ideally we wouldn't need SDL2_classic by 25.11, but for now it is around. pygames has similar-ish issues.

[7c6f434c]

So far I am not yet sure whether going deeper recursively helps (I suspect that in some cases yes, at least for pygames-sdl2)

[marcin-serwin]

Hmm, looks like RenPy triggers vague segfaults with sdl2-compat but works fine with SDL2_classic, so builds passing is not all…

Do you have some reliable repro or is it random? I've tried clicking through the tutorial but nothing happened.

[7c6f434c]

(Not to leave it without the answer completely) so far I have impression that some large games are more likely to crash when clicking through stuff quickly, so not close to reliable reproduction yet, sorry. Also, I have not yet found how much overriding is needed to bring segfaults to zero. I guess I should also check if the recent sdl2-compat bumps have improved things…

[7c6f434c]

Hm, backtrace makes no sense, the context is too large to be a usable reproduction, but the crash frequency change seems to be there…