Changeset 62028 in webkit

Timestamp:

Jun 28, 2010 1:05:25 PM (14 years ago)

Author:

abarth@webkit.org

Message:

2010-06-28 Adam Barth <​abarth@webkit.org>

Reviewed by Eric Seidel.

The new tree builder needs to call attach() on elements it attaches to
the DOM
​https://bugs.webkit.org/show_bug.cgi?id=41293

Update expected results to show that the contents of the iframe
actually get created. This only gets us a few lines down the test
harness before erroring out again, but I'll take it.

• html5lib/runner-expected-html5.txt:
• html5lib/webkit-resumer-expected-html5.txt:

2010-06-28 Adam Barth <​abarth@webkit.org>

Reviewed by Eric Seidel.

The new tree builder needs to call attach() on elements it attaches to
the DOM
​https://bugs.webkit.org/show_bug.cgi?id=41293

Apparently Nodes expect to have their attach() method called when they
are attached to the DOM. The new tree builder is happy to oblige.
Making this call requires some fancy footwork with RefPtr/PassRefPtr to
avoid extra ref churn while keeping each function small.

• html/HTMLTreeBuilder.cpp: (WebCore::HTMLTreeBuilder::insertHTMLStartTagBeforeHTML): (WebCore::HTMLTreeBuilder::processCharacter): (WebCore::HTMLTreeBuilder::insertDoctype): (WebCore::HTMLTreeBuilder::insertComment): (WebCore::HTMLTreeBuilder::insertCommentOnDocument): (WebCore::HTMLTreeBuilder::insertElement): (WebCore::HTMLTreeBuilder::insertSelfClosingElement): (WebCore::HTMLTreeBuilder::insertScriptElement):
• html/HTMLTreeBuilder.h: (WebCore::HTMLTreeBuilder::attach):

Location:

trunk

Files:

• LayoutTests/ChangeLog (modified) (1 diff)
• LayoutTests/html5lib/runner-expected-html5.txt (modified) (1 diff)
• LayoutTests/html5lib/webkit-resumer-expected-html5.txt (modified) (1 diff)
• WebCore/ChangeLog (modified) (1 diff)
• WebCore/html/HTMLTreeBuilder.cpp (modified) (8 diffs)
• WebCore/html/HTMLTreeBuilder.h (modified) (1 diff)

trunk/LayoutTests/ChangeLog

@@ +1 @@
+2010-06-28  Adam Barth  <abarth@webkit.org>
+
+        Reviewed by Eric Seidel.
+
+        The new tree builder needs to call attach() on elements it attaches to
+        the DOM
+        https://bugs.webkit.org/show_bug.cgi?id=41293
+
+        Update expected results to show that the contents of the iframe
+        actually get created.  This only gets us a few lines down the test
+        harness before erroring out again, but I'll take it.
+
+        * html5lib/runner-expected-html5.txt:
+        * html5lib/webkit-resumer-expected-html5.txt:
+
 2010-06-28  Robert Hogan  <robert@webkit.org>
 

trunk/LayoutTests/html5lib/runner-expected-html5.txt

@@ -1 @@
-CONSOLE MESSAGE: line 33: TypeError: Result of expression 'iframe.contentWindow' [null] is not an object.
+CONSOLE MESSAGE: line 36: TypeError: Result of expression 'iframe.contentWindow.document.lastChild' [null] is not an object.
 Content-Type: text/plain
 

trunk/LayoutTests/html5lib/webkit-resumer-expected-html5.txt

@@ -1 @@
-CONSOLE MESSAGE: line 33: TypeError: Result of expression 'iframe.contentWindow' [null] is not an object.
+CONSOLE MESSAGE: line 36: TypeError: Result of expression 'iframe.contentWindow.document.lastChild' [null] is not an object.
 Content-Type: text/plain
 

