Re: making g_object_connect() virtual

[Tim Janik <timj imendio com>]

On Wed, 2 Nov 2005, Robert Staudinger wrote:

> Hi,
>
> in the process of wrapping parts of the c++ mozilla API into gobjects
> [1] i came across the limitation of g_object_connect() being not
> virtual.
> Mozilla heavily uses listener classes that have to be derived from and
> handed over to the event source for receiving callbacks. In order to
> have the gobject wrapper use signals instead of callbacks it would be
> very helpful being able to override g_object_connect(). So the gobject
> wrapper would know which signals are expected by the user and can
> create the required listeners in a transparent manner.
> Or am i missing something and it's already possible to get some kind
> of notification when a signal is connected on a GObject subclass?

g_object_connect() is just a convenience wrapper around the
g_signal_connect*() familay and that doesn't necessarily operate
on only GObject types.
so adding a virtual funciton for g_signal_connect*() is not possible
and adding a virtual funciton for g_object_connect() would have little
use, since connections could always be bypassed via the g_signal*() API.
if you can make sure that this won't happen in your application, implementing
your own rober_object_connect() with a virtualized method might be
an option though.

> What is your opinion? Copious API padding seems to be there.

not sure what you mean by API padding.

> [1] http://cvs.gnome.org/viewcvs/gtkmozedit/
>
> Thanks,
> - Rob

---
ciaoTJ