Title: [139497] trunk/Source/WebCore
- Revision
- 139497
- Author
- to...@chromium.org
- Date
- 2013-01-11 14:20:13 -0800 (Fri, 11 Jan 2013)
Log Message
We should be able to checkpoint and restore the HTMLTokenizer across threads
https://bugs.webkit.org/show_bug.cgi?id=106597
Based on patch by Adam Barth.
This has the ability to create a checkpoint any time the parser is blocked on a script.
We clear m_appropriateEndTagName after each end tag is flushed so that the ASSERT in
canCreateCheckpoint() will pass.
Reviewed by Adam Barth.
No new tests because no new functionality.
* html/parser/HTMLDocumentParser.cpp:
(WebCore::HTMLDocumentParser::HTMLDocumentParser):
(WebCore::HTMLDocumentParser::pumpTokenizer):
* html/parser/HTMLDocumentParser.h:
(WebCore):
(HTMLDocumentParser):
* html/parser/HTMLTokenizer.cpp:
(WebCore):
(WebCore::HTMLTokenizer::canCreateCheckpoint):
(WebCore::HTMLTokenizer::createCheckpoint):
(WebCore::HTMLTokenizer::restoreFromCheckpoint):
* html/parser/HTMLTokenizer.h:
(HTMLTokenizer):
(Checkpoint):
(WebCore::HTMLTokenizer::Checkpoint::Checkpoint):
* xml/parser/MarkupTokenizerBase.h:
(WebCore::MarkupTokenizerBase::InputStreamPreprocessor::InputStreamPreprocessor):
(WebCore::MarkupTokenizerBase::InputStreamPreprocessor::skipNextNewLine):
(InputStreamPreprocessor):
(WebCore::MarkupTokenizerBase::InputStreamPreprocessor::reset):
Modified Paths
Diff
Modified: trunk/Source/WebCore/ChangeLog (139496 => 139497)
--- trunk/Source/WebCore/ChangeLog 2013-01-11 22:18:27 UTC (rev 139496)
+++ trunk/Source/WebCore/ChangeLog 2013-01-11 22:20:13 UTC (rev 139497)
@@ -1,3 +1,39 @@
+2013-01-11 Tony Gentilcore <to...@chromium.org>
+
+ We should be able to checkpoint and restore the HTMLTokenizer across threads
+ https://bugs.webkit.org/show_bug.cgi?id=106597
+
+ Based on patch by Adam Barth.
+
+ This has the ability to create a checkpoint any time the parser is blocked on a script.
+ We clear m_appropriateEndTagName after each end tag is flushed so that the ASSERT in
+ canCreateCheckpoint() will pass.
+
+ Reviewed by Adam Barth.
+
+ No new tests because no new functionality.
+
+ * html/parser/HTMLDocumentParser.cpp:
+ (WebCore::HTMLDocumentParser::HTMLDocumentParser):
+ (WebCore::HTMLDocumentParser::pumpTokenizer):
+ * html/parser/HTMLDocumentParser.h:
+ (WebCore):
+ (HTMLDocumentParser):
+ * html/parser/HTMLTokenizer.cpp:
+ (WebCore):
+ (WebCore::HTMLTokenizer::canCreateCheckpoint):
+ (WebCore::HTMLTokenizer::createCheckpoint):
+ (WebCore::HTMLTokenizer::restoreFromCheckpoint):
+ * html/parser/HTMLTokenizer.h:
+ (HTMLTokenizer):
+ (Checkpoint):
+ (WebCore::HTMLTokenizer::Checkpoint::Checkpoint):
+ * xml/parser/MarkupTokenizerBase.h:
+ (WebCore::MarkupTokenizerBase::InputStreamPreprocessor::InputStreamPreprocessor):
+ (WebCore::MarkupTokenizerBase::InputStreamPreprocessor::skipNextNewLine):
+ (InputStreamPreprocessor):
+ (WebCore::MarkupTokenizerBase::InputStreamPreprocessor::reset):
+
2013-01-11 Tony Chang <t...@chromium.org>
[chromium] Don't regenerate all bindings when any idl file changes
Modified: trunk/Source/WebCore/html/parser/HTMLTokenizer.cpp (139496 => 139497)
--- trunk/Source/WebCore/html/parser/HTMLTokenizer.cpp 2013-01-11 22:18:27 UTC (rev 139496)
+++ trunk/Source/WebCore/html/parser/HTMLTokenizer.cpp 2013-01-11 22:20:13 UTC (rev 139497)
@@ -136,6 +136,41 @@
m_additionalAllowedCharacter = '\0';
}
+#if ENABLE(THREADED_HTML_PARSER)
+
+bool HTMLTokenizer::canCreateCheckpoint() const
+{
+ if (!m_appropriateEndTagName.isEmpty())
+ return false;
+ if (!m_temporaryBuffer.isEmpty())
+ return false;
+ if (!m_bufferedEndTagName.isEmpty())
+ return false;
+ return true;
+}
+
+void HTMLTokenizer::createCheckpoint(Checkpoint& result) const
+{
+ ASSERT(canCreateCheckpoint());
+ result.options = m_options;
+ result.state = m_state;
+ result.additionalAllowedCharacter = m_additionalAllowedCharacter;
+ result.skipNextNewLine = m_inputStreamPreprocessor.skipNextNewLine();
+ result.shouldAllowCDATA = m_shouldAllowCDATA;
+}
+
+void HTMLTokenizer::restoreFromCheckpoint(const Checkpoint& checkpoint)
+{
+ m_token = 0;
+ m_options = checkpoint.options;
+ m_state = checkpoint.state;
+ m_additionalAllowedCharacter = checkpoint.additionalAllowedCharacter;
+ m_inputStreamPreprocessor.reset(checkpoint.skipNextNewLine);
+ m_shouldAllowCDATA = checkpoint.shouldAllowCDATA;
+}
+
+#endif
+
inline bool HTMLTokenizer::processEntity(SegmentedString& source)
{
bool notEnoughCharacters = false;
@@ -161,6 +196,7 @@
return true;
m_token->beginEndTag(m_bufferedEndTagName);
m_bufferedEndTagName.clear();
+ m_appropriateEndTagName.clear();
m_temporaryBuffer.clear();
return false;
}
@@ -196,6 +232,7 @@
// We started an end tag during our last iteration.
