[gst-devel] Re: [gst-cvs] wingo gstreamer: gstreamer/ gstreamer/gst/ gstreamer/testsuite/ gstreamer/testsuite/ghostpads/

[Andy Wingo]

On Thu, 05 Feb 2004, David Schleef wrote:

> On Thu, Feb 05, 2004 at 05:31:34PM -0700, Brett Kosinski wrote:
> > > > Then, once more with feeling, how the heck can bindings properly marshal
> > > > parameters in signal callbacks without resorting to special casing for
> > > > ever damned signal in the library?
> > >
> > > You don't.  And you only have to special-case some signals.

Incorrect, incorrect, incorrect!

> > Which ones?  Just the ones with GPOINTER as the type?
> Yes, probably.

Fucking insane!

> You can reasonable get away with checking all G_TYPE_POINTERs to
> see if they're pointers to GstCaps or GstStructure.

In a general piece of code, like guile-gobject? Sorry..

> We could also possibly create a special GType especially for
> marshalling GstCaps the way we want.  I'm not sure that will work
> for all types marshalled in this way, though.
> 

Document the way you want. And by the way, this is the same situation for all
language bindings. Until this is settled together, the issue will never be
settled.

> > Well, Gtk doesn't seem to have this problem, so I'm a little confused as
> > to why this issue is unique to Gstreamer.
> 
> GStreamer's internals are considerably different than Gtk+'s.  In
> particular, we have a number of structures that are not objects that
> are thrown around in signals.

What about GtkRequisition in some of the size events? That was my model when
doing this change. You're welcome by the way for value collection and lcopying.
What's the problem, excessive copying? Then or it withG_SIGNAL_TYPE_STATIC_SCOPE
or whatever that is.

This passes the testsuites. What's the problem?

BTW I won't be able to respond for another few days or so.

Regards,

Wingo.