Re: menuitem confusion

[Torsten Schoenfeld <kaffeetisch gmx de>]

On 30.06.2012 21:53, Dave M wrote:
> sub Gtk3::RadioMenuItem::new {
>          my ($class, $group) = @_;
>
>          return Glib::Object::new( $class, group => $group );
> }
>
> This returns
> (radiomenuitem.pl:29742): Gtk-CRITICAL **:
> gtk_radio_menu_item_get_group: assertion `GTK_IS_RADIO_MENU_ITEM
> (radio_menu_item)' failed

Note that the assertion refers to gtk_radio_menu_item_get_group.  It is 
due to the "group" property officially being of type GtkRadioMenuItem. 
So in the Glib::Object::new call above, Glib will create a GValue of 
this type containing NULL.  But under the hood, GtkRadioMenuItem 
actually accepts a list for the "group" property.  Here's the relevant code:

  if (G_VALUE_HOLDS_OBJECT (value))
    slist = gtk_radio_menu_item_get_group ((GtkRadioMenuItem*)
              g_value_get_object (value));
  else
    slist = NULL;
  gtk_radio_menu_item_set_group (radio_menu_item, slist);

So it uses G_VALUE_HOLDS_OBJECT, which is always true when Glib 
constructed the value, to decide whether the value is a single item or a 
list.  In the former case, it doesn't handle NULL/undef gracefully -- 
this is where the assertion occurs.

I think the following would work better, given the above constraints:

  sub Gtk3::RadioMenuItem::new {
    my ($class, $group) = @_;
    $group = [] unless defined $group;
    return $class->new ($group);
  }