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

[David Schleef]

On Sun, Feb 08, 2004 at 02:29:16PM +0100, Benjamin Otte wrote:
> > Btw, gobject signal register functions have two arguments indicating the
> > type, first a marshal macro, and later the type per argument. Can't you
> > use the combination of those two to solve it? I.e., use the first for
> > 'how to copy the argument during signalling' and the second for 'the
> > type of the argument'... Or is that a stupid remark?
> >
> I think that's killing kittens again...
> 
> Anyway, I've applied a patch that passed all caps tests and didn't have
> any issues for me.
> David (or anyone else), is that ok with you or do you think there are
> issues?

It doesn't seem necessary to deprecate gst_value_get_caps(), etc.
We should also use G_SIGNAL_TYPE_STATIC_SCOPE whenever it's useful,
which means essentially all the time with caps and structures.

Also, the boxed type registration can't take gst_caps_copy(), since
gst_caps_copy() barfs on a NULL.

I'm thinking now that my reservations about changing the signal
types were because of the old caps system.  I have not been able to
reproduce any of the problems I used to have.

Please note that there is only one test in testsuite/ that checks
correct behavior of a signal.  So relying on the testsuite as your
only means of testing is probably not wise.



dave...