Considering HTMLTokenizer state in HTMLSourceTracker

third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp

Index: third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp
diff --git a/third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp b/third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp
index aaec23e519e49a975365549964c783acefc9164b..9c9ee2400d272ad0ae797beeef76a49230f1e91f 100644
--- a/third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp
+++ b/third_party/WebKit/Source/core/html/parser/HTMLSourceTracker.cpp
@@ -37,7 +37,8 @@ void HTMLSourceTracker::start(SegmentedString& currentInput,
                               HTMLToken& token) {
   if (token.type() == HTMLToken::Uninitialized && !m_isStarted) {
     m_previousSource.clear();
-    if (tokenizer->numberOfBufferedCharacters())
+    if (needToCheckTokenizerBuffer(tokenizer) &&
+        tokenizer->numberOfBufferedCharacters())
       m_previousSource = tokenizer->bufferedCharacters();
   } else
     m_previousSource.append(m_currentSource);

[Charlie Harrison, 2017/03/07 18:37:48]

while you are here could you add braces to the else statement to fit our style
guide?

[wanchang, 2017/03/08 00:35:35]

Done.

@@ -56,8 +57,12 @@ void HTMLSourceTracker::end(SegmentedString& currentInput,
   m_cachedSourceForToken = String();
 
   // FIXME: This work should really be done by the HTMLTokenizer.
+  size_t numberOfBufferedCharacters = 0u;
+  if (needToCheckTokenizerBuffer(tokenizer)) {
+    numberOfBufferedCharacters = tokenizer->numberOfBufferedCharacters();
+  }
   token.end(currentInput.numberOfCharactersConsumed() -
-            tokenizer->numberOfBufferedCharacters());
+            numberOfBufferedCharacters);
 }
 
 String HTMLSourceTracker::sourceForToken(const HTMLToken& token) {
@@ -92,4 +97,11 @@ String HTMLSourceTracker::sourceForToken(const HTMLToken& token) {
   return m_cachedSourceForToken;
 }
 
+bool HTMLSourceTracker::needToCheckTokenizerBuffer(HTMLTokenizer* tokenizer) {
+  HTMLTokenizer::State state = tokenizer->getState();
+  // considering tokenizer temporarybuffer only if in the middle of

[Charlie Harrison, 2017/03/07 18:37:47]

Phrasing suggestion:
"Consider checking the tokenizer's temporary buffer if it is in an end tag
buffering state, or the token has finished parsing.

The temporary buffer must not be used unconditionally, because in some states
(e.g. ScriptDataDoubleEscapeStartState), data is appended to both the temporary
buffer and the token itself.
"

[wanchang, 2017/03/08 00:35:35]

Done.

+  // endtagbufferingstate or completed parsing the token
+  return state == HTMLTokenizer::DataState || isEndTagBufferingState(state);
+}
+
 }  // namespace blink