[RFA] Mark Sparc/Linux as multi-arch

["David S. Miller" <davem@redhat.com>]

This actually fixes some testsuite failures because the trap location
diddling in sparc.c:sparc_fix_call_dummy() only works if multiarch is
enabled.

I don't classify the sparc_fix_call_dummy behavior, as a bug.
Honestly there is no other way to implement that except perhaps
with a state variable, but that state variable would need to
be tied to the current task/thread/etc. and that would just be
unnecessarily complicated.

In any event, tested under sparc-linux-gnu.  There are no regressions,
in fact there are quite a few less :-)  Unexpected failures went from
125 (before the patch) to 112, this is using gcc-3.1 CVS.

Here is the patch:

2002-04-19  David S. Miller  <davem@redhat.com>

	* config/sparc/tm-linux.h (GDB_MULTI_ARCH): Define to
	GDB_MULTI_ARCH_PARTIAL
	* config/sparc/tm-sp64linux.h (GDB_MULTI_ARCH): Do not
	define, let tm-sp64.h do it.

--- config/sparc/tm-linux.h.~1~	Fri Apr 19 13:58:11 2002
+++ config/sparc/tm-linux.h	Fri Apr 19 14:54:47 2002
@@ -23,6 +23,8 @@
 #ifndef TM_SPARCLINUX_H
 #define TM_SPARCLINUX_H
 
+#define GDB_MULTI_ARCH GDB_MULTI_ARCH_PARTIAL
+
 #include "sparc/tm-sparc.h"
 
 #undef SPARC_TARGET_LONG_DOUBLE_BYTES
--- config/sparc/tm-sp64linux.h.~1~	Fri Apr 19 13:58:03 2002
+++ config/sparc/tm-sp64linux.h	Fri Apr 19 14:58:09 2002
@@ -21,8 +21,6 @@
 #ifndef TM_SPARC_LIN64_H
 #define TM_SPARC_LIN64_H
 
-#define GDB_MULTI_ARCH 0
-
 #include "sparc/tm-sp64.h"
 
 #undef SPARC_TARGET_LONG_DOUBLE_BYTES