Thanks, Alan (and Colin).

And thanks for the warning, too! I've already learned to save my image 
*before* running any tests!

Michael.

Alan Lovejoy wrote:
> Michael: 
> 
> You will find that your new Metaclass ("m,") although it has a
> MethodDictionary and a superclass, does not have the right MethodDictionary
> or superclass to actually create any instances.  That's why the code in
> ClassBuilder does the following:
> 
> privateNewSubclassOf: newSuper
> 	"Create a new meta and non-meta subclass of newSuper"
> 	"WARNING: This method does not preserve the superclass/subclass
> invariant!"
> 	| newSuperMeta newMeta |
> 	newSuperMeta := newSuper ifNil:[Class] ifNotNil:[newSuper class].
> 	newMeta := Metaclass new.
> 	newMeta
> 		superclass: newSuperMeta
> 		methodDictionary: MethodDictionary new
> 		format: newSuperMeta format.
> 	^newMeta new
> 
> If you provided the new Metaclass with a MethodDictionary with the right
> CompiledMethods in it, you wouldn't need to give it a superclass.  But the
> usual practice is to make the superclass of a Metaclass instance be either
> Class (which is an instance of an instance of Metaclass) or else some other
> already-provisioned Metaclass instance.
> 
> Be forewarned: You're dealing with the "moebius strip" of Smalltalk's
> class-metaclass metamodel, and it's easy to get yourself very confused.
> Careful thought will be required to avoid error.
> 
> Note: The class of Metaclass is 'Metaclass class,' and the class of
> 'Metaclass class' is Metaclass. This provides the transitive closure of
> Smalltalk's metamodel.  The superclass of a root class is nil, the
> superclass of the metaclass of a root class should normally be Class, so
> that the instance of the metaclass instance (a root class) will inherit
> "Class behavior" from Class itself.
> 
> --Alan
> 
> -----Original Message-----
> From: squeak-dev-bounces at lists.squeakfoundation.org
> [mailto:squeak-dev-bounces at lists.squeakfoundation.org] On Behalf Of Michael
> van der Gulik
> Sent: Sunday, March 19, 2006 1:35 PM
> To: squeak-dev at lists.squeakfoundation.org
> Subject: Metaclass issues.
> 
> Hi all.
> 
> I've found myself in the unfortunate position of needing to work with
> Metaclasses for my own project!
> 
> I'll spare the details; you can email me directly if you want them :-).
> Why does the following happen?:
> 
> m := Metaclass new. " Construct the Metaclass instance for my new class"
> c := m new. "Make my new class".
> 
> The code will evaluate fine, but if you try to do anything with c, CPU usage
> hits 100% and I can't seem to break out of it.
> 
> Ideas? I suspect it is something like m not being initialised properly with
> a superclass etc. I'm taking my inspiration from
> ClassBuilder>>privateNewSubclassOf:.
> 
> Michael.
> 
> 
> 
> 
>