Martin Sebor wrote:
FWIW, if the standard says that numeric_limits<cv-qualified T>
is different than numeric_limits<T> I would consider it a defect
in the standard
You saw my first reply, I tried to articulate in detail why, in fact,
I'm strongly convinced that the standard *forbids* specializations for
anything different from the cv-unqualified fundamental type. You also
read that from a general, philosophical, stand-point, I agree with you
and Roberto, seems rather obvious that a const int can assume the very
same values of an int! Therefore, if you agree with my reading of the
standard, well, we indeed have a defect! ;)