[PATCH 13/13] Use GCC builtins for copysign functions if desired.

[Joseph Myers]

On Wed, 4 Dec 2019, Stefan Liebler wrote:

> Okay. Then I've removed the macros for float and double and call the builtins
> unconditionally.
> But I still need the long double version as the f128 builtin is not available
> with every GCC version.

I think the _Float128 version could be defined to 1 for GCC 7 and later 
(so both long double and _Float128 versions could be removed once we 
require GCC 7 or later to build glibc).

-- 
Joseph S. Myers
joseph@codesourcery.com