ONLYOFFICE Desktop Editors (.deb) crashes on launch after MESA UPDATE to Mesa 25.2.8

PB19 · 27 June 2026 15:30

Environment

OS: Linux Mint 22 (Ubuntu 24.04 LTS base)
Desktop Environment: Cinnamon (X11)
Hardware: Lenovo ThinkPad X13 Gen 2
Application Version: ONLYOFFICE Desktop Editors 9.4.0-129 (installed via official upstream .deb repository)
Graphics Stack: Mesa 25.2.8-0ubuntu0.24.04.2 (AMD64 / I386)

Context & Chronology

May 22, 2026: ONLYOFFICE Desktop Editors was upgraded from version 9.3.1-8 to 9.4.0-129. The application performed as expected and launched successfully after this update.
June 2026: The system received the Mesa graphics library update (25.2.8-0ubuntu0.24.04.2). Immediately following this update, ONLYOFFICE Desktop Editors began crashing systematically on startup.

Steps to Reproduce

Install ONLYOFFICE Desktop Editors (9.4.0-129) via the official .deb package on a system running Mesa 25.2.8-0ubuntu0.24.04.2.
Launch the application via the terminal: onlyoffice-desktopeditors

Observed Behavior

The application fails to initialize its graphical interface and terminates immediately with a core dump during the XCB/Qt window initialization phase.

Terminal output:

(DesktopEditors:37629): Gtk-WARNING **: 14:51:30.351: gtk_disable_setlocale() must be called before gtk_init()
QXcbConnection: XCB error: 8 (BadMatch), sequence: 670, resource id: 79691839, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 686, resource id: 79691839, major code: 130 (Unknown), minor code: 3
Abandon (core dumped)

Verified Test Cases (Isolation)

Software Rendering: Forcing software rendering via LIBGL_ALWAYS_SOFTWARE=1 or GALLIUM_DRIVER=softpipe does not bypass the crash. The identical XCB BadMatch error and core dump occur.
Environment / Configuration: Purging local configuration directories (~/.config/onlyoffice and ~/.local/share/onlyoffice) has no effect.
Security Modules: Running with QT_IGNORE_PKCS11=1 does not prevent the crash.
Alternative Formats: The Flatpak distribution of ONLYOFFICE Desktop Editors runs successfully on the exact same host system configuration

Nurullo · 29 June 2026 14:23

Hi @PB19, welcome to the community!

Thanks for the detailed report!

Based on the timeline you described, the Mesa 25.2.8 update in June looks like the likely culprit rather than the ONLYOFFICE upgrade from May.

Could you try launching the app with this:

LIBGL_DRI3_DISABLE=1 onlyoffice-desktopeditors

If that doesn’t help, also try:

QT_XCB_GL_INTEGRATION=none onlyoffice-desktopeditors

In the meantime, the Flatpak version from Flathub works fine as a workaround — here’s how to set it up if you haven’t already: Installing ONLYOFFICE Desktop Editors on Linux on your computer using Flatpak

PB19 · 1 July 2026 16:36

Hi,

Thanks for the quick follow-up!

I just tested both commands with the .deb package reinstalled, and unfortunately, neither workaround resolved the issue. The behavior is identical to the initial crash.

Here are the exact terminal outputs for both tests:

Plaintext

$ LIBGL_DRI3_DISABLE=1 onlyoffice-desktopeditors

(DesktopEditors:7069): Gtk-WARNING **: 18:32:37.392: gtk_disable_setlocale() must be called before gtk_init()
QXcbConnection: XCB error: 8 (BadMatch), sequence: 671, resource id: 65011775, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 687, resource id: 65011775, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 752, resource id: 65011784, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 799, resource id: 65011784, major code: 130 (Unknown), minor code: 3
Aborted (core dumped)

Plaintext

$ QT_XCB_GL_INTEGRATION=none onlyoffice-desktopeditors

(DesktopEditors:7224): Gtk-WARNING **: 18:32:50.532: gtk_disable_setlocale() must be called before gtk_init()
QXcbConnection: XCB error: 8 (BadMatch), sequence: 652, resource id: 65011774, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 668, resource id: 65011774, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 733, resource id: 65011783, major code: 130 (Unknown), minor code: 3
QXcbConnection: XCB error: 8 (BadMatch), sequence: 780, resource id: 65011783, major code: 130 (Unknown), minor code: 3
Aborted (core dumped)

I will stick to the Flatpak version as a temporary workaround for now, but let me know if you need any other logs or tests to help fix the upstream .deb package.

Nurullo · 6 July 2026 10:27

Hi @PB19,

Thanks for testing both and sharing the outputs!
We’re looking into this and will get back to you.

Nurullo · 27 July 2026 12:24

Hi @PB19,

Thanks for your patience! To help us narrow down the root cause, could you share the outputs of the following commands:

lspci -k | grep -A5 -E 'VGA|3D|Display'
glxinfo -B
dpkg -l | grep mesa
echo "XDG_SESSION_TYPE=$XDG_SESSION_TYPE"
cat /etc/linuxmint/info
inxi -Fxxxrz 2>/dev/null | head -30

And this one is the most important — a full crash backtrace:

gdb -batch -ex "run" -ex "bt full" --args onlyoffice-desktopeditors 2>&1 | tail -100

You may need to install gdb first: sudo apt install gdb

One more workaround worth trying:

QT_X11_NO_MITSHM=1 onlyoffice-desktopeditors

And a question — would you be able to update Mesa to the latest available version and check if the crash still happens?

