[gst-devel] Re: GObject porting & gtk+-1.2 shim

[Erik Walthinsen]

On 19 Jun 2001, Havoc Pennington wrote:

>  - converting GtkObject to GObject will normally break memory
>    management, unless you were using GtkObject incorrectly
>    to begin with (or planned ahead by ref/sinking all
>    GtkObject on creation)
It wouldn't surprise me if we're doing something wrong, because we came
from the gtkobject world.  We'll be working on redoing the refcounting in
the gobject-based code, and we'll do whatever shimming is necessary to
keep that functioning on gtkobject.

The script itself is only designed to deal with the vast majority of the
typing work necessary to convert.  That's what s/x/y/ is for.

>  - allowing compilation with both 1.2 and 2.0 is insanity -
>    use a CVS branch
We have no choice.  GObject is not ready, and will not be shipped to
users' machines until early 2002 at the earliest.  However, we have
requirements at work that require the ability to run with zero X
dependencies.  The *only* way to do that is to move to GObject now and
shim backwards.  Which, BTW, is working perfectly.  I can do the same
things in HEAD (GtkObject), GOBJECT1, and GOBJECT1 with the GtkObject
shim.  Once GObject is common on user machines, we'll ditch the gtk
compatibility and start making use of the newer GObject features.

And yes, I am insane, so I'm allowed to do these things anyway <g>

      Erik Walthinsen <omega at temple-baptist.com> - System Administrator
        __
       /  \                GStreamer - The only way to stream!
      |    | M E G A        ***** http://gstreamer.net/ *****
      _\  /_