Segmentation Fault on Nobara Linux

[hallnewbie]

System Information

      ---------------

OS: Nobara Linux 43 (GNOME Edition) x86_64
Host: Alienware m15 Ryzen Ed. R5 (1.13.1)
Kernel: Linux 7.0.0-200.nobara.fc43.x86_64
Terminal Font: Adwaita Mono (11pt)
CPU: AMD Ryzen 9 5900HX (16) @ 4.68 GHz
GPU 1: NVIDIA GeForce RTX 3070 Mobile / Max-Q [Discrete]
GPU 2: AMD Radeon Vega Series / Radeon Vega Mobile Series [Integrated]
Memory: 9.10 GiB / 38.52 GiB (24%)
Swap: 732.00 KiB / 50.37 GiB (0%)
Disk (/): 497.46 GiB / 908.91 GiB (55%) - btrfs
Local IP (wlp4s0): 192.168.1.21/24
Battery (0x46,0x39,0x35,0x41,0x0B,0x0A,0x06): 100% [AC Connected]
Locale: en_GB.utf8

• Operating System: OPERATING SYSTEM
• OpenRA Version: ENGINE VERSION
• OpenRA Mod: Combined Arms, Red Alert, Tiberian Dawn, Dune 2000
• Source: CA github, Official Openra site

Exception log

It is to note here that the app does not write any logs when crashing like this. the files exist, but they remain empty.

Platform is Linux (X64)
Engine version is ca-engine/1.08.2
Runtime: .NET CLR 6.0.36
launch-game.sh: line 55: 197716 Segmentation fault         (core dumped) ${RUNTIME_LAUNCHER} bin/OpenRA.dll Engine.EngineDir=".." Engine.LaunchPath="${TEMPLATE_LAUNCHER}" "Engine.ModSearchPaths=${MOD_SEARCH_PATHS}" Game.Mod="${MOD_ID}" "$@"

Additional information

All tested mods crash when executed. Both .appimage and self compiled fail in the exact same manner. Executing the programm as sudo seems to solve the issue, but somehow makes sound devices unavailable in game.

[anvilvapre]

Please consider running below steps. It may give a little bit more info on where it goes wrong.

You can also try running dotnet with prepending strace or else ltrace. All three may give some hints on in which library it may occur.

Brave AI answer:

By default, Nobara Linux does not generate traditional core files in the current working directory.  Instead, it uses systemd-coredump to manage crash dumps.

Location: Core dumps are stored in the binary journal at /var/lib/systemd/coredump/. 
Management: They are handled by systemd-coredump and can be listed and extracted using the coredumpctl command. 
Instructions to Investigate a Segmentation Fault
Trigger the Crash: Run the program that causes the segmentation fault. 
List the Crash: Use coredumpctl list to find the entry for your crashed program. 
Extract the Core Dump: Use coredumpctl dump <PID> --output /path/to/corefile to save the core dump to a specific location (e.g., coredumpctl dump 1234 --output ~/core.1234). 
Analyze with GDB: Open the core dump in the GNU Debugger with gdb <program_name> <core_file>. For example: gdb ./my_program ~/core.1234. 
Get the Backtrace: Once in the GDB prompt, type bt full to get a complete backtrace of the program's state at the time of the crash, which will help identify the source of the segmentation fault. 

[anvilvapre]

Also try running some other SDL2 program. Just to see if SDL works. Also try to run glxgears - to see if OpenGL works as expected.

Consider forcing OpenGL to use a MESA software renderer to see if it relates to your graphic card/OpenGL driver.

[hallnewbie]

