This is an unofficial snapshot of the ISO/IEC JTC1 SC22 WG21 Core Issues List revision 113d. See http://www.open-std.org/jtc1/sc22/wg21/ for the official list.

2024-03-20


608. Determining the final overrider of a virtual function

Section: 11.7.3  [class.virtual]     Status: CD2     Submitter: Mike Miller     Date: 7 December 2006

[Voted into WP at October, 2009 meeting.]

According to 11.7.3 [class.virtual] paragraph 2:

Then in any well-formed class, for each virtual function declared in that class or any of its direct or indirect base classes there is a unique final overrider that overrides that function and every other overrider of that function. The rules for member lookup (6.5.2 [class.member.lookup]) are used to determine the final overrider for a virtual function in the scope of a derived class but ignoring names introduced by using-declarations.

I think that description is wrong on at least a couple of counts. First, consider the following example:

    struct A { virtual void f(); };
    struct B: A { };
    struct C: A { void f(); };
    struct D: B, C { };

What is the “unique final overrider” of A::f() in D? According to 11.7.3 [class.virtual] paragraph 2, we determine that by looking up f in D using the lookup rules in 6.5.2 [class.member.lookup]. However, that lookup determines that f in D is ambiguous, so there is no “unique final overrider” of A::f() in D. Consequently, because “any well-formed class” must have such an overrider, D must be ill-formed.

Of course, we all know that D is not ill-formed. In fact, 11.7.3 [class.virtual] paragraph 10 contains an example that illustrates exactly this point:

struct A {
    virtual void f();
};
struct B1 : A {     // note non-virtual derivation
    void f();
};
struct B2 : A {
    void f();
};
struct D : B1, B2 { // D has two separate A subobjects
};

In class D above there are two occurrences of class A and hence two occurrences of the virtual member function A::f. The final overrider of B1::A::f is B1::f and the final overrider of B2::A::f is B2::f.

It appears that the requirement for a “unique final overrider” in 11.7.3 [class.virtual] paragraph 2 needs to say something about sub-objects. Whatever that “something” is, you can't just say “look up the name in the derived class using 6.5.2 [class.member.lookup].”

There's another problem with using the 6.5.2 [class.member.lookup] lookup to specify the final overrider: name lookup just looks up the name, while the overriding relationship is based not only on the name but on a matching parameter-type-list and cv-qualification. To illustrate this point:

    struct X {
        virtual void f();
    };
    struct Y: X {
        void f(int);
    };
    struct Z: Y { };

What is the “unique final overrider” of X::f() in A? Again, 11.7.3 [class.virtual] paragraph 2 says you're supposed to look up f in Z to find it; however, what you find is Y::f(int), not X::f(), and that's clearly wrong.

Proposed Resolution (December, 2006):

Change 11.7.3 [class.virtual] paragraph 2 as follows:

Then in any well-formed class, for each virtual function declared in that class or any of its direct or indirect base classes there is a unique final overrider that overrides that function and every other overrider of that function. The rules for member lookup (6.5.2 [class.member.lookup]) are used to determine the final overrider for a virtual function in the scope of a derived class but ignoring names introduced by using-declaration s. A virtual member function vf of a class C is a final overrider unless the most derived class (6.7.2 [intro.object]) of which C is a base class (if any) declares or inherits another member function that overrides vf. In a derived class, if a virtual member function of a base class subobject has more than one final overrider, the program is ill-formed.

Proposed resolution (July, 2009):

Change 11.7.3 [class.virtual] paragraph 2 as follows:

...Then in any well-formed class, for each virtual function declared in that class or any of its direct or indirect base classes there is a unique final overrider that overrides that function and every other overrider of that function. The rules for member lookup (6.5.2 [class.member.lookup]) are used to determine the final overrider for a virtual function in the scope of a derived class but ignoring names introduced by using-declarations. A virtual member function C::vf of a class object S is a final overrider unless the most derived class (6.7.2 [intro.object]) of which S is a base class subobject (if any) declares or inherits another member function that overrides vf. In a derived class, if a virtual member function of a base class subobject has more than one final overrider, the program is ill-formed. [Example: ... —end example] [Example:

    struct A { virtual void f(); };
    struct B: A { };
    struct C: A { void f(); };
    struct D: B, C { };    // OK; A::f and C::f are the final overriders
                           // for the B and C subobjects, respectively

end example]