obtained gcc-4.6.4.tar.bz2 from upstream website;upstream

verified gcc-4.6.4.tar.bz2.sig;
imported gcc-4.6.4 source tree from verified upstream tarball.

downloading a git-generated archive based on the 'upstream' tag
should provide you with a source tree that is binary identical
to the one extracted from the above tarball.

if you have obtained the source via the command 'git clone',
however, do note that line-endings of files in your working
directory might differ from line-endings of the respective
files in the upstream repository.

1 files changed, 78 insertions, 0 deletions

diff --git a/gcc/testsuite/gcc.dg/torture/builtin-convert-2.c b/gcc/testsuite/gcc.dg/torture/builtin-convert-2.c
new file mode 100644
index 000000000..167ecddcb
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/builtin-convert-2.c
@@ -0,0 +1,78 @@
+/* Copyright (C) 2004  Free Software Foundation.
+
+   Verify that built-in math function conversion into integer rounding
+   functions is correctly performed by the compiler.
+
+   Written by Kaveh ghazi, 2004-04-26.  */
+
+/* { dg-do link } */
+/* { dg-options "-ffast-math" } */
+/* { dg-add-options c99_runtime } */
+
+#include "../builtins-config.h"
+
+/* Macro to do all FP type combinations.  The second half tests
+   narrowing the FP type.  */
+#define TEST_FP2FIXED(FN1, FN2) \
+  extern void link_error_##FN1##_##FN2(void); \
+  extern void link_error_##FN1##f_##FN2##f(void); \
+  extern void link_error_##FN1##l_##FN2##l(void); \
+  extern void link_error_##FN1##_l##FN2(void); \
+  extern void link_error_##FN1##f_l##FN2##f(void); \
+  extern void link_error_##FN1##l_l##FN2##l(void); \
+  if ((long)__builtin_##FN1(d) != __builtin_##FN2(d)) \
+    link_error_##FN1##_##FN2(); \
+  if ((long)__builtin_##FN1##f(f) != __builtin_##FN2##f(f)) \
+    link_error_##FN1##f_##FN2##f(); \
+  if ((long)__builtin_##FN1##l(ld) != __builtin_##FN2##l(ld)) \
+    link_error_##FN1##l_##FN2##l(); \
+  if ((long long)__builtin_##FN1(d) != __builtin_l##FN2(d)) \
+    link_error_##FN1##_l##FN2(); \
+  if ((long long)__builtin_##FN1##f(f) != __builtin_l##FN2##f(f)) \
+    link_error_##FN1##f_l##FN2##f(); \
+  if ((long long)__builtin_##FN1##l(ld) != __builtin_l##FN2##l(ld)) \
+    link_error_##FN1##l_l##FN2##l(); \
+  extern void link_error_##FN1##_##FN2##f(void); \
+  extern void link_error_##FN1##l_##FN2(void); \
+  extern void link_error_##FN1##l_##FN2##f(void); \
+  extern void link_error_##FN1##_l##FN2##f(void); \
+  extern void link_error_##FN1##l_l##FN2(void); \
+  extern void link_error_##FN1##l_l##FN2##f(void); \
+  if (sizeof(double) > sizeof(float) \
+      && (long)__builtin_##FN1(f) != __builtin_##FN2##f(f)) \
+    link_error_##FN1##_##FN2##f(); \
+  if (sizeof(long double) > sizeof(double) \
+      && (long)__builtin_##FN1##l(d) != __builtin_##FN2(d)) \
+    link_error_##FN1##l_##FN2(); \
+  if (sizeof(long double) > sizeof(float) \
+      && (long)__builtin_##FN1##l(f) != __builtin_##FN2##f(f)) \
+    link_error_##FN1##l_##FN2##f(); \
+  if (sizeof(double) > sizeof(float) \
+      && (long long)__builtin_##FN1(f) != __builtin_l##FN2##f(f)) \
+    link_error_##FN1##_l##FN2##f(); \
+  if (sizeof(long double) > sizeof(double) \
+      && (long long)__builtin_##FN1##l(d) != __builtin_l##FN2(d)) \
+    link_error_##FN1##l_l##FN2(); \
+  if (sizeof(long double) > sizeof(float) \
+      && (long long)__builtin_##FN1##l(f) != __builtin_l##FN2##f(f)) \
+    link_error_##FN1##l_l##FN2##f()
+
+void __attribute__ ((__noinline__)) foo (double d, float f, long double ld)
+{
+#ifdef __OPTIMIZE__
+# ifdef HAVE_C99_RUNTIME
+  /* The resulting transformation functions are all C99.  */
+  TEST_FP2FIXED (ceil, lceil);
+  TEST_FP2FIXED (floor, lfloor);
+  TEST_FP2FIXED (round, lround);
+  TEST_FP2FIXED (nearbyint, lrint);
+  TEST_FP2FIXED (rint, lrint);
+# endif
+#endif
+}
+
+int main()
+{
+  foo (1.0, 2.0, 3.0);
+  return 0;
+}