Crash under Wayland on Fedora 40 [libnvidia-glcore.so.570.86.16+0xa58f7c]

granitrocky

Thank you for that. I tried the old launcher and the bug is still there. I am using the latest Nvidia drivers for Fedora 40 (570.86.16), but this has been happening across versions. You can actually google around and find a bunch of Minecraft threads that reference the same issue.

If there's a significant distance between the java library versions, then I'll just have to swap to X11 when I want to use this app natively.

It runs just fine under Wine and Wayland, which is frustrating haha.

Nate

Re: Wine, that is annoying! 😆

I did a quick search and found lots of similar LWJGL (2 and 3) crashes, but very few crash in nDestroyWindow like we're seeing here.

LWJGL2 nDestroyWindow code is here:
LWJGL/lwjglblob/2df01dd762e20ca0871edb75daf670ccacc89b60/src/native/linux/opengles/org_lwjgl_opengl_Display.c#L542

destroyWindow is here:
LWJGL/lwjglblob/2df01dd762e20ca0871edb75daf670ccacc89b60/src/native/linux/opengles/org_lwjgl_opengl_Display.c#L227

static void destroyWindow(JNIEnv *env, Display *disp, Window window) {
	XDestroyWindow(disp, window);
	XFreeColormap(disp, cmap);
}

The crash is in one of those two calls, and they are both in X11 here:
mirror/libX11blob/master/include/X11/Xlib.h

Maybe Spine or LWJGL2 is doing something that causes those X11 functions to explode, but it seems more likely it's an Nvidia driver issue.

granitrocky

Hm... Yeah it seems like Wayland shouldn't be calling X11 functions. Maybe under Xwayland it's doing it. So maybe it's not an Nvidia driver issue but a Wayland bug that is passing a null pointer to destroywindow

granitrocky

UPDATE: I just tried running on Fedora 39 on my laptop with Wayland and there was no crash. Then I remembered that I have a passthrough GPU that is disabled on my desktop and it's in slot 0, so I bet the NVIDIA/Wayland driver are trying to use that one.

Long story short, this seems like a very personal bug. If someone comes across this thread in the future from googling the "nv_glcore" stuff, I will try to see what I can find to help them solve this.

Nate

Ah, interesting. Thanks for following up!

Nate

Can you please try setting __GL_THREADED_OPTIMIZATIONS=0?

export __GL_THREADED_OPTIMIZATIONS=0
./Spine.sh

granitrocky

Nate Well that's progress at least! It starts without crashing, but now throws a different error about Wayland. I'll get more details soon

granitrocky

Looks like it all is working perfectly with that environment variable. Was glitchy a bit at first, but on relaunching it works great. Thanks!

Nate

Great! We can put it in Spine.sh to reduce the pain.

thomaswashere

I've also ran into this issue on my desktop machine:

# A fatal error has occurred:
#
# SIGBUS (0x7) at pc=0x000072f3c665949c, pid=132985, tid=133021
# Problematic frame:
# C  [libnvidia-glcore.so.570.133.07+0xa5949c]

I can provide the log file if needed. I am using Sway (Wayland) on Arch Linux with a NVIDIA GeForce RTX 3080 Ti/PCIe/SSE2, 4.6.0 NVIDIA 570.133.07. I will also try the environment variable and report back.

Works fine with an Intel integrated graphics on my laptop.

Davide

thomaswashere

Works fine with Intel integrated graphics on my laptop.

Did you also use Sway (Wayland) on Arch Linux on your laptop?

I will also try the environment variable and report back.

We personally experienced the exact same error you linked above and were able to resolve it by setting that environment variable. Hope it works for you as well!

thomaswashere

Davide

Did you also use Sway (Wayland) on Arch Linux on your laptop?

Yes, an almost identical configuration (using Mesa drivers instead).

Setting the environment variable solved the issue! Happy to continue using Spine!

Davide

thomaswashere

Thanks for reporting back 🙂
Glad it solved your problem!