Harden NumberToSize against overflows.

src/v8conversions.h

Index: src/v8conversions.h
diff --git a/src/v8conversions.h b/src/v8conversions.h
index 3a7b5242ab79af659b15e99b0a6d53f967cd91d1..f3d0992d62b2ccb030b54f260cce604d0ec20ec5 100644
--- a/src/v8conversions.h
+++ b/src/v8conversions.h
@@ -60,10 +60,15 @@ inline size_t NumberToSize(Isolate* isolate,
                            Object* number) {
   SealHandleScope shs(isolate);
   if (number->IsSmi()) {
+    int value = Smi::cast(number)->value();
+    CHECK(value >= 0

[Jakob Kummerow, 2013/11/14 10:17:48]

Not that it makes a big difference, but it's sufficient to:

CHECK(value >= 0);
ASSERT(static_cast<unsigned>(value) <= std::numeric_limits<size_t>::max());

or even:

STATIC_ASSERT(Smi::kMaxValue <= std::numeric_limits<size_t>::max());

if that numeric_limits call is static assert capable.

[Dmitry Lomov (no reviews), 2013/11/14 10:22:58]

Great suggestion, this is by far the commonest case so shaving a couple cycles
here is useful.

+        && static_cast<unsigned>(value) <= std::numeric_limits<size_t>::max());
     return Smi::cast(number)->value();
   } else {
     ASSERT(number->IsHeapNumber());
     double value = HeapNumber::cast(number)->value();
+    CHECK(value >= 0 &&
+        value <= std::numeric_limits<size_t>::max());

[Jakob Kummerow, 2013/11/14 10:17:48]

nit: I'd align "value" with "value".

     return static_cast<size_t>(value);
   }
 }