After today's teleconference, I noticed what I believe is a defect in
the grammar for __has_cpp_attribute.

has-attribute-expression:
  __has_cpp_attribute ( attribute-scoped-token )

I believe attribute-scoped-token isn't what we want; this production
would only be useful for vendor-specific attributes (since
attribute-scoped-token is attribute-namespace :: identifier). This
would be problematic for our examples of
__has_cpp_attribute(deprecated). ;-)

I think the correct production should be:

has-attribute-expression:
  __has_cpp_attribute ( attribute-token )

Thoughts?

~Aaron