Re: GtkCellRenderer.t failure.

[Ross McFarland <rwmcfa1 neces com>]

On Sun, 2004-02-01 at 14:00, muppet wrote:
> On Sunday, February 1, 2004, at 11:14 AM, Ross McFarland wrote:
>
> which versions of everything are you using?  i've seen this on some 
> combinations, usually if you don't actually add any signals or 
> properties to the derived type (add_signals in GType.xs calls ref on 
> the type).

stock rh9, gtk2-perl HEAD

> i don't think the change to ref actually hurts anything; the leaked 
> class ref is not a problem in practice, since perl does not let classes 
> go away once they've been defined.  it may be problematic for embedded 
> situations, but we'll burn that bridge when we're standing on it.

what about something like:

/* peek should suffice, as the bindings should keep this class
 * alive for us. */
class = g_type_class_peek (gtype);
if (! class)
{
        class = g_type_class_ref (gtype);
        if (! class)
                croak ("internal problem: can't peek at type class "
                       "for %s(%d)", g_type_name (gtype), gtype);
}
gtk2perl_cell_renderer_class_init (class);

would seem to try to catch and handle the case i'm seeing, without added
an extra unwanted ref to the class.

-rm