m_token->beginEndTag(m_bufferedEndTagName);
m_bufferedEndTagName.clear();
+ m_appropriateEndTagName.clear();
m_temporaryBuffer.clear();
if (m_state == HTMLTokenizerState::DataState) {
// We're back in the data state, so we must be done with the tag.
@@ -317,9 +354,11 @@
HTML_BEGIN_STATE(EndTagOpenState) {
if (isASCIIUpper(cc)) {
m_token->beginEndTag(static_cast<LChar>(toLowerCase(cc)));
+ m_appropriateEndTagName.clear();
HTML_ADVANCE_TO(TagNameState);
} else if (isASCIILower(cc)) {
m_token->beginEndTag(static_cast<LChar>(cc));
+ m_appropriateEndTagName.clear();
HTML_ADVANCE_TO(TagNameState);
} else if (cc == '>') {
parseError();
Modified: trunk/Source/WebCore/html/parser/HTMLTokenizer.h (139496 => 139497)
--- trunk/Source/WebCore/html/parser/HTMLTokenizer.h 2013-01-11 22:18:27 UTC (rev 139496)
+++ trunk/Source/WebCore/html/parser/HTMLTokenizer.h 2013-01-11 22:20:13 UTC (rev 139497)
@@ -124,6 +124,33 @@
void reset();
+#if ENABLE(THREADED_HTML_PARSER)
+
+ struct Checkpoint {
+ HTMLParserOptions options;
+ HTMLTokenizerState::State state;
+ UChar additionalAllowedCharacter;
+ bool skipNextNewLine;
+ bool forceNullCharacterReplacement;
+ bool shouldAllowCDATA;
+
+ Checkpoint()
+ : options(0)
+ , state()
+ , additionalAllowedCharacter('\0')
+ , skipNextNewLine(false)
+ , forceNullCharacterReplacement(false)
+ , shouldAllowCDATA(false)
+ {
+ }
+ };
+
+ bool canCreateCheckpoint() const;
+ void createCheckpoint(Checkpoint&) const;
+ void restoreFromCheckpoint(const Checkpoint&);
+
+#endif
+
// This function returns true if it emits a token. Otherwise, callers
// must provide the same (in progress) token on the next call (unless
// they call reset() first).
Modified: trunk/Source/WebCore/xml/parser/MarkupTokenizerBase.h (139496 => 139497)
--- trunk/Source/WebCore/xml/parser/MarkupTokenizerBase.h 2013-01-11 22:18:27 UTC (rev 139496)
+++ trunk/Source/WebCore/xml/parser/MarkupTokenizerBase.h 2013-01-11 22:20:13 UTC (rev 139497)
@@ -61,9 +61,8 @@
public:
InputStreamPreprocessor(MarkupTokenizerBase<Token, State>* tokenizer)
: m_tokenizer(tokenizer)
- , m_nextInputCharacter('\0')
- , m_skipNextNewLine(false)
{
+ reset();
}
ALWAYS_INLINE UChar nextInputCharacter() const { return m_nextInputCharacter; }
@@ -124,6 +123,14 @@
return peek(source);
}
+ bool skipNextNewLine() const { return m_skipNextNewLine; }
+
+ void reset(bool skipNextNewLine = false)
+ {
+ m_nextInputCharacter = '\0';
+ m_skipNextNewLine = skipNextNewLine;
+ }
+
static const UChar endOfFileMarker = 0;
private:
_______________________________________________
webkit-changes mailing list
webkit-changes@lists.webkit.org
http://lists.webkit.org/mailman/listinfo/webkit-changes