Thank you!

PB19 · 28 July 2026 20:43

Hello,

Thanks for getting back to me!

Before gathering all the graphics and Mesa logs, I found the remaining cause of the crash (with some troubleshooting help from Claude). It turns out the issue was entirely unrelated to Mesa or the GPU (the QT_X11_NO_MITSHM=1 workaround didn’t change anything). Maybe the Mesa has been updated.

The remaining crash was caused by CEF/NSS attempting to load a specific PKCS#11 module on startup: the official Belgium eID middleware (libbeidpkcs11.so.0) registered in my user NSS database (~/.pki/nssdb).

Here is the exact backtrace from the crash, showing the unhandled exception during C_Initialize():

Plaintext

Thread 11 "ThreadPoolForeg" received signal SIGABRT, Aborted.
[Switching to Thread 0x7fffc1ffc6c0 (LWP 101412)]
#0  __pthread_kill_implementation (no_tid=0, signo=6, threadid=<optimized out>) at ./nptl/pthread_kill.c:44
#1  __pthread_kill_internal (signo=6, threadid=<optimized out>) at ./nptl/pthread_kill.c:78
#2  __GI___pthread_kill (threadid=<optimized out>, signo=signo@entry=6) at ./nptl/pthread_kill.c:89
#3  0x00007ffff3a4527e in __GI_raise (sig=sig@entry=6) at ../sysdeps/posix/raise.c:26
#4  0x00007ffff3a288ff in __GI_abort () at ./stdlib/abort.c:79
#5  0x00007ffff38137c2 in ?? () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#6  0x00007ffff383617d in __gcc_personality_v0 () from /lib/x86_64-linux-gnu/libgcc_s.so.1
#7  0x00007ffff5ebd993 in ?? () from /opt/onlyoffice/desktopeditors/converter/libicuuc.so.74
#8  0x00007ffff5ebe04a in ?? () from /opt/onlyoffice/desktopeditors/converter/libicuuc.so.74
#9  0x00007ffff5e3762c in __cxa_throw () from /opt/onlyoffice/desktopeditors/converter/libicuuc.so.74
#10 0x00007fffa8020d81 in ?? () from /usr/lib/x86_64-linux-gnu/libbeidpkcs11.so.0
#11 0x00007fffa803e69b in ?? () from /usr/lib/x86_64-linux-gnu/libbeidpkcs11.so.0
#12 0x00007fffa8044d8c in ?? () from /usr/lib/x86_64-linux-gnu/libbeidpkcs11.so.0
#13 0x00007fffa802664a in C_Initialize () from /usr/lib/x86_64-linux-gnu/libbeidpkcs11.so.0
#14 0x00007fffe0321e28 in ?? () from /lib/x86_64-linux-gnu/libnss3.so
#15 0x00007fffe0322dab in ?? () from /lib/x86_64-linux-gnu/libnss3.so
#16 0x00007fffe0329c05 in SECMOD_LoadModule () from /lib/x86_64-linux-gnu/libnss3.so
#17 0x00007fffe0329ff8 in SECMOD_LoadModule () from /lib/x86_64-linux-gnu/libnss3.so
#18 0x00007fffe02ef118 in ?? () from /lib/x86_64-linux-gnu/libnss3.so
#19 0x00007fffe02ef447 in ?? () from /lib/x86_64-linux-gnu/libnss3.so
#20 0x00007fffe02f36cd in NSS_InitReadWrite () from /lib/x86_64-linux-gnu/libnss3.so
#21 0x00007fffed196cde in ?? () from /opt/onlyoffice/desktopeditors/libcef.so

The Fix: I ran modutil -delete "Belgium eID" -dbdir sql:$HOME/.pki/nssdb, and OnlyOffice immediately started working without any crashes.

Surprisingly, after re-adding the exact same module manually using modutil -add, OnlyOffice continues to work perfectly. (Note: I haven’t re-tested actual eID card authentication in the browser yet, but the module is properly registered and loaded in NSS).

Since the application is running flawlessly now and the root cause is isolated to the NSS PKCS#11 module loading, I assume you won’t need the lspci, glxinfo, or inxi outputs anymore.

Here is the precise sequence:

Fresh profile / First launch: If libbeidpkcs11.so.0 is registered in NSS before OnlyOffice runs for the first time, CEF attempts to call C_Initialize() during its initial setup, triggering the unhandled exception and SIGABRT.
The bypass: Removing the module (modutil -delete) allows OnlyOffice to complete its first-run setup cleanly and structure its profile.
Subsequent launches: Re-registering the module (modutil -add) after that initial run causes no crashes at all. OnlyOffice opens normally, even with the eID module fully loaded in NSS.

Nurullo · 29 July 2026 13:24

Thank you for the detailed investigation and the backtrace — really appreciate you tracking this down!

We’ll pass this along on our side for a closer look. Thanks again for the thorough troubleshooting!

PB19 · 29 July 2026 15:34

Thank you. It is really possible that there has been a mesa update between the differents posts, that solved the first issue. The second is already know and quite for a long tile without real fix, from what I read.

Nurullo · 12 August 2026 12:03

Hi @PB19, just a quick update — we were able to reproduce the issue on our side and it’s been logged as a bug. Thanks again for the thorough investigation, it really helped us get to the bottom of this!