This change is incorrect. Here's how this section works today:

• The "set of overloaded functions" is the set that we form through name lookup and deduction.
• Then we apply the "A function with type \tcode{F} is selected for the function type \tcode{FT} of the target type required in the context if" rule to select some subset.
• Then we apply the "set of selected functions" rule to pick between the selected functions.

This change effectively moves the second bullet after the third by making the third bullet apply to the "set of overloaded functions" instead of the "set of selected functions".

Ok, I was of the mistaken belief that "selected" in the overload resolution context means "single function". I'm now seeing the "selected" state is an intermediate set where additional partial ordering etc. is applied to find the "single chosen one". For example, p3 talks about "is selected" and then requiring an ampersand to form the pointer-to-member, which should probably be the very last check after you've actually selected a function in p4.

It would probably be helpful to move the phrasing for the second bullet, which is in p1, to below p2, where template argument deduction is discussed.

This is getting large enough to warrant a core issue, it seems.

@jensmaurer: can we close this in favour of a CWG issue?

Let's discuss in Jacksonville how to proceed here.

Parent issue #1691 to become a core issue.