Revert "Revert of [oilpan]: Change marking to do precise roots first and conservative second. (http…

Source/platform/heap/HeapTest.cpp

Index: Source/platform/heap/HeapTest.cpp
diff --git a/Source/platform/heap/HeapTest.cpp b/Source/platform/heap/HeapTest.cpp
index 1309cddf5f5ba9aba34d6124711fd8468dcc55d5..679b55c023f9887efc5ce9f13edcb9ce89a3664e 100644
--- a/Source/platform/heap/HeapTest.cpp
+++ b/Source/platform/heap/HeapTest.cpp
@@ -4660,8 +4660,6 @@ public:
             // shutting it down.
             stackPtrValue = reinterpret_cast<uintptr_t>(cto.get());
         }
-        RELEASE_ASSERT(stackPtrValue);
-
         // At this point it is "programatically" okay to shut down the worker thread
         // since the cto object should be dead. However out stackPtrValue will cause a
         // trace of the object when doing a conservative GC.
@@ -4685,6 +4683,13 @@ public:
         EXPECT_EQ(0, CrossThreadObject::s_destructorCalls);
         EXPECT_EQ(1, IntWrapper::s_destructorCalls);
 
+        // This release assert is here to ensure the stackValuePtr is not
+        // optimized away before doing the above conservative GC. If the
+        // EXPECT_EQ(0, CrossThreadObject::s_destructorCalls) call above
+        // starts failing it means we have to find a better way to ensure
+        // the stackPtrValue is not optimized away.
+        RELEASE_ASSERT(stackPtrValue);
+
         // Do a GC with no pointers on the stack to see the cto being collected.
         Heap::collectGarbage(ThreadState::NoHeapPointersOnStack);
         EXPECT_EQ(1, CrossThreadObject::s_destructorCalls);