RFC: [GDL2/base] unableToSetNullForKey: / unableToSetNilForKey:

[David Ayers]

Hello...

I'm stuck in a "little" ObjC runtime dilema.  I know that if mutiple 
implementations of categories exists, that the choice of implementation 
is undefined.  I was wondering if someone could nudge me in the right 
direction on  handeling this while trying to follow OS X and WO 4.5 API.

Current gnustep-base and OS X Foundation call unableToSetNilForKey: when 
takeValue:forKey: is invoked with NSNull/nil value on scalar key.  EOF 
of WO4.5 should call unableToSetNullForKey:.  I was hoping to find some 
way to override NSObject (NSKeyValueCoding) unableToSetNilForKey: in 
GDL2 to invoke unableToSetNullForKey:.

That way, if any base classes ever implemented unableToSetNilForKey: 
there behavior would be unchanged.  But code implementing 
unableToSetNullForKey: instead would  have a 2nd chance of handling the 
situation before GDL2 NSObject (EOKeyValueCoding) unableToSetNullForKey: 
would raise the coresponding exception.

Now on the one hand we might be able to add a hook in gnustep base that 
would allow "a second chance" selector to be invoked in 
unableToSetNilForKey: but then this would only work for GDL2 when 
compiled with base and not with OS X.

I thought we reimplemented KVC in GDL2 to check if the reciever responds 
to unableToSetNullForKey: in GSObjCSetValue before calling 
unableToSetNilForKey: but that won't work either.  If we can't override 
unableToSetNilForKey: for NSObject, then we won't be able to override 
takeValue:forKey: either.

How can we invoke unableToSetNullForKey: on OS X if 
unableToSetNilForKey: of NSObject is invoked?

Cheers,
Dave