Hm... sorry, I can't reproduce, seems to work just fine now.
So nevermind.

Daniel

Avi Bryant <avi at beta4.com> wrote:
> 
> On Tuesday, October 14, 2003, at 11:48 AM, Daniel Vainsencher wrote:
> 
> > When I try to load
> > http://modules.squeakfoundation.org/People/dvf/Packages/SMBase- 
> > dvf.13.mc
> > z into a more or less clean image 3.7 using MC, I get an error because
> > of adding a variable to a class that exists in a superclass (IIUC). I'd
> > guess what is happening is that the superclass gets redefined to not
> > have the variable, but it comes later in the load order. This can be
> > worked around by proceeding in the walkback. After that I loaded the
> > file again, and things seemed ok.
> 
> Hm, that doesn't quite make sense - the superclass should always be  
> loaded before the subclass.  I would expect this problem to happen when  
> pushing ivars up from a subclass to a superclass, but not the other way  
> around.
> 
> Either way, of course, it needs to be fixed.  I'm not sure exactly what  
> the best approach is.  Incidentally, I've found in the past that  
> changesets can exhibit a similar problem.
> 
> Perhaps MC should catch such exceptions, and defer those loads, trying  
> them again at the end?  Does it throw a generic Error or is there a  
> specific exception for it?