running Openra inside the GNU Debugger and looking at the stack yields this result:
(gdb) bt full #0 ___pthread_mutex_lock (mutex=0x133) at pthread_mutex_lock.c:80 type = <optimized out> __PRETTY_FUNCTION__ = "___pthread_mutex_lock" id = <optimized out> #1 0x00007fff74007bf6 in wl_proxy_create_wrapper (proxy=proxy@entry=0x5555559e3ec0) at ../src/wayland-client.c:2701 wrapped_proxy = 0x5555559e3ec0 wrapper = 0x555555dfaa70 #2 0x00007fff43e7a5d7 in GetDisplayRegistry (wdpy=0x5555559e3ec0, queue=0x555555b38bc0, names=0x7fffffffbb00) at ../src/wayland/wayland-display.c:423 wrapper = 0x0 success = 0 done = <optimized out> #3 eplWlDisplayInstanceCreate (pdpy=pdpy@entry=0x555555b392b0, from_init=from_init@entry=0) at ../src/wayland/wayland-display.c:849 inst = 0x555555e280f0 names = {registry = 0x0, version_overrides = 0x0, zwp_linux_dmabuf_v1 = {name = 0, version = 0}, wp_linux_drm_syncobj_manager_v1 = {name = 0, version = 0}, wp_presentation = {name = 0, version = 0}, wp_fifo_manager_v1 = {name = 0, version = 0}, wp_commit_timing_manager_v1 = {name = 0, version = 0}, wl_drm = {name = 0, version = 0}} queue = 0x555555b38bc0 mainDevice = 0 drmNode = 0x0 drmFd = -1 serverDevice = 0x0 renderDevice = 0x0 fmt = <optimized out> supportsLinear = 0 ext = 0x0 --Type <RET> for more, q to quit, c to continue without paging--ret success = 0 done = <optimized out> __PRETTY_FUNCTION__ = "eplWlDisplayInstanceCreate" #4 0x00007fff43e7cf34 in eplWlGetPlatformDisplay (plat=0x55555590d610, pdpy=0x555555b392b0, native_display=<optimized out>, attribs=<optimized out>, existing_displays=<optimized out>) at ../src/wayland/wayland-display.c:204 env = <optimized out> inst = <optimized out> #5 0x00007fff43e76431 in eplGetPlatformDisplayExport (platformData=0x55555590d610, platform=<optimized out>, nativeDisplay=0x5555559e3ec0, attribs=<optimized out>) at ../src/base/platform-base.c:706 plat = 0x55555590d610 remainingAttribs = 0x7fffffffbce0 pdpy = 0x555555b392b0 node = <optimized out> ret = 0x0 attribCount = <optimized out> attribIndex = <optimized out> i = <optimized out> track_references = <optimized out> #6 0x00007fff2bf73bc9 in ?? () from /lib64/libEGL_nvidia.so.0 No symbol table info available. #7 0x00007fff2bf0e50f in ?? () from /lib64/libEGL_nvidia.so.0 No symbol table info available. #8 0x00007fff747bb96e in GetPlatformDisplayCommon (platform=12760, native_display=0x5555559e3ec0, attrib_list=0x0, funcName=0x7fff747c2cf5 "eglGetDisplay") at /usr/src/debug/libglvnd-1.7.0-8.fc43.x86_64/src/EGL/libegl.c:324 dpy = <optimized out> --Type <RET> for more, q to quit, c to continue without paging--c vendor = 0x555555e3c2f0 dpyInfo = 0x0 errorCode = 12288 anyVendorSuccess = <optimized out> vendorList = 0x7fff747c7430 <__eglVendorList> #9 0x00007fff744eb7d6 in SDL_EGL_LoadLibrary (_this=_this@entry=0x5555559e2f60, egl_path=egl_path@entry=0x0, native_display=0x5555559e3ec0, platform=platform@entry=0) at /__w/SDL2-CS/SDL2-CS/SDL2-2.32.0/src/video/SDL_egl.c:539 library_load_retcode = <optimized out> #10 0x00007fff74526721 in X11_GLES_LoadLibrary (_this=_this@entry=0x5555559e2f60, path=path@entry=0x0) at /__w/SDL2-CS/SDL2-CS/SDL2-2.32.0/src/video/x11/SDL_x11opengles.c:56 data = <optimized out> #11 0x00007fff745255ef in X11_GL_LoadLibrary (_this=0x5555559e2f60, path=<optimized out>) at /__w/SDL2-CS/SDL2-CS/SDL2-2.32.0/src/video/x11/SDL_x11opengl.c:270 display = <optimized out> handle = <optimized out> #12 0x00007fff744ff106 in SDL_GL_LoadLibrary_REAL (path=path@entry=0x0) at /__w/SDL2-CS/SDL2-CS/SDL2-2.32.0/src/video/SDL_video.c:3495 retval = <optimized out> #13 0x00007fff74500ada in SDL_CreateWindow_REAL (title=0x7fffffffc0c0 "", x=0, y=0, w=1, h=1, flags=10) at /__w/SDL2-CS/SDL2-CS/SDL2-2.32.0/src/video/SDL_video.c:1717 window = <optimized out> type_flags = 0 graphics_flags = 2 #14 0x00007fff7e82897e in ?? () No symbol table info available. #15 0x0000003700000036 in ?? () No symbol table info available. #16 0x000000000002b12d in ?? () No symbol table info available. #17 0x00007ffff76ce8c0 in ?? () from /home/groot/Downloads/Combined_Arms_1.08.2/Combined Arms 1.08.2/Linux/squashfs-root/usr/lib/openra/libcoreclr.so No symbol table info available. #18 0xffffffffffffffff in ?? () No symbol table info available. #19 0x00007fff7ed199c0 in ?? () No symbol table info available. #20 0x00007fff7ed199c0 in ?? () No symbol table info available. #21 0x00007fffffffc020 in ?? () No symbol table info available. #22 0x00007fff7e82897e in ?? () No symbol table info available. #23 0x00007fffffffc0b0 in ?? () No symbol table info available. #24 0x0000000000000000 in ?? () No symbol table info available.

