Fix incorrect Math.pow() calculations with MinGW-w64

src/assembler.cc

Index: src/assembler.cc
===================================================================
--- src/assembler.cc	(revision 11248)
+++ src/assembler.cc	(working copy)
@@ -1153,6 +1153,22 @@
 
 
 double power_double_double(double x, double y) {
+#ifdef __MINGW64_VERSION_MAJOR
+  if ((x == 0.0 || isinf(x)) && isfinite(y)) {
+    double f;
+
+    if (modf(y, &f) != 0.0)
+      return (x == 0.0) ^ (y > 0) ? V8_INFINITY : 0;
+  }
+
+  if (x == 2.0) {
+    int y_int = static_cast<int>(y);
+
+    if (y == y_int)
+      return ldexp(1.0, y);

[alexeif, 2012/04/17 16:40:13]

nit: y_int

+  }
+#endif
+
   // The checks for special cases can be dropped in ia32 because it has already
   // been done in generated code before bailing out here.
   if (isnan(y) || ((x == 1 || x == -1) && isinf(y))) return OS::nan_value();