Re: GtkGC shared auto-release

[Torsten Schoenfeld <kaffeetisch gmx de>]

On Mon, 2008-01-14 at 23:51 -0500, muppet wrote:

> >> Should we just add this DESTROY override to the bindings?  (Taking
> >> care to retain compatibility with code that calls ->release
> >> explicitly, of course.)
> >
> > Yes, I think that would make sense.  The only approach I can think  
> > of is
> > actually the one Kevin implemented.  Attached is an adaptation of it.
> > The in-place wrapping of Gtk2::GC methods makes it look and feel  
> > rather
> > dirty.  Is there a better way?
> 
> Not sure if it's really better, but how about a more hermetically- 
> sealed modification of the methods themselves?

Yes, that's much better.  I initially wanted to implement this in XS
too, but storing and modifying the release count inside the hash using
the XS API didn't appear pleasant.  For some reason, I didn't think of
qdata.

What about Kevin's suggestion of guarding against release_count == 0 in
the xsub for gtk_gc_release?  Otherwise, the patch looks fine to me.

-- 
Bye,
-Torsten