[PATCH,testsuite] xfail pr23098.c on vxworks targets

[PATCH,testsuite] xfail pr23098.c on vxworks targets

[Nathan Froyd]

[-- Attachment #1: Type: text/plain, Size: 756 bytes --]

The attached patch XFAILs gcc.target/i386/pr23098.c on x86 vxworks
targets.  The testcase was written to ensure that loads of a "easy"
floating-point constant (e.g. 0.0 or 1.0) get turned into the
appropriate x86 instructions, rather than loading from memory.
Unfortunately, the testcase fails because VxWorks takes a somewhat
different approach to loading such constants, one that the
deligitimize_address hook does not recognize.  Rewriting the x86 backend
to help VxWorks out here was judged too hairy and therefore XFAILing the
testcase seemed like the best way forward.

Tested on i586-wrs-vxworks.  OK to commit?

-Nathan

2007-07-26  Nathan Froyd  <froydnj@codesourcery.com>

	gcc/testsuite/
	* gcc.target/i386/pr23098.c: XFAIL on vxworks targets.


[-- Attachment #2: pr23098.patch --]
[-- Type: text/plain, Size: 572 bytes --]

Index: gcc/testsuite/gcc.target/i386/pr23098.c
===================================================================
--- gcc/testsuite/gcc.target/i386/pr23098.c	(revision 177473)
+++ gcc/testsuite/gcc.target/i386/pr23098.c	(working copy)
@@ -1,7 +1,7 @@
 /* PR rtl-optimization/23098 */
 /* { dg-do compile } */
 /* { dg-options "-O2 -fPIC" } */
-/* { dg-final { scan-assembler-not "\.LC\[0-9\]" } } */
+/* { dg-final { scan-assembler-not "\.LC\[0-9\]" { xfail *-*-vxworks* } } } */
 /* { dg-require-effective-target ilp32 } */
 /* { dg-require-effective-target fpic } */
 

Re: [PATCH,testsuite] xfail pr23098.c on vxworks targets

[Janis Johnson]

On Thu, 2007-07-26 at 10:40 -0700, Nathan Froyd wrote:
> The attached patch XFAILs gcc.target/i386/pr23098.c on x86 vxworks
> targets.  The testcase was written to ensure that loads of a "easy"
> floating-point constant (e.g. 0.0 or 1.0) get turned into the
> appropriate x86 instructions, rather than loading from memory.
> Unfortunately, the testcase fails because VxWorks takes a somewhat
> different approach to loading such constants, one that the
> deligitimize_address hook does not recognize.  Rewriting the x86 backend
> to help VxWorks out here was judged too hairy and therefore XFAILing the
> testcase seemed like the best way forward.
> 
> Tested on i586-wrs-vxworks.  OK to commit?

OK.

Janis