Mahh braainn huhts! (libstdc++-v3 vs Darwin)

[Stan Shebs]

Mainline libstdc++-v3 fails to build on Darwin because of a missing
definition for _BSD_WCHAR_T_.  After chasing tangled ifdefs for awhile,
we came up with the following patch to ginclude/stddef.h that allows
compilation to continue:

RCS file: /cvs/repository/CoreTools/gcc3/gcc/ginclude/stddef.h,v
retrieving revision 1.4
diff -c -3 -p -r1.4 stddef.h
*** stddef.h    2001/01/25 01:52:31     1.4
--- stddef.h    2001/05/15 22:09:02
*************** Boston, MA 02111-1307, USA.  */
*** 83,89 ****
  #undef _SIZE_T_
  #undef _BSD_SIZE_T_
  #endif
! #if defined (__need_wchar_t) || defined (_STDDEF_H_)
  #undef _WCHAR_T_
  #undef _BSD_WCHAR_T_
  #endif
--- 83,89 ----
  #undef _SIZE_T_
  #undef _BSD_SIZE_T_
  #endif
! #if defined (__need_wchar_t) //|| defined (_STDDEF_H_)
  #undef _WCHAR_T_
  #undef _BSD_WCHAR_T_
  #endif

This is almost certainly wrong, but I don't understand how things
are *supposed* to work here - testing the include guard seems pretty
lame to me, I guess it's connected to glibc including this file
multiple times or something.  Anyway, the #undef _BSD_WCHAR_T_ is
wanted because later on there is a #ifndef test that gives us the
value we really want, but __need_wchar_t is *not* defined because
Darwin is a BSD variant - or maybe I have all this backwards. :-) :-(

If anybody knows the theory of this stuff, and especially how it's
intended to work for BSD flavors, I'd be grateful for any hints.

Stan