It seems that Matthieu found the cause of the issue: ----- Forwarded message from Walter Alejandro Iglesias <w...@roquesor.com> -----
Date: Tue, 30 Aug 2022 20:43:07 +0200 From: Walter Alejandro Iglesias <w...@roquesor.com> To: Matthieu Herrb <matth...@herrb.eu> Subject: Seems you got it! On Aug 30 2022, Matthieu Herrb wrote: > On Tue, Aug 30, 2022 at 06:47:55PM +0200, Matthieu Herrb wrote: > > On Tue, Aug 30, 2022 at 06:22:21PM +0200, Walter Alejandro Iglesias wrote: > > > On Aug 30 2022, Matthieu Herrb wrote: > > > > Another path is that you try to get an X server stack trace. Since > > > > you're using startx it's easier than what is documented in > > > > /usr/xenocara/README. Just create a /etc/X11/xorg.conf file containing > > > > > > > > Section "ServerFlags" > > > > Option "NoTrapSignals" "true" > > > > EndSection > > > > > > > > and also rebuild the X server once more with : > > > > > > > > doas make -f Makefile.bsd-wrapper build CFLAGS="-O0 -g" > > > > > > > > Once it has crashed, you should have a Xorg.core file in your home > > > > directory. To get a backtrace install egdb from packages (pkg_add gdb) > > > > and run it > > > > > > > > egdb /usr/xobj/xserver/hw/xfree86/Xorg Xorg.core > > > > > > > > inside egdb run the 'bt' command. > > > > > > The xmovepointer didn't triger the issue. Perhaps it's not the pointer > > > but the keyboard. After recompiling the xserver as you told me I > > > reproduced the bug as I've been doing. This is the output of egdb: > > > > Yes the trace seems to point to a keyboard issue. I'm going to look > > closer. Thanks a lot this is really useful. > > > > Hmm it looks like a bug that was already fixed upstreams. Can you try > the patch below that applies all the remaining parts of 21.1.4 release > to the xenocara tree (the security related ChangeLog items are already > there). > > Apply by running: > > patch -p0 -E < /path/to/this/file > > and rebuild the X server like you already did several times. I applied that patch to both machines and recompiled xserver. So far I tried a hundred times in both machines and I couldn't reproduce the bug! -- Matthieu Herrb Walter