This wording still doesn't make any sense. Did this perhaps mean R T0::*f() cv? Or is this supposed to be a declaration of a member function? Perhaps LWG can give us some guidance about what was intended here.

I think it makes sense with the fix. As an example of what the item is talking about, consider the member function:

size_t string::size() const

(R=size_t, T0=string, f=size, cv=const)

The item says that if T (reference_wrapper's parameter) is a pointer to this thing, then T1 (the resulting argument_type) is cv T0*, which is const string*.

So it is saying that reference_wrapper<size_t(string::*)()const>::argument_type is const string*, which is also what I get with libstdc++ and libc++.

(I agree it's sloppy and confusing phrasing though.)

This is talking about a type T. There's no way a type can be "a pointer to member function size_t string::size() const" -- that's not a property of a type.

My concerns here are twofold:

1. This may be a normative change, depending on whether an argument_type was also supposed to be defined for a pointer to member function with parameters or not. (With your change we seem to say only the no-parameter case is covered, whereas without it, this appears to apply regardless of whether the function has parameters.)
2. This change still leaves the wording meaningless.

I think we do want to make a change in this direction (and I suspect the right change would be to use R T0::*f()cv), but I'd like LWG to look at this and decide what exactly to do.

I think it is supposed to be R T0::f() cv, and should be read as:

a pointer, to a member function like R T0::f() cv

i.e.

a pointer to member function that points to a member function such as R T0::f() cv

It certainly would be clearer to talk about the type of the pointer to member function, not show a declaration of a member function.

i.e. Eelis's suggestion is not a normative change, the intention, and I believe what all implementations do, is to only define argument_type when it's a pointer to member function for a member function taking no parameters.

For a member function taking exactly one parameter first_argument_type and second_argument_type are defined.

For member functions taking two or more parameters none of those typedefs is defined.

tr1::reference_wrapper says:

— a pointer to member function type with cv-qualifier cv and no arguments; the type T1 is cv T* and R is the return type of the pointer to member function

and then for the R T0::f(T2) case:

— a pointer to member function with cv-qualifier cv and taking one argument of type T2; the type T1 is cv T* and R is the return type of the pointer to member function

So at some point we rephrased that and mangled it.

It was changed by http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#521 and it was meant to have the () but the editor didn't apply the DR correctly.