fix HOST_WIDE_INT mismatch on i386-apple-darwin

[Geoffrey Keating]

If you configured GCC with '--target=i386-apple-darwin', spelt exactly
like that, then on a trivial line of code like

char __mbstate_t[128];

you would get a complaint that the array size was too large, and so
the build would fail at the first use of the new compiler (usually for
crt3.o).

The problem was that libcpp and gcc were disagreeing on the size of
HOST_WIDE_INT, causing lexing of numbers to be wrong.  gcc/config.gcc
has

i[34567]86-*-darwin*)
	need_64bit_hwint=yes

but libcpp/configure.ac had 'i[34567]86-*-darwin[0-9]*'.

Fixed thusly.

-- 
- Geoffrey Keating <geoffk@apple.com>

===File ~/patches/gcc-libcpp-64hwidarwin86.patch============
2006-09-22  Geoffrey Keating  <geoffk@apple.com>

	* configure.ac: Make need_64_bit_hwint case for x86-darwin
	match exactly the glob in gcc/config.gcc.
	* configure: Regenerate.

Index: configure.ac
===================================================================
--- configure.ac	(revision 117150)
+++ configure.ac	(working copy)
@@ -118,7 +118,7 @@
 	x86_64-*-* | \
 	ia64-*-* | \
 	hppa*64*-*-* | parisc*64*-*-* | \
-	i[34567]86-*-darwin[0-9]* | \
+	i[34567]86-*-darwin* | \
 	i[34567]86-*-solaris2.1[0-9]* | \
 	mips*-*-* | \
 	mmix-*-* | \
============================================================