Re: Gtk-CRITICAL **: gtk_widget_event: assertion `WIDGET_REALIZED_FOR_EVENT (widget, event)' failed

[David Nečas (Yeti) <yeti physics muni cz>]

On Thu, Mar 29, 2007 at 08:41:39AM +0200, Stefan Kost wrote:
> I am looking for some tips of how to get the culprit for the above  
> mentioned warning. I have a notebook with 5 pages. On the first tab I  
> have a gnome-canvas with some items on. For that I have registered an  
> event handler to move objects on the canvas. The moving part works  
> fine. When I press a key I get the assertion. Still eveything works.  
> Of course I'd like to fix it anyway. So I run the app as
>    G_DEBUG="fatal_warnings" gdb ~/buzztard/bin/bt-edit
> and dump a backtrace:
> [Switching to Thread -1225189712 (LWP 20028)]
> 0xb748509b in g_logv (log_domain=<value optimized out>,  
> log_level=G_LOG_LEVEL_CRITICAL, format=0xb74ccb67 "%s: assertion `%s'  
> failed",
>      args1=0xbf93f4bc "]\207???k?? m??\002") at gmessages.c:493
> 493                     G_BREAKPOINT ();
> (gdb) bt
> #0  0xb748509b in g_logv (log_domain=<value optimized out>,  
> log_level=G_LOG_LEVEL_CRITICAL, format=0xb74ccb67 "%s: assertion `%s'  
> failed",
>      args1=0xbf93f4bc "]\207???k?? m??\002") at gmessages.c:493
> #1  0xb7485325 in g_log (log_domain=0xb7ac6ba7 "Gtk",  
> log_level=G_LOG_LEVEL_CRITICAL, format=0xb74ccb67 "%s: assertion `%s'  
> failed") at gmessages.c:517
> #2  0xb748540b in g_return_if_fail_warning (log_domain=0xb7ac6ba7  
> "Gtk", pretty_function=0xb7b7875d "gtk_widget_event",
>      expression=0xb7b76be8 "WIDGET_REALIZED_FOR_EVENT (widget,  
> event)") at gmessages.c:532
> #3  0xb7a87769 in IA__gtk_widget_event (widget=0x833e288,  
> event=0x834b900) at gtkwidget.c:3927
> #4  0xb7a95354 in IA__gtk_window_propagate_key_event  
> (window=0x81f0808, event=0x834b900) at gtkwindow.c:4799

The key point is probably here.  Gtk+ must think the widget
has keyboard focus for some reason -- or rather Gtk+ gives
the widget focus for some reason.  So you have to find out
why.

> ...
> 
> and on it goes. None of my stuff there except below gtk_main. Now  
> unfortunately the assertion dos not tell which widget is not realized.  
> So I do:
>    (gdb) select-frame 3
>    (gdb) print
> g_type_name(((GTypeClass*)((GTypeInstance*)widget)->g_class)->g_type)
>    $1 = (gchar *) 0x8335450 "BtPatternView"
> 
> Well this is a subclassed widget that is on the 2nd tab of the  
> notbook. The "BtPatternView" subclasses a GtkTreeView, overrides  
> realize, unrealize, expose, dispose and finalize. It does the chaining  
> up.

Still, knowing normal GtkTreeView behaves the same would
decrease the number of unknonws.

> When I initially click thru all tabs (so that they get realized it  
> infact works). Explicitely calling gtk_widget_show() on it doesn't  
> help. So any idea what goes wrong here?

No ideas what goes wrong, but some ideas what to try to
debug it:
- run with G_OBJECT_DEBUG=signals and look for is-focus in
  the youtput, maybe you'll see a pattern, unfortunately it
  may not be detailed enough
- watch all changes of window->focus_widget in gtkwindow.c
  and actions of set_focus_child() of relevant containers,
  probably by manual insertions of some debugging messages,
  although one can use gdb facilities too

Yeti

--
http://gwyddion.net/