Fix 23563: double unary- operator unstable for NANs

runtime/lib/double.cc

Index: runtime/lib/double.cc
diff --git a/runtime/lib/double.cc b/runtime/lib/double.cc
index 4173c1403d034ad917f12dea13d7f23cd2fc1f4b..8250d25f594d0f15da096fe544b22eb59d3e2951 100644
--- a/runtime/lib/double.cc
+++ b/runtime/lib/double.cc
@@ -300,6 +300,16 @@ DEFINE_NATIVE_ENTRY(Double_getIsNegative, 1) {
   return Bool::Get(signbit(dval) && !isnan(dval)).raw();
 }
 
+
+DEFINE_NATIVE_ENTRY(Double_flipSignBit, 1) {
+  const Double& arg = Double::CheckedHandle(arguments->NativeArgAt(0));
+  const double in_val = arg.value();
+  const int64_t bits =
+      *(reinterpret_cast<const int64_t*>(&in_val)) ^ 0x8000000000000000LL;

[koda, 2015/06/02 23:41:02]

Consider putting the constant in globals.h. Also, do all compilers accept this
syntax for a 64-bit literal?

[koda, 2015/06/02 23:41:02]

Just curious: I wonder if there is any spec/de-facto difference between doing
reinterpret_cast on pointers to int64/double vs. doing it directly on the
values? If either works, the latter seems simpler, but I don't feel strongly
about it.

[srdjan, 2015/06/03 00:02:51]

AFAIK yes, we are using it in simd28.cc and assembler_{ia32,x64}.cc. I think it
is part of standard.
http://gcc.gnu.org/onlinedocs/gcc/Long-Long.html

Just in case, added to globals.h:
const int64_t kSignBitDouble = DART_INT64_C(0x8000000000000000);

[srdjan, 2015/06/03 00:02:51]

As far as I remember, I had problems with doing it on values (some compilers
complained) and had to do it on pointers instead.

+  const double out_val = *(reinterpret_cast<const double*>(&bits));
+  return Double::New(out_val);
+}
+
 // Add here only functions using/referring to old-style casts.
 
 }  // namespace dart