V2.32.5: deconz.service: Main process exited, code=killed, status=11/SEGV

Can, you try to swap the used protocol ? for exemple use the http/ws if you use https/wss or the reverse ?

If you want to downgrade, give me your OS I can give you the folder with all versions.

I don’t even know what protocol Home Assistant uses?

And even if switches helps - it’s still a way to remotly crash deCONZ. This should not be possible. This is a bug

There are no changes in protocols from the integration perspective, its the same as has always been.

I am seeing this too. Definitely some odd behaviour since updating to 2.32.

I can also confirm the issue with Home Assistant.
I notice it easily because I have a virtual sensor that initializes only once to a specific value when deCONZ starts up.
When I start Home Assistant, I see the “lastupdated” value update accordingly.

Home assistant : docker
DeCONZ 2.32.5 : raspberry pi 3 with the image provided by deCONZ

How many confirmations do we need until its accepted as a bug?

I just reinstalled deCONZ using the Raspberry Pi Imager tool, selecting Debian Trixie (64-bit), and then installed deCONZ by following the instructions at : ConBee III Installation

I restarted the Home Assistant Docker container several times, and deCONZ did not restart :ok_hand:.

Is there a problem with the image provided by deCONZ ? :thinking: Personally, I checked the MD5.

Was the same deconz version ?
This bug is random, need more time to be sure.

Yes with deCONZ 2.32.5.

Update: Deconz even crash if I reboot my router, so it is definetly network re-connection related.

I hope for this issue is similar to fd leak with https_port enabled (sockets to :80 in CLOSE_WAIT) · Issue #8494 · dresden-elektronik/deconz-rest-plugin · GitHub

Same issue here with 2.32.5, Windows 10. Running 32 bit version. I have both Homeseer and Home Assistant connected, I barely get an hour before it crashes.

v2.32.5-beta

Same issue on my side also when I modify something in Home Assistant related with MQTT suddenly Deconz crash. Running on raspberry 4 bulleyes 32bits

This patch will be in 2.33.0

any ETA for stable release containing the fix?

It’s out, but on beta Index of /raspbian/beta/

I updated to the beta (had to run apt --fix-broken install after that), restarted deconz. Restarted Home Assistant and it did not crash. Looks good so far :tada:

The new release 2.33.0 don’t help on my side deconz crash 2 times after I installed this new release.
The process can be reproduce when I restart HA after some minutes (~30min or less) deconz-gui on a other raspberry crash and restart.

Apr 19 11:29:41 phoscon deCONZ[30414]: malloc(): unsorted double linked list corrupted
Apr 19 11:29:41 phoscon systemd[1]: deconz-gui.service: Main process exited, code=killed, status=6/ABRT
Apr 19 11:29:41 phoscon systemd[1]: deconz-gui.service: Failed with result 'signal'.
Apr 19 11:29:41 phoscon systemd[1]: deconz-gui.service: Consumed 37.450s CPU time.
Apr 19 11:30:11 phoscon systemd[1]: deconz-gui.service: Scheduled restart job, restart counter is at 1.
Apr 19 11:30:11 phoscon systemd[1]: Stopped deCONZ: ZigBee gateway -- GUI/REST API.
Apr 19 11:30:11 phoscon systemd[1]: deconz-gui.service: Consumed 37.450s CPU time.
Apr 19 11:30:11 phoscon systemd[1]: Started deCONZ: ZigBee gateway -- GUI/REST API.
Apr 19 11:30:11 phoscon deCONZ[10250]: libpng warning: iCCP: known incorrect sRGB profile
Apr 19 11:30:11 phoscon deCONZ[10250]: libpng warning: iCCP: known incorrect sRGB profile

This is a memory corruption bug in the beta version can you fix it?

I just updated Home Assistant and deCONZ crashed again. So it’s not fixed in the beta :frowning:

Probably the same crash than here V2.32.5: deconz.service: Main process exited, code=killed, status=11/SEGV - #3 by Smanar

But I have probably missed an update on the fonction openClientTimerFired() …