Re: testsuite tweaks for powerpc64

[Alan Modra <amodra at bigpond dot net dot au>]

On Mon, Feb 09, 2004 at 10:02:19AM -0800, Dale Johannesen wrote:
> On Feb 9, 2004, at 4:56 AM, Alan Modra wrote
> 
> >Index: gcc/testsuite/gcc.dg/darwin-longlong.c
> >===================================================================
> >RCS file: /cvs/gcc/gcc/gcc/testsuite/gcc.dg/darwin-longlong.c,v
> >retrieving revision 1.1
> >diff -u -p -r1.1 darwin-longlong.c
> >--- gcc/testsuite/gcc.dg/darwin-longlong.c	26 Jan 2004 17:48:49 
> >-0000	1.1
> >+++ gcc/testsuite/gcc.dg/darwin-longlong.c	9 Feb 2004 12:16:56 -0000
> >@@ -25,7 +25,7 @@ int main()
> >
> >   /* Exit on systems without 64bit instructions.  */
> >   signal (SIGILL, sig_ill_handler);
> >-  asm volatile ("extsw r0,r0");
> >+  asm volatile ("extsw 0,0");
> >   signal (SIGILL, SIG_DFL);
> 
> This is wrong (assuming the test is Darwin-specific as its name 
> suggests).
> http://gcc.gnu.org/ml/gcc-patches/2004-02/msg00122.html
> The -mregnames option doesn't exist on Darwin.

The test is run for powerpc64-linux too, where it fails due to r0
being a symbol.  How about using

asm volatile ("extsw %r0,%r0");

instead?

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre