Re: Remove __GNUC__ conditions for "long long" from string.h and wchar.h

[Roland McGrath <roland at hack dot frob dot com>]

We have not yet reached consensus on unconditional use of long long.
The presence of unconditional uses elsewhere has no bearing on that
policy choice.  When the policy becomes clear, then those may be
existing bugs rather than correct precedents.

I, for one, think that at least -std=c89 -D_GNU_SOURCE ought to work
and that it ought not to rely on system-headers exceptions.  So for GCC
(and its emulators) that means more consistent use of __extension__,
which is already used correctly in the cases you're touching here.

So the effect here is for non-GCC compilers.  The change
conservatively within what IMHO is the pre-existing policy intent
is to use __GLIBC_HAVE_LONG_LONG.

The pure redundancy eliminations like __USE_GNU when __USE_ISOC99 is
already tested are fine, and fall well enough into the "obvious" category.


Thanks,
Roland