Updated the check for unmodfied objects to handle Smi Objects.

src/heap/heap.cc

Index: src/heap/heap.cc
diff --git a/src/heap/heap.cc b/src/heap/heap.cc
index 1105c6868dc4699e0c0ffb9c057d7a6c16775112..81271d44a8be1f40c0ffc1c45ee35bcf353805b8 100644
--- a/src/heap/heap.cc
+++ b/src/heap/heap.cc
@@ -1493,6 +1493,7 @@ static bool IsUnscavengedHeapObject(Heap* heap, Object** p) {
 
 static bool IsUnmodifiedHeapObject(Object** p) {
   Object* object = *p;
+  if (object->IsSmi()) return true;

[mythria, 2015/12/02 09:45:49]

I thought we can collect all Smi objects. If you think it is more expensive to
collect them and recreate them when necessary, I will just return false. In that
case they will not be collected by minor GC.

[mythria, 2015/12/02 09:45:49]

I am a little surprised that we have Smi objects that are marked weak in new
space nodes. Is it expected? This is the code which generates a Weak Smi object:

Promise.reject(42); from the test
inspector/console/console-uncaught-promise.html in LayoutTests

[jochen (gone - plz use gerrit), 2015/12/02 09:48:28]

actually, I think we should return false for all objects that weren't created
via the C++ API.

While it's possible to recreate e.g. 42, blink has no way of knowing that it
should recreate 42

[mythria, 2015/12/02 12:32:38]

Done.

   DCHECK(object->IsHeapObject());

[mythria, 2015/12/02 09:45:49]

I think we could remove this DCHECK right? or is it safer to have it?

[jochen (gone - plz use gerrit), 2015/12/02 09:48:28]

you can remove it

[mythria, 2015/12/02 12:32:38]

Done.

   HeapObject* heap_object = HeapObject::cast(object);
   if (!object->IsJSObject()) return false;