(gdb) bt
#0 __pthread_create_2_1 (newthread=0x5555572d4f10, attr=0x0, start_routine=0x7ffff7e33950 <red_worker_main(void*)>, arg=0x5555572d4f10) at ./nptl/pthread_create.c:621
#1 0x00007ffff7eb11e6 in red_worker_run.isra.0 (worker=worker@entry=0x5555572d4f10) at ../server/red-worker.cpp:1127
#2 0x00007ffff7e2c4eb in red_qxl_init (qxl=0x555557194c08, reds=<optimized out>) at ../server/red-qxl.cpp:583
#3 spice_server_add_interface(SpiceServer*, SpiceBaseInstance*) (reds=<optimized out>, sin=0x555557194c08) at ../server/reds.cpp:3248
#4 0x0000555555a8c078 in qemu_spice_add_interface (sin=0x555557194c08) at ../ui/spice-core.c:874
#5 0x0000555555a8c2d1 in qemu_spice_add_display_interface (qxlin=0x555557194c08, con=0x555557133c00) at ../ui/spice-core.c:944
Now that we know that much this changelog entry in spice is suspicious
"Fix some potential thread race condition in RedClient"
But none of the commits was immediately obvious as being "the one"
Gladly I can reproduce this from spice git and spice builds fast,
so we can bisect spice to be sure what change it was without making
too many assumptions.
With debuginfo into libspice this looks like:
(gdb) bt create_ 2_1 (newthread= 0x5555572d4f10, attr=0x0, start_routine= 0x7ffff7e33950 <red_worker_ main(void* )>, arg=0x5555572d4f10) at ./nptl/ pthread_ create. c:621 run.isra. 0 (worker= worker@ entry=0x5555572 d4f10) at ../server/ red-worker. cpp:1127 4c08, reds=<optimized out>) at ../server/ red-qxl. cpp:583 add_interface( SpiceServer* , SpiceBaseInstance*) (reds=<optimized out>, sin=0x555557194c08) at ../server/ reds.cpp: 3248 add_interface (sin=0x55555719 4c08) at ../ui/spice- core.c: 874 add_display_ interface (qxlin= 0x555557194c08, con=0x555557133c00) at ../ui/spice- core.c: 944
#0 __pthread_
#1 0x00007ffff7eb11e6 in red_worker_
#2 0x00007ffff7e2c4eb in red_qxl_init (qxl=0x55555719
#3 spice_server_
#4 0x0000555555a8c078 in qemu_spice_
#5 0x0000555555a8c2d1 in qemu_spice_
Now that we know that much this changelog entry in spice is suspicious
"Fix some potential thread race condition in RedClient"
But none of the commits was immediately obvious as being "the one"
Gladly I can reproduce this from spice git and spice builds fast,
so we can bisect spice to be sure what change it was without making
too many assumptions.