Re: [DotGNU]Overriding C++ virtual methods in C#

[Gopal V]

If memory serves me right, Marcus wrote:
> I've come up with one way to achieve this goal in effect. 

Nope ... you haven't ... at least you haven't made it far enough ..

Let me explain ... 

I have 

QWidget which inherits from QObject , and QObject has a virtual method
`vmethod' declared for it ... 

So I create QWidgetBridge and QObjectBridge to wrap these ... so here
QWidgetBridge inherits from QObjectBridge to inherit bridging (which
seems to be the obvious plan here, correct me if I'm wrong).

So the inheritance pattern looks like this 

QObject 
 |     \__ QObjectBridge
 |              |
QWidget         |
       \__ QWidgetBridge

(which is the Multiple Inheritance diamond drawn in ASCII art :-)

Since QWidgetBridge should inherit QObjectBridge *and* QWidget to be 
useful for the purpose .... 

Now if I call QWidgetBridge.vmethod() which one will get called ?,
the options are

1) QWidget::vmethod           (inherited via QWidget)
2) QObjectBridge::vmethod     (inherited via QWidgetBridge)

Since QWidgetBridge does not override vmethod, we really cannot be sure.

(DISCLAIMER: I don't know if there's a way in C++ to avoid this)

Naming the bridge function differently doesn't work either (that was almost 
too obvious,right ?) ..

So all bridge classes must override *all* the previous methods in their
wrapped class and not use inheritance from the bridge of their parent class
to do avoid this ambiguity .... 

Which means you cannot override methods in parent bridges or cast them to 
parent bridges, and thus the system breaks down....

What this really needs is a way to allow PInvoke to check for C# functions
like they do in JNI , delegates need to be stored in C++ wrappers to be 
accessible , causing all the problems I have described so far ...

Hope that makes it clear,
Gopal
-- 
The difference between insanity and genius is measured by success