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, 94 insertions, 0 deletions

diff --git a/gcc/testsuite/c-c++-common/dfp/convert-dfp.c b/gcc/testsuite/c-c++-common/dfp/convert-dfp.c
new file mode 100644
index 000000000..d52ba51b0
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/dfp/convert-dfp.c
@@ -0,0 +1,94 @@
+/* { dg-options "-O0" } */
+
+/* N1150 5.2 Conversions among decimal floating types and between
+   decimal floating types and generic floating types.
+   C99 6.3.1.5(3) New.
+
+   Test various conversions involving decimal floating types. */
+
+#ifndef	__STDC_WANT_DEC_FP__
+#define __STDC_WANT_DEC_FP__ 1
+#endif
+
+#include "dfp-dbg.h"
+#include <float.h>
+
+volatile _Decimal32 d32;
+volatile _Decimal64 d64;
+volatile _Decimal128 d128;
+
+int
+main ()
+{
+  /* Conversions to larger types.  */
+  d32 = 123.4df;
+  d64 = d32;
+  if (d64 != 123.4dd)
+    FAILURE
+  d128 = d32;
+  if (d128 != 123.4dl)
+    FAILURE
+  d64 = 345.678dd;
+  d128 = d64;
+  if (d128 != 345.678dl)
+    FAILURE
+
+  /* Conversions to smaller types for which the value fits.  */
+  d64 = 3456.789dd;
+  d32 = d64;
+  if (d32 != 3456.789df)
+    FAILURE
+  d128 = 123.4567dl;
+  d32 = d128;
+  if (d32 != 123.4567df)
+    FAILURE
+
+  d128 = 1234567890.123456dl;
+  d64 = d128;
+  if (d64 != 1234567890.123456dd)
+    FAILURE
+
+  /* Test demotion to non-representable decimal floating type. */
+
+  /* Assumes a default rounding mode of 'near'.  This uses the rules
+     describe in the 27 July 2005 draft of IEEE 754r, which are much
+     more clear that what's described in draft 5 of N1107.  */
+
+  /* Rounds to what _Decimal32 can handle.  */
+  d64 = 9.99999949E96DD;
+  d32 = d64;
+  if (d32 != DEC32_MAX)
+    FAILURE
+
+  /* Rounds to more than _Decimal32 can handle.  */
+  d64 = 9.9999995E96DD;
+  d32 = d64;
+  if (d32 != __builtin_infd32())
+    FAILURE
+
+  /* Rounds to what _Decimal32 can handle.  */
+  d128 = 9.99999949E96DD;
+  d32 = d128;
+  if (d32 != DEC32_MAX)
+    FAILURE
+
+  /* Rounds to more than _Decimal32 can handle.  */
+  d128= 9.9999995E96DD;
+  d32 = d128;
+  if (d32 != __builtin_infd32())
+    FAILURE
+
+  /* Rounds to what _Decimal64 can handle.  */
+  d128 = 9.99999999999999949E384DL;
+  d64 = d128;
+  if (d64 != DEC64_MAX)
+    FAILURE
+
+  /* Rounds to more than _Decimal64 can handle.  */
+  d128 = 9.9999999999999995E384DL;
+  d64 = d128;
+  if (d64 != __builtin_infd64())
+    FAILURE
+
+  FINISH
+}