PERFORCE change 109347 for review

[Warner Losh]

http://perforce.freebsd.org/chv.cgi?CH=109347

Change 109347 by imp at imp_lighthouse on 2006/11/06 08:30:32

	Use the new MACHINE_CPU instead of MACHINE_ARCH as the basis for
	TARGET_ARCH.  I'm unsure if the amd64 -> x86_64 translation should
	happen like it does in this file, or if that's another case for
	substitution in bsd.own.mk.  I kinda think this is the right place,
	since it is a gnuism and the rest of the tree is setup so that
	amd64 is the right thing to do.
	
	The binultils infrastructure wants 'TARGET_BIG_ENDIAN' defined
	when we're building big endian architectures for those CPUs
	that can support many different byte sexes.  Since I'm uneasy
	about doing a regular expression match against "*eb" I have
	listed the two cases where we have multi-endian architectures
	(mips and arm).  powerpc may also be one, and some that NetBSD
	supports that FreeBSD does not.
	
	Note: armeb support isn't there yet, and I believe this will not
	break the present TARGET=arm TARGET_ARCH=arm TARGET_BIG_ENDIAN=t
	way of building for xscale based big endian boards.

Affected files ...

.. //depot/projects/arm/src/gnu/usr.bin/binutils/Makefile.inc0#2 edit

Differences ...

==== //depot/projects/arm/src/gnu/usr.bin/binutils/Makefile.inc0#2 (text+ko) ====

@@ -6,13 +6,16 @@
 
 VERSION=	"2.15 [FreeBSD] 2004-05-23"
 
-TARGET_ARCH?=	${MACHINE_ARCH}
+TARGET_ARCH?=	${MACHINE_CPU}
 .if ${TARGET_ARCH} == "amd64"
 BINUTILS_ARCH=x86_64
 .else
 BINUTILS_ARCH=${TARGET_ARCH}
 .endif
 TARGET_TUPLE?=	${BINUTILS_ARCH}-obrien-freebsd
+.if ${MACHINE_ARCH} == "armeb" || ${MACHINE_ARCH} == "mipseb"
+TARGET_BIG_ENDIAN=t
+.endif
 
 # RELTOP is the relative path to this point in the source or object
 # tree, from any subdirectory of same.  It gets extra "../" prefixes