Xorg session fails to launch, probably due to mesa 24.2.1 -> 24.2.2 update

[acid-bong]

Describe the bug

I start my X session with sx (https://codeberg.org/acidbong/nixos/src/commit/369eb18078e2702053c90c38829fd6ba04457383/modules/desktop/default.nix) and it fails.

Update: it also happens with startx/xinit

Steps To Reproduce

1. Just log in (the X session is autostarted)

Expected behavior

Normally it starts without crashing.

Additional context

• Xorg log: https://gist.github.com/acid-bong/f6720ba1f6017ae4c0861c48a817bee2
• Nixpkgs branch is nixos-unstable
• My hardware:
  • CPU: Intel Core i3-7100
  • GPU: integrated

Notify maintainers

@vcunat @primeos @K900

Metadata

Please run nix-shell -p nix-info --run "nix-info -m" and paste the result.

[user@system:~]$ nix-shell -p nix-info --run "nix-info -m"
 - system: `"x86_64-linux"`
 - host os: `Linux 6.6.49, NixOS, 24.11 (Vicuna), 24.11.20240906.574d1ea`
 - multi-user?: `yes`
 - sandbox: `yes`
 - version: `nix-env (Lix, like Nix) 2.91.0
System type: x86_64-linux
Additional system types: i686-linux, x86_64-v1-linux, x86_64-v2-linux, x86_64-v3-linux
Features: gc, signed-caches
System configuration file: /etc/nix/nix.conf
User configuration files: /home/acid/.config/nix/nix.conf:/etc/xdg/nix/nix.conf:/home/acid/.nix-profile/etc/xdg/nix/nix.conf:/nix/profile/etc/xdg/nix/nix.conf:/home/acid/.local/state/nix/profile/etc/xdg/nix/nix.conf:/etc/profiles/per-user/acid/etc/xdg/nix/nix.conf:/nix/var/nix/profiles/default/etc/xdg/nix/nix.conf:/run/current-system/sw/etc/xdg/nix/nix.conf
Store directory: /nix/store
State directory: /nix/var/nix
Data directory: /nix/store/dzs6wmlzg8hdh9mjailpghvxrzs43jdn-lix-2.91.0/share`
 - channels(root): `""`
 - nixpkgs: `/nix/store/5w3dp0m37794iffsbm9vd9f1xmmhda6i-source`

---

Add a 👍 reaction to issues you find important.

[K900]

Disable xserver.videoDrivers = ["intel"].

[acid-bong]

Removing "intel" did help, but now I can't get rid of tearing 🤔

[vcunat]

X -> Wayland should solve that, but you probably didn't want to hear that 🙂

[Atemu]

Could you bisect or at least check before and after the staging-next merge containing the Mesa update?

[acid-bong]

https://codeberg.org/acidbong/nixos/commit/0ab36d9069742ba4bf84c061014140dce8a1eca7
between 574d1ea and 99dc878?

[Atemu]

If that's where you saw the regression come into effect, yes.

[vcunat]

If you think it's due to the mesa update, you could try to just revert that update on the commit that fails for you. (much less work than a full bisect)

[acid-bong]

@Atemu, @vcunat: git said 28e9b6d is the first bad commit, which is exactly the recentmost staging merge. What are my next steps?

[Atemu]

In that case do as @vcunat suggested and revert the mesa update. This will cause quite a lot of rebuilds, so I'd recommend you keep your closure as minimal as possible in this and e.g. remove all of your systemPackages, disable any optional desktop component that is not strictly necessary; basically strip your config down to the minimal config where you can still reproduce the bug and then rebuild your system with the mesa update reverted.

Though tbh, this smells like an upstream Mesa bug.

[acid-bong]

Huh, I tried reverting Mesa, but the error is still there. What else can I check out?

[Atemu]

Unless @vcunat has another hunch, you're going to have to bisect staging which is painful at best and, depending on your hardware, possibly not feasible.

[vcunat]

I don't. As long as you rebuilt the OS and packages for this test and rebooted.

[acid-bong]

Do I bisect staging-next between two Merge staging-next into master?

[acid-bong]

Phew, finally bisected (thank Linus and the team for git bisect start --first-parent): the culprit is Mesa, but #339741 specifically.

Should I file an issue to the Mesa devs?