Vcall/vbase offset clarification

[Jason Merrill]

>>>>> Jim Dehnert <dehnert at baalbek.engr.sgi.com> writes:

 >> From: Martin von Loewis <loewis at informatik.hu-berlin.de>
 >> To: mark at codesourcery.com
 >> 
 >> > > From mitchell at codesourcery.com  Tue Jan 25 13:26:12 2000
 >> > >   struct A { void f();}
 >> > >   struct B : public virtual A {};
 >> [...]
 >> > >   o First, we convert to the most derived class in which `f' was
 >> > >     overridden.  That's `A'.  We know statically what offset
 >> > >     to adjust `b' by to get an `A*'.
 >> 
 >> Maybe you didn't mean it that way, but: How do we know statically the
 >> offset needed to adjust a B* to an A*? Wouldn't we have to look in the
 >> vtable at run time to fetch the vbase offset?
 >> 
 >> > >   o We pass this value to the function found in the vtable for `A'.
 >> > > 
 >> > >   o That function will first perform the constant adjustment from 
 >> > >     `A' to `B'.  
 >> 
 >> Or perhaps that shouldn't have been virtual inheritance in the B
 >> bases?

 > I think the right statement is 'we know statically how to adjust "b" to
 > get an "A*".  If A is a non-virtual base, we know the offset
 > statically.  If it is a virtual base, we know statically where to fetch
 > the vbase offset in the vtable.'

No, I'm with Martin on this.  We're trying to adjust from A* to C*; the
adjustment from B* to A* is not useful.

Jason