[RFC] Second half of libstdc++/8949

[Paolo Carlini]

Hi,

the first part of the PR seems already fixed for 3.3.
The second half, however, about:

numeric_limits<short>::is_iec559
numeric_limits<unsigned short>::is_iec559
numeric_limits<int>::is_iec559
numeric_limits<unsigned int>::is_iec559
numeric_limits<long>::is_iec559
numeric_limits<unsigned long>::is_iec559

seems still an issue: according to submitter, those should be *false* 
not *true*. Indeed, in my reading of 18.2.1.2, 53 and 18.2.1.5 he's right.

Do you agree?

In case, a straightforward patch could be envisaged as soon as the 
mainline reopens (perhaps safe for 3.3 too).

Ciao, Paolo.