[Gtk2-perl-bugs] [Bug 570792] Support interface properties when subclassing

[gnome-perl (bugzilla.gnome.o. <bug...@gn...>]

https://bugzilla.gnome.org/show_bug.cgi?id=570792
  gnome-perl | Glib | unspecified

--- Comment #1 from Kevin Ryde <us...@zi...> 2010-11-08 19:45:42 UTC ---
Created an attachment (id=174092)
 View: https://bugzilla.gnome.org/attachment.cgi?id=174092
 Review: https://bugzilla.gnome.org/review?bug=570792&attachment=174092

patch installing properties, but they don't reach GET_PROPERTY/SET_PROPERTY

I think saying "interfaces => [ ... ]" could automatically add the interface
properties.  If you're implementing an interface then that'll be usually what
you want.  Worry about subtle variations later.

I wondered if it was as easy as the gobject manual suggested and you just had
to g_object_class_override_property() each of the
g_object_interface_list_properties().  I even threw down the attached few lines
to that end.  However,

The pspec which reaches gperl_type_get_property() is the underlying one out of
the interface, and its pspec->owner_type is the interface, not the class
implementing the interface.  This means gperl_type_get_property() doesn't go to
the class' GET_PROPERTY.

The way g_object_get_property() looks through an override to the underlying can
be seen in object_get_property() of gobject.c.  I suppose someone thought that
was a good idea.

I suspect this is another case of the C code being oriented towards individual
class funcs such as class->get_property implicitly knowing which class they're
for, rather than a generic like gperl_type_get_property().  Perhaps the
property_id number could encode something to identify the actual target class. 
Normally those ids are consecutive 1 to N, but I get the impression it's an
arbitrary integer for use by the class in question.

Second, I see g_object_class_override_property() looks first in parent classes
and then in interfaces for a property of the given name.  If a subclass has an
interface with an interface property with the same name as a superclass
property then g_object_class_override_property() might end up taking the
superclass one as the target, not the interface.

I suppose that affects C code as well.  Unless the gobject manual has something
better to say I suspect an explicit

    pspec = g_param_spec_override (name, iface_pspec);
    g_object_class_install_property (oclass, property_id, pspec);

might be necessary to ensure it's the interface one which is installed.

This sort of thing might come up if a subclass re-implements an interface
already implemented in a superclass.  That's fairly common of course if a
subclass wants to do things slightly differently.  Eg. I've made various
subclasses redoing TreeDragSource to have different answers about draggability.
 Whether there's any like that with interface properties would be another
matter.

-- 
Configure bugmail: https://bugzilla.gnome.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.