http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51749
--- Comment #23 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
So, perhaps this is solveable only through cooperation with glibc folks.
I guess a start for that would be to do an audit of the libstdc++ headers, what
symbols does it need from the glibc headers, and categorize them (symbols that
are defined in the C headers unconditionally, not guarded by any feature test
macros, symbols guarded by feature test macros that C++ wants to have visible
in the *.h headers (candidates for _ISOCXX98_SOURCE, _ISOCXX11_SOURCE,
_ISOCXX14_SOURCE feature test macros?), symbols we don't want to be visible in
the C headers but libstdc++ headers use privately for the implementation
(export as __ prefixed alternatives?). If glibc is changed, the question would
be what should be the new G++ feature test macro defaults, for
-std=gnu++{98,11,1y} I guess it should include the POSIX etc. stuff by default
too, for -std=c++{98,11,1y} perhaps it should be strict and require users to
use feature test macros explicitly.
