This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[testsuite] fix gcc.dg/torture/builtin-modf-1.c

The test modified by this patch started failing on powerpc*-linux,
powerpc-*-darwin*, and powerpc-*-aix* when -funsafe-math-optimizations
began to imply -fno-signed-zeros.  The test fails with earlier versions
of GCC when -fno-signed-zeros is used explicitly.  Furthermore, the
option -funsafe-math-optimizations wasn't really needed on

This patch removes the dg-options directive for
-funsafe-math-optimiztions, which should make the test pass again for
powerpc targets.  If it turns out to still be needed for mips-*-irix6
it can be added back for that target.  I plan to check this in
tomorrow if there are no objections.

Tested on powerpc64-linux with -m32/-m64.

2007-12-18  Janis Johnson  <>

	* gcc.dg/torture/builtin-modf-1.c: Remove -funsafe-math-optimizations.

Index: gcc/testsuite/gcc.dg/torture/builtin-modf-1.c
--- gcc/testsuite/gcc.dg/torture/builtin-modf-1.c	(revision 131004)
+++ gcc/testsuite/gcc.dg/torture/builtin-modf-1.c	(working copy)
@@ -7,9 +7,6 @@
 /* { dg-do link } */
 /* { dg-options "-fno-finite-math-only" { target sh*-*-* } } */
-/* In order to fold algebraic exprs below, targets with "composite"
-   floating point formats need -funsafe-math-optimizations.  */
-/* { dg-options "-funsafe-math-optimizations" { target mips*-*-irix6*
powerpc*-*-* } } */
 extern void link_error(int);

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]