You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hmmm, odd, Linux is a touchy one for us. I know Ubuntu 25.10 runs Vector perfectly as that's the machine I test for mass Linux support on, I'll keep this in mind, but I only run Arch and Ubuntu on my machines. 🙏
Thank you for answering!
So Cachy is Arch based, so if you happen to find the time to test replicating the same conditions, should be useful too (eg, Plasma, X11). That being said, the problem is that I have no idea what fails, so not sure what to try.
What I see in the terminal: identical to this issue:
[AudioEngine] Initialized successfully
Could not create default EGL display: EGL_BAD_PARAMETER. Aborting...
[WARN ...] Frontend never revealed the window in 6s — showing it anyway
The window shows up and the main process keeps running until killed — but the app content never renders.
Additionally, on every launch the WebKitWebProcess child aborts (SIGABRT, si_code: SI_TKILL — a self-inflicted abort(), not a segfault). Coredumps both times, identical signature. From coredumpctl + gdb:
The caller of abort() is inside the AppImage's bundledlibwebkit2gtk-4.1.so.0 (offset 0x28dc4d4 from the library's load base). Bundled libs are stripped with no debug symbols, so the exact function can't be resolved — but the depth into the library puts it in WebKit's own code, most likely a failed assert()/g_assert during web-process startup.
All other threads were idle in poll/IPC at crash time — the abort happens during web-process initialization, nothing in flight.
Ruled out: OOM (41 GiB free, zero OOM events), GPU/driver (no amdgpu kernel messages in the window), recent package updates.
Host GL stack is fine on this system: Mesa 26.2.1 with radeonsi_dri.so present, host libEGL loads cleanly, no EGL_*/LIBGL_* env overrides. So the EGL failure is in Vector's bundled stack, not the system's.
So the failure extends past X11/KWin to Wayland, and past the main process's EGL error into the web process itself. Could be the same root cause (the bundled Ubuntu-based GTK/WebKit stack not coping with a newer Arch-based system) or two bugs — the crash is too deep to tell without symbols.
Happy to upload the two coredumps if that helps.
Filed by Qwen3.8-27B-Cyankiwi-AWQ-INT4 via Hermes Agent (Nous Research).
Activity
JSKitty commentedon Jul 19, 2026
Hmmm, odd, Linux is a touchy one for us. I know Ubuntu 25.10 runs Vector perfectly as that's the machine I test for mass Linux support on, I'll keep this in mind, but I only run Arch and Ubuntu on my machines. 🙏
Novack commentedon Jul 19, 2026
Thank you for answering!
So Cachy is Arch based, so if you happen to find the time to test replicating the same conditions, should be useful too (eg, Plasma, X11). That being said, the problem is that I have no idea what fails, so not sure what to try.
Novack commentedon Aug 29, 2026
Note that up to this day, I get exactly the same on 4.3.
BonesGit commentedon Sep 1, 2026
Same
EGL_BAD_PARAMETERline plus a related second failure — adding for comparison since this thread is thin on data.Setup: Arch-based (Omarchy) + Hyprland (Wayland, no X11), AMD Radeon 880M APU, Mesa 26.2.1,
Vector_0.4.4_amd64.AppImage.What I see in the terminal: identical to this issue:
The window shows up and the main process keeps running until killed — but the app content never renders.
Additionally, on every launch the
WebKitWebProcesschild aborts (SIGABRT,si_code: SI_TKILL— a self-inflictedabort(), not a segfault). Coredumps both times, identical signature. Fromcoredumpctl+ gdb:abort()is inside the AppImage's bundledlibwebkit2gtk-4.1.so.0(offset0x28dc4d4from the library's load base). Bundled libs are stripped with no debug symbols, so the exact function can't be resolved — but the depth into the library puts it in WebKit's own code, most likely a failedassert()/g_assertduring web-process startup.amdgpukernel messages in the window), recent package updates.Host GL stack is fine on this system: Mesa 26.2.1 with
radeonsi_dri.sopresent, hostlibEGLloads cleanly, noEGL_*/LIBGL_*env overrides. So the EGL failure is in Vector's bundled stack, not the system's.So the failure extends past X11/KWin to Wayland, and past the main process's EGL error into the web process itself. Could be the same root cause (the bundled Ubuntu-based GTK/WebKit stack not coping with a newer Arch-based system) or two bugs — the crash is too deep to tell without symbols.
Happy to upload the two coredumps if that helps.
Filed by Qwen3.8-27B-Cyankiwi-AWQ-INT4 via Hermes Agent (Nous Research).