trunk/WebCore/ChangeLog

@@ +1 @@
+2010-06-28  Adam Barth  <abarth@webkit.org>
+
+        Reviewed by Eric Seidel.
+
+        The new tree builder needs to call attach() on elements it attaches to
+        the DOM
+        https://bugs.webkit.org/show_bug.cgi?id=41293
+
+        Apparently Nodes expect to have their attach() method called when they
+        are attached to the DOM.  The new tree builder is happy to oblige.
+        Making this call requires some fancy footwork with RefPtr/PassRefPtr to
+        avoid extra ref churn while keeping each function small.
+
+        * html/HTMLTreeBuilder.cpp:
+        (WebCore::HTMLTreeBuilder::insertHTMLStartTagBeforeHTML):
+        (WebCore::HTMLTreeBuilder::processCharacter):
+        (WebCore::HTMLTreeBuilder::insertDoctype):
+        (WebCore::HTMLTreeBuilder::insertComment):
+        (WebCore::HTMLTreeBuilder::insertCommentOnDocument):
+        (WebCore::HTMLTreeBuilder::insertElement):
+        (WebCore::HTMLTreeBuilder::insertSelfClosingElement):
+        (WebCore::HTMLTreeBuilder::insertScriptElement):
+        * html/HTMLTreeBuilder.h:
+        (WebCore::HTMLTreeBuilder::attach):
+
 2010-06-28  Xan Lopez  <xlopez@igalia.com>
 

trunk/WebCore/html/HTMLTreeBuilder.cpp

@@ -282 +282 @@
     RefPtr<Element> element = HTMLHtmlElement::create(m_document);
     element->setAttributeMap(token.attributes(), m_fragmentScriptingPermission);
-    m_document->addChild(element);
-    m_openElements.push(element.release());
+    m_openElements.push(attach(m_document, element.release()));
 }
 
@@ -667 +666 @@
 {
     if (insertionMode() == TextMode) {
-        currentElement()->addChild(Text::create(m_document, token.characters()));
+        attach(currentElement(), Text::create(m_document, token.characters()));
         return;
     }
@@ -793 +792 @@
 {
     ASSERT(token.type() == HTMLToken::DOCTYPE);
-    m_document->addChild(DocumentType::create(m_document, token.name(), String::adopt(token.publicIdentifier()), String::adopt(token.systemIdentifier())));
+    attach(m_document, DocumentType::create(m_document, token.name(), String::adopt(token.publicIdentifier()), String::adopt(token.systemIdentifier())));
     // FIXME: Move quirks mode detection from DocumentType element to here.
     notImplemented();
@@ -803 +802 @@
 {
     ASSERT(token.type() == HTMLToken::Comment);
-    currentElement()->addChild(Comment::create(m_document, token.comment()));
+    attach(currentElement(), Comment::create(m_document, token.comment()));
 }
 
@@ -809 +808 @@
 {
     ASSERT(token.type() == HTMLToken::Comment);
-    m_document->addChild(Comment::create(m_document, token.comment()));
+    attach(m_document, Comment::create(m_document, token.comment()));
 }
 
@@ -815 +814 @@
 {
     ASSERT(token.type() == HTMLToken::StartTag);
-    RefPtr<Element> element = createElement(token);
-    currentElement()->addChild(element);
-    m_openElements.push(element.release());
+    m_openElements.push(attach(currentElement(), createElement(token)));
 }
 
@@ -823 +820 @@
 {
     ASSERT(token.type() == HTMLToken::StartTag);
-    currentElement()->addChild(createElement(token));
+    attach(currentElement(), createElement(token));
     // FIXME: Do we want to acknowledge the token's self-closing flag?
     // http://www.whatwg.org/specs/web-apps/current-work/multipage/tokenization.html#acknowledge-self-closing-flag
@@ -857 +854 @@
     RefPtr<HTMLScriptElement> element = HTMLScriptElement::create(scriptTag, m_document, true);
     element->setAttributeMap(token.attributes(), m_fragmentScriptingPermission);
-    currentElement()->addChild(element);
-    m_openElements.push(element.release());
+    m_openElements.push(attach(currentElement(), element.release()));
     m_tokenizer->setState(HTMLTokenizer::ScriptDataState);
     m_originalInsertionMode = m_insertionMode;

trunk/WebCore/html/HTMLTreeBuilder.h

@@ -178 +178 @@
     bool processStartTagForInHead(AtomicHTMLToken&);
 
+    template<typename ChildType>
+    PassRefPtr<ChildType> attach(Node* parent, PassRefPtr<ChildType> prpChild)
+    {
+        RefPtr<ChildType> child = prpChild;
+        parent->addChild(child);
+        // It's slightly unfortunate that we need to hold a reference to child
+        // here to call attach().  We should investigate whether we can rely on
+        // |parent| to hold a ref at this point.  In the common case (at least
+        // for elements), however, we'll get to use this ref in the stack of
+        // open elements.
+        child->attach();
+        return child.release();
+    };
+
     void insertDoctype(AtomicHTMLToken&);
     void insertComment(AtomicHTMLToken&);