GNOME Bugzilla – Bug 622442
Make GObject criticals fatal
Last modified: 2013-05-31 16:56:34 UTC
This turns them into crashes, so ABRT will pick them up, we'll get more stack traces, and we'll fix them.
Created attachment 164349 [details] [review] Make GObject criticals fatal
Review of attachment 164349 [details] [review]: ::: src/gnome-shell-plugin.c @@ +315,3 @@ + * Assume the OS is shipping with a crash catching system. + */ + g_log_set_fatal_mask ("GLib-GObject", G_LOG_LEVEL_CRITICAL); Why this one domain?
To add some context I was hoping to be able to catch errors like: Window manager warning: Log level 16: invalid uninstantiatable type `(null)' in cast to `GObject' Window manager warning: Log level 8: g_object_run_dispose: assertion `G_IS_OBJECT (object)' failed Window manager warning: Log level 8: g_object_unref: assertion `G_IS_OBJECT (object)' failed Window manager warning: Log level 16: invalid uninstantiatable type `(null)' in cast to `GObject' Window manager warning: Log level 8: g_object_run_dispose: assertion `G_IS_OBJECT (object)' failed Window manager warning: Log level 8: g_object_unref: assertion `G_IS_OBJECT (object)' failed
(In reply to comment #3) > To add some context I was hoping to be able to catch errors like: > Window manager warning: Log level 16: invalid uninstantiatable type `(null)' in > cast to `GObject' > Window manager warning: Log level 8: g_object_run_dispose: assertion > `G_IS_OBJECT (object)' failed > Window manager warning: Log level 8: g_object_unref: assertion `G_IS_OBJECT > (object)' failed > Window manager warning: Log level 16: invalid uninstantiatable type `(null)' in > cast to `GObject' > Window manager warning: Log level 8: g_object_run_dispose: assertion > `G_IS_OBJECT (object)' failed > Window manager warning: Log level 8: g_object_unref: assertion `G_IS_OBJECT > (object)' failed G_DEBUG=fatal_criticals ?
Comment on attachment 164349 [details] [review] Make GObject criticals fatal Um...so I guess let's defer this pending some more GNOME-global decision on making GObject criticals fatal rather than handling them per-component.
(In reply to comment #5) > (From update of attachment 164349 [details] [review]) > Um...so I guess let's defer this pending some more GNOME-global decision on > making GObject criticals fatal rather than handling them per-component. Is that topic discussed somewhere we can track?
(In reply to comment #6) > (In reply to comment #5) > > (From update of attachment 164349 [details] [review] [details]) > > Um...so I guess let's defer this pending some more GNOME-global decision on > > making GObject criticals fatal rather than handling them per-component. > > Is that topic discussed somewhere we can track? I am not aware of anyone discussing this at the moment. Were we to try, there's lots of criticals to fix first that happen just on bootup in a standard qemu configuration... (How/why did you find this bug anyways?)
(In reply to comment #7) > (How/why did you find this bug anyways?) I was just trying to do some triaging. :)
(In reply to comment #8) > I was just trying to do some triaging. :) ... which is very much appreciated!
I think G_DEBUG=fatal-criticals is a good enough solution, we don't really need a patch for this.