changes for approval, part three (solaris)

[David Edelsohn]

	Some observation:

1) libio/libio.h defines the __P() and __PMT() macros if _G_HAVE_SYS_CDEFS
is defined as 0.

2) libio/_G_config.h defines _G_HAVE_SYS_CDEFS as 1.

3) libio/_G_config.h was generated in v2 builds via gen-params and
contains target-dependent information.

4) libio/_G_config.h now uses the GNU/Linux libc version.

5) One cannot override the values in _G_config.h via os_defines.h because
os_defines.h is included via bits/c++config.h at the beginning of
_G_config.h (as it needs to be for other reasons).

What is the purpose / intent of the current use of _G_config.h?  At least
part of Phil's patch is working around a definition in _G_config.h which
is not appropriate for Solaris.  Not surprisingly, AIX has problems with
the assumptions in _G_config.h as well.

David