@cmeerw, something like this?

yes, looks good to me.

@jensmaurer Should we also improve the second sentence of [temp.deduct.general] p8 to make [temp.deduct.general.note.4] have a formal explanation? The current sentence is that

Only invalid types and expressions in the immediate context of the function type, its template parameter types, and its explicit-specifier can result in a deduction failure.

I think we may change it to

Only invalid types and expressions in the immediate context of the function type outside of the exception specification, its template parameter types, and its explicit-specifier can result in a deduction failure.

In other words, the substitution in exception specification that would result in invalid types and expressions can make the program be ill-formed, which is what the note saying

Incidentally, the deduction failure can also occur in the default argument of a template parameter. For instance

struct A{
    using type = int;
};
template<class T, int U =  T::v>
void fun(T){};

template<class T>
void fun(T*){}
int main(){
  A* ptr = 0;
  fun(ptr);
}

The current wording only mentions the template parameter types.

@xmh0511, since this is not the place where we specify exactly where we substitute, we can be a bit hazy here. We should not be incorrect. I've fixed the obvious problem, and I've moved the "only" to apply to "immediate context", not to "invalid stuff".

@jensmaurer Could this be something you could quickly look at in CWG? No objections otherwise, but if you have a slot for GitHub reviews, it would be nice to get core sign-off.

Created CWG2618 for CWG review and visibility.

Should have been superseded by #5984 (bb4acc8).