Re: Fake inheritance in GDK

[murrayc t-online de (Murray Cumming)]

On Thu, 2001-10-11 at 17:05, Owen Taylor wrote:
> 
> murrayc t-online de (Murray Cumming) writes:
> 
> > I'm auto-generating C++ wrappers for GDK 2, and I've discovered some
> > small difficulties:
> > 
> > GdkVisualClass is declared as this, in gdkvisual.h:
> > 
> > typedef struct _GdkVisualClass    GdkVisualClass;
> > 
> > But _GdkVisualClass is not defined anywhere. I think that it should, at
> > least, be defined as 
> > typedef struct _GdkDrawableClass    GdkVisualClass
> > as has been done for GdkPixmap and GdkBitmap in gdktypes.h
> 
> GdkVisual is not a GdkDrawable:

Sorry, I mean it should be

typedef struct _GObjectClass    GdkVisualClass;
 
> 
>  http://developer.gnome.org/doc/API/2.0/gdk/gdk-visuals.html
>  
> explains what a GdkVisual is.
> 
> > Actually, I'd like to change all of these pretend derivations to real
> > derivations. The C++ compiler can't tell the difference between
> > GdkDrawable*, GdkBitmap*, GdkPixmap*, and GdkVisual* because they are
> > all simple typedefs of each other. Would you accept a patch to do that?
> 
> No. Breaking every GTK+ program out there not allowed. :-)

I didn't think the there would be any source breakage, and possibly not
even any binary breakage, because

struct GdkBitmapClass
{
  GObjectClass
}

might look just like

typdef struct _GdkDrawableClass GdkBitmapClass

Of course I was only guessing.

> 
> (They actually are real derivations, in the GObject sense, we just
> use the same typedef for all of the objects.)

I can see that there isn't much advantage for C coders, but surely it's
not 'real derivation' unless you register a new GType. There is no
GDK_BITMAP_TYPE or gdk_bitmap_get_type().

Aren't you glad that the whining from the gtk-- developers has already
started again?

-- 
Murray Cumming
murrayc usa net
www.murrayc.com