Re: Chaining of dispose and finalize

[Russell Shaw <rjshaw netspace net au>]

John Cupitt wrote:
> On Sat, 04 Sep 2004 00:39:27 +1000, Russell Shaw <rjshaw netspace net au> wrote:
>
> > I understand that, and is what i do. What i mean is that most
> > objects do in class_init something like:
> >
> >   gobject_class->dispose=myobject_dispose;
> >
> > But this makes the original dispose code in gobject
> > inaccessible. Same for gobject_class->dispose=myobject_finalize.
>
>
> That assignment isn't overwriting gobject's dispose, it's setting the
> gobject dispose for this class. If you have a class called myclass
> derived from gobject, then that class has two ->dispose() function
> pointers: the standard gobject one, and the one you set. That's
> because every time you make a class, you get a fresh copy of every
> class above this one in the inheritance hierarchy.

I didn't realize copies of classes were stored in the g_type system.

I found that "The Glib Object system v0.9.0" has the information
in it:

  http://le-hacker.org/papers/gobject/ch05.html#howto-gobject-chainup

I had read only V0.8.0.

> Classes and class instances don't behave the same. When you make a
> class instance, you are building a single big, overlapping struct. You
> can override things in your parent classes.
>
> When you derive a new class, you build a lot of structs: one for each
> class in the hierarchy. When you chain up at the end of dispose,
> you're jumping up one struct in this pile of structs.

I knew that bit;)

> (hope I've understood you here, sorry if I've missed the point)