CWG2717 [temp.variadic] should specify a special case for the expansion of alignment-specifier

[xmh0511]

Full name of submitter (unless configured in github; will be published with the issue): Jim X

[dcl.align] p1 says

An alignment-specifier with an ellipsis is a pack expansion ([temp.variadic]).

template<class ...T>
struct Align{
  alignas(T...) unsigned char buffer[128];
};

If there is no special specification, the expansion is defined as [temp.variadic] p8

The instantiation of a pack expansion considers items E1,E2,…,EN, where N is the number of elements in the pack expansion parameters. Each Ei is generated by instantiating the pattern and replacing each pack expansion parameter with its ith element.

[temp.variadic] p11

The instantiation of any other pack expansion produces a list of elements E1,E2,…,EN.

That is, given Align<int, short> will instantiate the pack expansion to produce alignas(int), alignas(short), which is not a valid attribute-specifier-seq in a declaration. That is, a valid attribute-specifier-seq in a declaration is not comma-separated. Even if the note in the [temp.variadic] p11 may imply the intent, however, it's not a formal rule

[Note 1: The variety of list varies with the context: expression-list, base-specifier-list, template-argument-list, etc.
— end note]

[jensmaurer]

I'm not seeing that the default rule would give you alignas(T1), alignas(T2). It seems you'd get alignas(T1,T2) for alignas(T...). Which is also ill-formed; I think we want alignas(T1) alignas(T2).

Update: That's wrong, [temp.variadic] p5.9 says "In an alignment-specifier (9.12.2); the pattern is the alignment-specifier without the ellipsis."

[jensmaurer]

CWG2717