Bug fix: Add BlinkGC allocation hooks to CSS, Node and vector/table backing

third_party/WebKit/Source/core/css/CSSValue.h

Index: third_party/WebKit/Source/core/css/CSSValue.h
diff --git a/third_party/WebKit/Source/core/css/CSSValue.h b/third_party/WebKit/Source/core/css/CSSValue.h
index e5d7588113b10b3b19a61b753c0af83bc8de975b..041907d450887c5e13e681bd7e3ecc4f20555f61 100644
--- a/third_party/WebKit/Source/core/css/CSSValue.h
+++ b/third_party/WebKit/Source/core/css/CSSValue.h
@@ -44,7 +44,8 @@ public:
     static void* allocateObject(size_t size, bool isEager)
     {
         ThreadState* state = ThreadStateFor<ThreadingTrait<CSSValue>::Affinity>::state();
-        return Heap::allocateOnHeapIndex(state, size, isEager ? BlinkGC::EagerSweepHeapIndex : BlinkGC::CSSValueHeapIndex, GCInfoTrait<CSSValue>::index());
+        const char* typeName = WTF_HEAP_PROFILER_TYPE_NAME(CSSValue);

[Primiano Tucci (use gerrit), 2016/02/29 17:23:14]

Considering that IIUC this is a pretty heavy user and that the string "CSSValue"
is just 9 bytes, should this be just
const char typeName[] = "CSSValue", so we get this also in official builds?
You don't really need the <T> expansion here, right?

[hajimehoshi, 2016/03/03 10:12:31]

Right. It'd be best if we could get children class names of CSSValue and in this
case we would need <T> expansion, but I'm fine with using the string literal so
far.

BTW, "blink::CSSValue" is the more accurate name?

Done.

+        return Heap::allocateOnHeapIndex(state, size, isEager ? BlinkGC::EagerSweepHeapIndex : BlinkGC::CSSValueHeapIndex, GCInfoTrait<CSSValue>::index(), typeName);
     }
 #else
     USING_FAST_MALLOC_WITH_TYPE_NAME(blink::CSSValue);