kernel: caja trap divide error error:0 in libglib-2.0.so.0.6800.4

[L-U-T-i]

Expected behaviour

No divide error

Actual behaviour

kernel trap - caja divide error 0 in libglib

Steps to reproduce the behaviour

From the log file, I guess it happens upon the first start (login). Probably caja is restarted after, and seem to run fine when logged in...

MATE general version

1.27.2 git commit 84fa213 (from Dec. 1st, 2023), with pull-request "image properties: Handle rotated images #1657" patch applied

Package version

caja-1.27.2-0.3.git20231201.84fa213.el9.ml.x86_64 (custom build, based on a spec file from Fedora)

Linux Distribution

Rocky Linux 9.3

Link to bugreport of your Distribution (requirement)

No bug report, as it is not their package.

The following error is logged in /var/log/messages:
kernel: traps: caja[6013] trap divide error ip:7f82424949b0 sp:7ffcbd869eb0 error:0 in libglib-2.0.so.0.6800.4[7f8242472000+91000]

I wouldn't report it, as it is my custom build (what may cause the error), but googling I've found a very similar error here:
traps: caja[1878] trap divide error ip:7f6c1039bfe8 sp:7ffc005057d0 error:0 in libglib-2.0.so.0.6400.2 (deleted)[7f6c10379000+84000]
what make me guess that there is actually a part of the caja code, which in certain circumstances (doesn't happen on all machines and always, only on one machine and only on the forst start after boot...) can lead to a divide by zero condition in libglib.

Is there a way to find out which part of the caja code is "responsible" for this behavior (from ip:7f82424949b0 address, for instance)?

[superkuh]

I get this every now and then on Debian 12 MATE 1.26.0. I've never seen it in my logs (or the screen flash that occurs with it) on multiple Debian 11 MATE or Ubuntu MATE 14.04 cajas. I guess it could just be a hardware caused error on my machine and the original poster's machine.

2024-11-06T21:59:45.802304-06:00 crasti kernel: [19824.111224] traps: caja[1256] trap divide error ip:7f97bf9d1f77 sp:7ffc2d680740 error:0 in libglib-2.0.so.0.7400.6[7f97bf9ad000+8d000]

MATE general version
1.26.0

Package version
Caja 1.26.1 (Version: 1.26.1-1+deb12u1)
libglib2.0-bin (Version: 2.74.6-2+deb12u4)

Linux Distribution
Debian 12 linux

[lukefromdc]

Any chance this is controlled by window geometry-or by something assuming a zero width or height value? There is code in some places in Caja to prevent exactly that, but it only takes one place where its needed but not present to cause something like this.

To find the culprit, first make sure you have the debugging symbols for caja installed. Then, either run caja from a non-MATE session or remove it from org>mate>session>required-components>filemanager (just use an empty string)

and then from that session fire up a terminal and run
gdb caja
and from the gdb prompt type

run --force-desktop

No guarantee the bug will trigger under GDB or not under a MATE session but it it does, typing
thread apply all bt full
will bring up the backtrace. type c at the prompt when it breaks, then select all the text from the error on down and post it here as code.