Fix liveedit heap traversal bug with write barrier

src/liveedit.cc

Index: src/liveedit.cc
diff --git a/src/liveedit.cc b/src/liveedit.cc
index 7f6455443ef5a506ea2f665b84230782e8df80ee..8f82cfc81e633269d62efc15bf257457828e5259 100644
--- a/src/liveedit.cc
+++ b/src/liveedit.cc
@@ -968,6 +968,12 @@ class ReplacingVisitor : public ObjectVisitor {
 static void ReplaceCodeObject(Code* original, Code* substitution) {
   ASSERT(!HEAP->InNewSpace(substitution));
 
+  // Perform a full GC in order to avoid dead objects and to
+  // notify the GC with a write barrier so that the it can keep track of where
+  // there are pointers in case some of the objects move in a compacting GC.

[Erik Corry, 2012/06/25 11:43:51]

I think this comment should read:

// Perform a full GC in order to ensure that we are not in the middle of an
incremental marking phase when we are replacing the code object.  Since we are
not in an incremental marking phase we can write pointers to code objects (that
are never in new space) without worrying about write barriers.

[Peter Rybin, 2012/06/25 21:43:42]

Thank you!

+  HEAP->CollectAllGarbage(Heap::kMakeHeapIterableMask,
+                          "liveedit.cc ReplaceCodeObject");
+
   AssertNoAllocation no_allocations_please;
 
   ReplacingVisitor visitor(original, substitution);