Re: gstreamermm ghost pad incomplete

[José Alburquerque <jaalburquerque cox net>]

Marcus Brinkmann wrote:
> At Thu, 19 Jun 2008 00:11:37 -0400,
> José Alburquerque <jaalburquerque cox net> wrote:
>   
> > I just submitted a bug report about this to gstreamer because it looks 
> > like it's related to the way that the underlying GObject is constructed 
> > without construction properties 
> > (http://bugzilla.gnome.org/show_bug.cgi?id=539055).  I want to wait and 
> > see what gstreamer says about this before I make a decision about how to 
> > fix this warning, but the good news is that though the warning appears, 
> > this does not affect object creation.  I admit the warning is not pretty 
> > so I'll see what I can do in the mean time.
> >     
>
> Uhm, thanks, I guess, but please let me assure you that the object is
> not properly created.  Otherwise, it would have a target pad or
> direction after creation, and it doesn't.  Thus, the follow up errors.
>
>   
<snipped>
> Ok, hope this helps:
>
> C program:
> -------------------------------
> #include <gst/gst.h>
>
> int
> main (int   argc,
>       char *argv[])
> {
>   GstPad *gpad;
>   gst_init (&argc, &argv);
>   gpad = gst_ghost_pad_new_no_target ("gpad", GST_PAD_SINK);
>   printf ("direction %i\n", gst_pad_get_direction (GST_PAD (gpad)));
>   return 0;
> }
> --------------------------------
>
> $ make good CFLAGS="`pkg-config  --cflags --libs gstreamer-0.10`"
> $ ./good
> direction 2
>
> Note that the direction is set to 2, as expected.  Below, it is 0.
>
> C++ program:
>
> ----------------------------------
> #include <iostream>
> #include <gstreamermm.h>
>
> int
> main (int argc, char *argv[])
> {
>   Gst::init (argc, argv);
>   Glib::RefPtr<Gst::GhostPad> gpad
>     = Gst::GhostPad::create ("gpad", Gst::PAD_SINK);
>   std::cout << "direction " << gpad->get_direction() << std::endl;
>   return 0;
> }
> ----------------------------------
>
> $ make bad CXXFLAGS="`pkg-config  --cflags --libs gstreamermm-0.10`"
> $ ./bad
>
> (bad:11166): glibmm-WARNING **: Glib::ConstructParams::ConstructParams(): object class "gtkmm__GstGhostPad" has no property named "dir"
> direction 0
>
> (bad:11166): GStreamer-CRITICAL **: gst_pad_set_activatepull_function: assertion `GST_IS_PAD (pad)' failed
>
> (bad:11166): GStreamer-CRITICAL **: gst_pad_set_activatepush_function: assertion `GST_IS_PAD (pad)' failed
>
> (bad:11166): GLib-GObject-WARNING **: invalid (NULL) pointer instance
>
> (bad:11166): GLib-GObject-CRITICAL **: g_signal_handler_disconnect: assertion `G_TYPE_CHECK_INSTANCE (instance)' failed
>
> (bad:11166): GStreamer-CRITICAL **: gst_pad_get_peer: assertion `GST_IS_PAD (pad)' failed
> Segmentation fault
>
>   

Okay.  I think you're right on this and the errors are a reflection that 
there is something wrong when pads are created.  I traced your first c++ 
example and even the one you've just submitted and the pad does seem to 
exist after creation (you can tell this from the `direction 0' output of 
your c++ example: if the pad did not exist, the `std::cout << ... << 
pad->get_direction() << ...' could not execute), but it simply is not 
created properly.  The reason for this is that it is probably necessary 
to use the gst_ghost_pad_new_* functions directly in the Gst::GhostPad 
constructors (and probably something similar in the Gst::Pad 
constructors) and not the generic g_object_newv() 
<http://library.gnome.org/devel/gobject/2.17/gobject-The-Base-Object-Type.html#g-object-newv> 
which I think should work (I'm still picking things up about how things 
with GObject work) but doesn't seem to produce the desired results.  The 
errors are actually a result of the destruction of the pad and look like 
they're also related to how the Gst::GhostPad is created in the first 
place.  Again, thanks for pointing this out and I'll try to fix this 
shortly.

--
José Alburquerque
jaalburquerque cox net