Re: Segfault with a combo with 3 columns

[muppet <scott asofyet org>]

On Mar 26, 2007, at 8:59 PM, Daniel Kasak wrote:

> I've attached a short script which produces a crash ( segfault )  
> for me when an item in a combo is selected.
>
> The demo has 2 combos / models.
>
> The 1st, 'good' model, has a Glib::Int column, and 2 Glib::String  
> columns. This works fine.
>
> The 2nd, 'bad' model, has a 2 Glib::Int columns, and a Glib::String  
> column.
>
> When an item is selected in the 2nd combo, I get a segfault. This  
> happens if the user selects an item with the mouse, or if the item  
> is selected in code, ie:
> $widget->set_active_iter( $iter );
>
> Note that I'm ONLY loading numeric data into the 2nd column.
>
> My hunch is that this is a gtk+ bug, and not related to the  
> bindings. Am I right?

Not so sure.

On one hand, you're telling the ComboBoxEntry to get the text to show  
in the entry from column 1, which is a Glib::Int column, and it does  
so, blindly trusting you, and crashing.  (The crash is in strcmp(),  
deep within gtk_entry_set_text() from  
gtk_combo_box_entry_active_changed().)  It's a fairly trivial patch  
to your code to use a real Glib::String column for the text column.

 sub create_combo {

     my $this_model = shift;

-    my $combo = Gtk2::ComboBoxEntry->new( $this_model, 1 );
+    my $combo = Gtk2::ComboBoxEntry->new( $this_model, 2 );

-    # Create a renderer for the 3rd column ( the text column )
+    # Create a renderer for the 2nd column ( the number column )
     my $renderer = Gtk2::CellRendererText->new;
     $combo->pack_start( $renderer, FALSE );
-    $combo->set_attributes( $renderer, text => 2 );
+    $combo->set_attributes( $renderer, text => 1 );

     return $combo;
 }


On the other hand, in perl we're used to type coercion, and when  
dealing with GtkTreeViewColumns and GtkCellRendererTexts you get type  
coercion from G_TYPE_INT to G_TYPE_STRING.  I've opened a gtk+ bug  
about it, if you'd like to follow it: http://bugzilla.gnome.org/ 
show_bug.cgi?id=423201


--
If the monkey could type one keystroke every nanosecond, the expected  
waiting time until the monkey types out Hamlet is so long that the  
estimated age of the universe is insignificant by comparison ... this  
is not a practical method for writing plays.
  -- Gian-Carlo Rota