fix va-arg-22 and 931104-12 on powerpc-darwin

fix va-arg-22 and 931104-12 on powerpc-darwin

[Geoffrey Keating]

The varargs routines were doing the right thing, and the right value
was on the stack, but the wrong value was in register.

This does change the ABI on Darwin!  (It changes it back to what GCC
3.3 used to do.)

Bootstrapped & tested on powerpc-darwin.

===File ~/patches/gcc-darwin-varargpadding.patch============
2004-10-24  Geoffrey Keating  <geoffk@geoffk.org>

	* config/rs6000/darwin.h (BLOCK_REG_PADDING): Define.

Index: config/rs6000/darwin.h
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/rs6000/darwin.h,v
retrieving revision 1.65
diff -u -p -r1.65 darwin.h
--- config/rs6000/darwin.h	14 Oct 2004 22:30:42 -0000	1.65
+++ config/rs6000/darwin.h	24 Oct 2004 22:25:43 -0000
@@ -352,6 +352,12 @@ do {									\
    ? MAX (MAX ((COMPUTED), (SPECIFIED)), 128)          			 \
    : MAX ((COMPUTED), (SPECIFIED)))
 
+/* Specify padding for the last element of a block move between
+   registers and memory.  FIRST is nonzero if this is the only
+   element.  */
+#define BLOCK_REG_PADDING(MODE, TYPE, FIRST) \
+  (!(FIRST) ? upward : FUNCTION_ARG_PADDING (MODE, TYPE))
+
 /* XXX: Darwin supports neither .quad, or .llong, but it also doesn't
    support 64 bit PowerPC either, so this just keeps things happy.  */
 #define DOUBLE_INT_ASM_OP "\t.quad\t"
============================================================

Re: fix va-arg-22 and 931104-12 on powerpc-darwin

[Andrew Pinski]

On Oct 24, 2004, at 10:01 PM, Geoffrey Keating wrote:

>
> The varargs routines were doing the right thing, and the right value
> was on the stack, but the wrong value was in register.
>
> This does change the ABI on Darwin!  (It changes it back to what GCC
> 3.3 used to do.)
>
> Bootstrapped & tested on powerpc-darwin.

Thanks for fixing this, I see that we now have only one testsuite
regression still left, that is not xfailed "debug/dwarf2/dwarf-die7.c".
This was not noticed on powerpc-darwin until after darwin enabled 
dwarf-2
debugging support so it will not show up on the regression tester at 
all.

Thanks,
Andrew Pinski

Re: fix va-arg-22 and 931104-12 on powerpc-darwin

[Andreas Tobler]

Geoffrey Keating wrote:

> The varargs routines were doing the right thing, and the right value
> was on the stack, but the wrong value was in register.
> 
> This does change the ABI on Darwin!  (It changes it back to what GCC
> 3.3 used to do.)

Thanks, I can confirm that libffi is down to zero failures.

Andreas