Resolution of B-2 not reflected in B-6 (aka covariant returns)

[Martin von Loewis]

> I think that is what was meant.  To restate:
> 
>   When a virtual function is declared in a class, it gets a new vtable slot
>   unless it overrides a function from the primary base and conversion
>   between the two return types does not require an adjustment.
> 
> It seems unnecessary to force a new entry if the return types are related
> by single inheritance.

In the B-2 discussion, the notes say

# Agreement was reached to avoid the complication of eliminating some
# of the Vtable entries. Thus, the Vtable will have one entry for each
# accessible return type of a covariant virtual function.

I agree with that resolution; it follows that a new slot is allocated
even if there are no return adjustments. Of course, a smart compiler
would use the same function entry as in the base part, instead of
generating a new one.

Please note that this rule would also apply in case of different cv
qualification, as allowed per 10.3/5

# both pointers or references have the same cv­qualification and the
# class type in the return type of D::f has the same cv­qualification
# as or less cv­qualification than the class type in the return type
# of B::f.

So, in case of

struct Base{
  virtual const Base* foo();
};

struct Derived{
  virtual Base* foo();
};

there would be two vtable slots for Derived::foo, which would most
likely be implemented by the same entry point.

If that is undesirable, I would not object to re-opening B-2, and
indeed eliminating some vtable entries. However, I think it is fine as
it is now.

Martin