Sadly i dont know how to read this and AI is unusualy unhelpfull. But maybe you can read whatever is wrong. Write to me if you need anything more.

[PunkPun]

I'm not sure. But it looks like it fails when either creating the window or trying to create the graphics context for the window. I see traces to SDL2, to OpenGL ES, to Wayland, to X11, to libcoreclr

Maybe drivers are setup wrong? Or the libraries on your machine have mistakes? One thing that's repeated is library failing to load

[hallnewbie]

I managed to get it to work by extracting the .appimage and using __EGL_VENDOR_LIBRARY_FILENAMES=/usr/share/glvnd/egl_vendor.d/50_mesa.json \ ./AppRun
to launch the game. While this is a bit of a hassle, it is better than using sudo to launch my video game. If that is an acceptable state of things for you people i will close the issue now.

[anvilvapre]

Brave ai answer below. The recursion is not visible. But at the bottom it mentions the likely root cause.

The segmentation fault is likely caused by a recursive mutex lock or threading issue within the Wayland EGL (egl-wayland) library.

The stack trace shows the crash occurs in ___pthread_mutex_lock while calling wl_proxy_create_wrapper. This specific pattern, involving wlExternalApiLock followed by a mutex lock failure in the EGL library, is documented in issues related to the NVIDIA egl-wayland implementation. The root cause is often a recursive attempt to acquire the same mutex when eglQueryString is called while already holding the external API lock, which leads to a deadlock or invalid lock state, ultimately causing the segfault.

This issue commonly surfaces in multi-threaded applications or toolkits (like WebKit, SDL) that interact with Wayland and EGL, especially when display or registry operations are performed across different threads without proper synchronization.

In your case, the segfault within ___pthread_mutex_lock strongly suggests the mutex being passed is invalid, most likely due to being uninitialized or having its memory corrupted or freed.

The segfault is caused by an invalid mutex address. The parameter mutex=0x133 in ___pthread_mutex_lock is not a valid pointer to a pthread_mutex_t structure. Instead, it's an extremely low, invalid memory address (307 in decimal), indicating that the pointer was either:

Never initialized, or
Overwritten due to a buffer overflow or memory corruption elsewhere in the program.

[anvilvapre]

For future reference. Ways to force mesa software GL. Any of these may be sufficient.

env __GLX_VENDOR_LIBRARY_NAME=mesa __EGL_VENDOR_LIBRARY_FILENAMES=/usr/share/glvnd/egl_vendor.d/50_mesa.json LIBGL_ALWAYS_SOFTWARE=1 your_application   

[hallnewbie]

For future reference. Ways to force mesa software GL. Any of these may be sufficient.

env __GLX_VENDOR_LIBRARY_NAME=mesa __EGL_VENDOR_LIBRARY_FILENAMES=/usr/share/glvnd/egl_vendor.d/50_mesa.json LIBGL_ALWAYS_SOFTWARE=1 your_application   

Still fries the .appimage.

I managed to find the issue after some testing. It turns out that the Extracted Appimage that i managed to get working was from a previous try where i replaced the bundled SDL2 with my systems SDL 2.

It seems two things are happening here, one is an issue with the bundled version of SDL2 and the other is mesa.