CWG2949 [temp.func.order] p6 Partial ordering rules don't take ellipsis into account

[ddvamp]

Full name of submitter: Artyom Kolpakov

Reference (section label): [temp.func.order] p6

Issue description: Firstly, the standard does not formally define what the length of the template-parameter-list is. This probably means the number of template-parameter.
Secondly, it is unclear what a function parameter list is. I suppose that means parameter-type-list. It is also unclear how its length is determined. This is the number of parameters only, or the ellipsis is also counted. Here and further, I omit the function parameter pack because according to CWG2942 it is proposed to remove it from the definition of parameter-type-list.
Then, in clauses (6.2.1.2.2) and (6.2.2) it is unclear what the positionally correspond for parameters mean when parameter corresponds to ellipsis. I don't see a specific clause, however, this example says that ellipsis is not a parameter, so you can think in two ways: the first, because there is no positionally corresponding parameter, their types cannot match; the second, because there is no positionally corresponding parameter, their types do not need to be checked (It's like unfolding empty packs for || and &&). But what happens if several parameters positionally correspond to ellipsis.

If ellipsis is not taken into account, it looks like [temp.func.order] p6 implies that parameter-type-lists must be the same in order to be checked against constraints. Also [over.match.best.general] p2.6 talks about the same (non-object-)parameter-type-lists (with some reservations).

Suggested resolution: Edit [temp.func.order] p6.1:

If their template-parameter-lists (possibly including template-parameters invented for an abbreviated function template ([dcl.fct])) contain a different number of template-parameters, or if their function parameter lists differ in length parameter-type-lists contain a different number of function parameters, or if exactly one of the templates contains an ellipsis, neither template is more specialized than the other.

[jensmaurer]

CWG2949

[jwakely]

@jensmaurer the new issue says it's from submission 631 but this is #632

[jensmaurer]

@jwakely Thanks, fixed.