Re: GtkOrientable vs property "override"

[Torsten Schoenfeld <kaffeetisch gmx de>]

Kevin Ryde wrote:
> Is the GtkOrientable bit about making a property "override" still
> pending?

Yes, I committed Emmanuel's patch without the _ADD_INTERFACE xsub.  So any 
attempt to implement Gtk2::Orientable in a subclass will fail[1].  The bug 
entries are <http://bugzilla.gnome.org/show_bug.cgi?id=570792> and 
<http://bugzilla.gnome.org/show_bug.cgi?id=570793>.

> Coming to understand it very very slowly ... is it about making
> g_object_class_override_property() available so a perl-code widget can
> create itself a property which takes the pspec of the GtkOrientable?

Yes.  The basic open question is when do we want 
g_object_class_override_property() to happen.  Do we want it to happen 
automatically whenever a subclass is registered that implements Gtk2::Orientable 
but which doesn't have its own "orientation" property?  That is, do we want

  use Glib::Object::Subclass
    Parent::,
    interfaces => [ Gtk2::Orientable:: ];

to call g_object_class_override_property() for "orientation"?

Or do we want to make the programmer actively ask for the property override, 
using a new syntax?

  use Glib::Object::Subclass
    Parent::,
    properties => [ "orientation" ], # override for Gtk2::Orientable
    interfaces => [ Gtk2::Orientable:: ];

The first approach should be easy to implement and doesn't need new syntax, but 
it is rather automatic.  It would mean that [GS]ET_PROPERTY were called for 
properties ("orientation" in the example) that the programmer didn't mention at all.

[1] With an obscure error message about _ADD_INTERFACE not being found.  We 
should try to provide better diagnostics.