Red Hat Bugzilla – Bug 1293088
[abrt] evince: g_type_check_instance(): evince killed by SIGSEGV
Last modified: 2016-11-08 16:59:48 EST
Description of problem:
If you let evince in full screen mode, it'll crash in the next launch (only the first launch, then it'll work ok)
Version-Release number of selected component:
cmdline: evince /home/neil/Documentos/Text/Books/Happiness-Beyond-Thought-A-Practical-Guide-to-Awakening.pdf
runlevel: N 3
Thread no. 1 (10 frames)
#0 g_type_check_instance at gtype.c:4138
#1 g_signal_handler_disconnect at gsignal.c:2620
#2 ev_page_action_widget_set_document at ev-page-action-widget.c:265
#8 g_object_notify_by_spec_internal at gobject.c:1154
#9 g_object_notify at gobject.c:1202
#10 ev_document_model_set_document at ev-document-model.c:381
#11 ev_window_load_job_cb at ev-window.c:1727
#12 _g_closure_invoke_va at gclosure.c:864
#15 emit_finished at ev-jobs.c:181
#19 g_main_context_iteration at gmain.c:3901
Potential duplicate: bug 1191772
Created attachment 1107854 [details]
Created attachment 1107855 [details]
Created attachment 1107856 [details]
Created attachment 1107857 [details]
Created attachment 1107858 [details]
Created attachment 1107859 [details]
Created attachment 1107860 [details]
Created attachment 1107861 [details]
Created attachment 1107862 [details]
Created attachment 1107863 [details]
Created attachment 1107864 [details]
Created attachment 1107865 [details]
Created attachment 1107866 [details]
This package has changed ownership in the Fedora Package Database. Reassigning to the new owner of this component.
*** Bug 1280106 has been marked as a duplicate of this bug. ***
cannot reproduce fullscreen crash in current F24.
action_widget->doc_model is apparently a weak pointer
if it were 0 that would explain it but gdb shows it as non-0.
the glib docs say that this weak reference isn't thread safe
so it's theoretically possible that another thread
deletes the object concurrently but i don't know enough
about evince to say if that is possible.
upstream bug with very similar stack:
... which i can't reproduce either.
g_object_notify (object=0x5616082b60c0, ...
but then ev_page_action_widget_set_document calls
g_signal_handler_disconnect (instance=0x5616086045b0, ...
is it expected that there are 2 different instances of
the only place where EvDocumentModel is created is
ev_window->priv->model = ev_document_model_new ();
the signal that is being broadcast here has a handler registered
from ev_page_action_widget_set_model, but that handler is
so it could be possible that ev_page_action_widget_set_model
is called twice, and the second call sets a model that is then
destroyed before the first one emits this "notify::document" signal.
but i think i'm not able to fix this if i can't reproduce it,
so i'll just close it upstream since there is already an
upstream bug about it, maybe somebody will fix it...