Re: AtkHyperlink question

[Bill Haneman <bill haneman sun com>]

Jeff wrote:

>   Hi,
> 
> I was going through ATK again and was just wondering about a few
> inconsistencies.
> 
> 1. Should the return value from 'atk_hyperlink_get_uri' be a
> G_CONST_RETURN ghar* rather than ghar*.

I think this would have been better 'const-ified' as you suggest.
Pretty late in the game to change it, but AFAIK nobody
other than AT-SPI is using this API at all yet.

Opinions, g2r team?

> 2. Should these two fucntion pointers declared in AtkObjectFactoryClass:
> 
> AtkObject* (* create_accessible) (GObject *obj);
> 
> GType (* get_accessible_type) (void);
> 
> actually be
> 
> AtkObject* (* create_accessible) (AtkObjectFactory *factory, GObject *obj);
> 
> GType (* get_accessible_type) (AtkObjectFactory *factory);

Thanks for catching this, Jeff;

This is indeed odd looking.  Of course the exported wrappers,
atk_object_factory_create_accessible () and 
atk_object_factory_get_accessible_type () 
are "normal" in their signatures, only the function
pointers are peculiar.

Again, consistency with GTK+ would require the API you
suggest, Jeff, and I tend to think you are right.
However as Padraig says, it works as-is.

Any opinions from g2r?  I think that if we don't change this
now it'll be ensconced as an eccentricity for the lifetime of
GTK+-2.X, but then again, it's not overtly harmful.  

The only technical limitation introduced is that 
instance-specific data for AtkObjectFactory-s can't 
readily be passed to the method implementation code.
 
-Bill

> Jeff.