[runtime function] Fix IndexOf when start is -Infinity

src/runtime/runtime-array.cc

Index: src/runtime/runtime-array.cc
diff --git a/src/runtime/runtime-array.cc b/src/runtime/runtime-array.cc
index 5538fb4c28afc84838552b8bdc36bc8f306bd12b..0ac965566c9ab46d2993b11e82e0ac2b6f3f1264 100644
--- a/src/runtime/runtime-array.cc
+++ b/src/runtime/runtime-array.cc
@@ -568,7 +568,15 @@ RUNTIME_FUNCTION(Runtime_ArrayIndexOf) {
                                        Object::ToInteger(isolate, from_index));
     double fp = from_index->Number();
     if (fp > len) return Smi::FromInt(-1);
-    start_from = static_cast<int64_t>(fp);
+    if (V8_LIKELY(
+            fp >= static_cast<double>(std::numeric_limits<int64_t>::min()) &&
+            fp < static_cast<double>(std::numeric_limits<int64_t>::max()))) {

[Jakob Kummerow, 2017/05/09 11:03:32]

You don't need the second condition; the "fp > len" check above guarantees that
fp <= std::numeric_limits<int64_t>::max().

[predrag.rudic, 2017/05/09 12:43:08]

Acknowledged.

+      start_from = static_cast<int64_t>(fp);
+    } else {
+      start_from = fp < static_cast<double>(std::numeric_limits<int64_t>::min())

[Jakob Kummerow, 2017/05/09 11:03:32]

...and then you don't need a case distinction here either; always assigning
std::numeric_limits<int64_t>::min() is fine.

[predrag.rudic, 2017/05/09 12:43:07]

Acknowledged.

+                       ? std::numeric_limits<int64_t>::min()
+                       : std::numeric_limits<int64_t>::max();
+    }
   }
 
   int64_t index;