Replace raw pointers to GC allocated objects by members in stack allocated objects.

Source/core/css/resolver/StyleResolverState.cpp

Index: Source/core/css/resolver/StyleResolverState.cpp
diff --git a/Source/core/css/resolver/StyleResolverState.cpp b/Source/core/css/resolver/StyleResolverState.cpp
index f7299230631d84a0ab562eca3f1db4c6dc1e6e5f..96fa5e1a29355c0ad4cf0f40186b875584970e8b 100644
--- a/Source/core/css/resolver/StyleResolverState.cpp
+++ b/Source/core/css/resolver/StyleResolverState.cpp
@@ -37,9 +37,9 @@ StyleResolverState::StyleResolverState(Document& document, Element* element, Ren
     , m_parentStyle(parentStyle)
     , m_applyPropertyToRegularStyle(true)
     , m_applyPropertyToVisitedLinkStyle(false)
-    , m_lineHeightValue(0)
+    , m_lineHeightValue(nullptr)
     , m_styleMap(*this, m_elementStyleResources)
-    , m_currentRule(0)
+    , m_currentRule(nullptr)

[haraken, 2014/03/24 07:05:03]

Can we detect an error if we forget to initialize a Member in a stack-allocated
object with nullptr ?

[Mads Ager (chromium), 2014/03/24 07:24:43]

Drive-by answer: Member initializes itself with zero in any case, so we do not
have to initialize them with nullptr. The thing that becomes dangerous is
RawPtr. In the end those will become raw pointers that we have to properly
initialize.

Come to think of it, maybe we should initialize RawPtr with a recognizable zap
value instead of 0. They represent raw pointers and therefore they should always
be explicitly initialized to zero if that is what we want. Otherwise, we are in
for surprises when we move from: RawPtr<A> a; to A* a;

[haraken, 2014/03/24 07:36:21]

This sounds like a good idea.

 {
     if (!parentStyle && m_elementContext.parentNode())
         m_parentStyle = m_elementContext.parentNode()->renderStyle();