MQParser memory leak fix

Source/core/css/CSSParserValues.cpp

Index: Source/core/css/CSSParserValues.cpp
diff --git a/Source/core/css/CSSParserValues.cpp b/Source/core/css/CSSParserValues.cpp
index 484272de759de25a1b11febfb763ab2920198f03..72334ec78192bb7a841c4c5a6fa79cb5e7dc6504 100644
--- a/Source/core/css/CSSParserValues.cpp
+++ b/Source/core/css/CSSParserValues.cpp
@@ -29,17 +29,28 @@ namespace WebCore {
 
 using namespace WTF;
 
-CSSParserValueList::~CSSParserValueList()
+static void destroy(Vector<CSSParserValue, 4>& values)
 {
-    size_t numValues = m_values.size();
+    size_t numValues = values.size();
     for (size_t i = 0; i < numValues; i++) {
-        if (m_values[i].unit == CSSParserValue::Function)
-            delete m_values[i].function;
-        else if (m_values[i].unit == CSSParserValue::ValueList)
-            delete m_values[i].valueList;
+        if (values[i].unit == CSSParserValue::Function)
+            delete values[i].function;
+        else if (values[i].unit == CSSParserValue::ValueList)
+            delete values[i].valueList;
     }
 }
 
+void CSSParserValueList::destroyAndClear()
+{
+    destroy(m_values);
+    clear();
+}
+
+CSSParserValueList::~CSSParserValueList()
+{
+    destroy(m_values);

[eseidel, 2014/05/07 15:26:10]

Seems like this could end up destroying twice.  I guess the logic is that
destory i never called w/o clear except here?

Why don't we use a wrapper for CSSParserValue which knows how to destory itself?
 Or perhaps we should just give CSSParserValue a working destructor?

[Yoav Weiss, 2014/05/07 15:34:43]

Yeah, and calling destroyAndClear immediately followed by the destructor does
not damage, since m_values would be empty.
We probably could do that, but since CSSParserValueList has several clients, not
just my code, I'd need to further study what would be the implications of that.

+}
+
 void CSSParserValueList::addValue(const CSSParserValue& v)
 {
     m_values.append(v);