Re: GtkAda : Trying to derive a widget

[DrPi <314@drpi.fr>]

Le 09/04/2021 à 08:12, J-P. Rosen a écrit :
> Le 09/04/2021 à 07:28, DrPi a écrit :
>>> Without "use" Initialize is unambiguous.
>>>
>> That makes sense.
>> However, I don't understand why there is ambiguity when using "use".
>> Debug_Panel (the type) is of type Debug_Panel_Record, so Initialize 
>> should resolve to the one using this type. Well, I guess I'm wrong.
> No, the type is "access Debug_Panel_Record'Class", and the other 
> initialize is for "access Gtk_Scrolled_Window_Record'Class", which 
> covers the other one.
> 
What I mean is that one "Initialize" procedure has a parameter of type 
"access Debug_Panel_Record'Class" and the other of type "access 
Gtk_Scrolled_Window_Record'Class".
The call to "Initialize" is made with a "Debug_Panel" type which is an 
access to "Debug_Panel_Record'Class" not " an access to 
"Gtk_Scrolled_Window_Record'Class".
So, why is there an ambiguity here ?


> Remember that a class wide type ('Class) is a different type that covers 
> all descendants.
> 
>>> P.S. When you create new widget it is better to use a more general 
>>> ancestor hiding insufficient details, e.g.
>>>
>>>    type Debug_Panel_Record is
>>>       new Gtk.Widget.Gtk_Widget_Record with private;
>>>    ...
>>> private
>>>    type Debug_Panel_Record is
>>>       new Gtk.Scrolled_Window.Gtk_Scrolled_Window_Record with
>>>    record
>>>       ...
>>>    end record;
>>
>> I'm surprized this is possible to write such a thing in Ada.
>> What does the compiler do with this ?
>> There is no problem, since 
> Gtk.Scrolled_Window.Gtk_Scrolled_Window_Record is a descendant of 
> Gtk.Widget.Gtk_Widget_Record. There is no lie: a Debug_Panel_Record IS A 
> Gtk_Widget_Record. The private view has more information: it actually IS 
> A Gtk_Scrolled_Window_Record, but the extra properties are not 
> accessible outside from the package body.
> 
Ok. That's interesting.
One more thing learned today :)