Fix CopyCode to properly record object pointer writes.

src/heap.cc

Index: src/heap.cc
diff --git a/src/heap.cc b/src/heap.cc
index 6a6afc022e54aff302e78a78abf2ecdaaf7dd423..57ab8796739445d31f48ca6434ac26c1cc289f89 100644
--- a/src/heap.cc
+++ b/src/heap.cc
@@ -3862,6 +3862,11 @@ MaybeObject* Heap::CopyCode(Code* code) {
   ASSERT(!isolate_->code_range()->exists() ||
       isolate_->code_range()->contains(code->address()));
   new_code->Relocate(new_addr - old_addr);
+
+  // Make sure new object is handled properly by the GC.
+  incremental_marking()->RecordWrites(new_code);

[Michael Starzinger, 2013/05/14 16:43:10]

So I looked into this write-barrier and I don't think we need one here. The
reasons for this are the following ...

1) All objects reachable from the code object header are pretenured and hence
cannot be in new-space. That's why the assertion in Heap::RecordWrites that the
address is not in code-space makes sense.

2) The newly allocated code object is guaranteed to still be white after we
copied over it's contents. So we couldn't have introduced black-to-white
pointers with the block copy.

3) The newly allocated code object will eventually be scanned by the marker
before a GC (it's still white) and hence all slots that point into evacuation
candidates will be discovered.

+  //  RecordWrites(new_code->address(), 0, Code::kHeaderSize);
+
   return new_code;
 }