Segfault in XrmDestroyDatabase in dunst 1.9.2

[cdown]

Didn't debug further yet since I'm going to be away for a bit, but I figure you'd like to know:

(gdb) bt
#0  ___pthread_mutex_lock (mutex=0x0) at pthread_mutex_lock.c:80
#1  0x00007f292eeb1733 in XrmDestroyDatabase (db=0x7f2918003860) at /usr/src/debug/libx11/libX11-1.8.4/src/Xrm.c:2653
#2  0x00007f292eeab8c6 in XrmSetDatabase (display=0x5560554e43c0, database=0x5560555675c0) at /usr/src/debug/libx11/libX11-1.8.4/src/Xrm.c:354
#3  0x00005560546707b1 in XRM_update_db () at src/x11/x.c:493
#4  0x0000556054670c38 in x_mainloop_fd_dispatch (source=<optimized out>, callback=<optimized out>, user_data=<optimized out>) at src/x11/x.c:354
#5  0x00007f292eac853b in g_main_dispatch (context=0x5560554c90d0) at ../glib/glib/gmain.c:3460
#6  g_main_context_dispatch (context=0x5560554c90d0) at ../glib/glib/gmain.c:4200
#7  0x00007f292eb25219 in g_main_context_iterate.constprop.0 (context=0x5560554c90d0, block=1, dispatch=1, self=<optimized out>) at ../glib/glib/gmain.c:4276
#8  0x00007f292eac7c7f in g_main_loop_run (loop=0x5560554d4820) at ../glib/glib/gmain.c:4479
#9  0x000055605466063d in dunst_main (argv=<optimized out>, argc=<optimized out>) at src/dunst.c:227
#10 main (argc=<optimized out>, argv=<optimized out>) at /usr/src/debug/dunst/dunst-1.9.2/main.c:5

So somehow ___pthread_mutex_lock is getting NULL.

This only happens immediately after starting X/wm/etc, so my guess is something is racing.

[cdown]

This is a bug report, maybe it's a mistake that this got turned into a discussion? Cc @fwsmit

[fwsmit]

I moved it, because it was not immediately clear to me something could be done with it (see reasoning in #1247). Basically, I want issues to be directly actionable.
Upon reading it a second time, it might be enough for investigation by someone interested in the problem. I'll move it back

[fwsmit]

See #1256

[bynect]

could you provide a bit more information on when this happens? Is this reproducible or just random?

[cdown]

Thanks for looking! From the original report:

This only happens immediately after starting X/wm/etc, so my guess is something is racing.

It's consistently reproducible. I looked a while ago but couldn't immediately work out what's going on.

[bynect]

I am trying to reproduce it to no avail. How do you start dunst? Could you provide a debug log (just in case)? Also, do you start X with startx or a de?

[cdown]

I worked out the problem -- it's contradictory-to-manual behaviour in XrmSetDatabase. I put up a fix in #1291.

Conditions for repro are:

1. Dunst starts before wm
2. Dunst is displaying a notification
3. WM starts and sends PropertyNotify