REGRESSION (STP): Video won't play on bilibili.com

https://bugs.webkit.org/show_bug.cgi?id=263196 rdar://117020123 Reviewed by Chris Dumez. bilibili injects DOM objects in the tree by calling: ``` var t = (new DOMParser).parseFromString(e, "text/html"); return document.adoptNode(t.body.firstChild) ``` Per spec, when parsing HTML to create a document through DOMParser::parseFromString() [1] you are to skip the whitespace tokens for the "initial" insertion mode [2]. The trailing spaces however must be kept [3] This behaviour was regressed by 267202@main. On this site, the observable behaviour was that the firstChild became a "#text" node rather the expected first DIV one. So we trim the leading whitespaces before calling the fast parser. Which restore the original behaviour prior 267202@main [1] https://html.spec.whatwg.org/multipage/dynamic-markup-insertion.html#parse-html-from-a-string [2] https://html.spec.whatwg.org/multipage/parsing.html#the-initial-insertion-mode [3] https://html.spec.whatwg.org/multipage/parsing.html#parsing-main-afterbody * LayoutTests/fast/dom/document-contentType-DOMParser-expected.txt: * LayoutTests/fast/dom/document-contentType-DOMParser.html: Added tests. * Source/WebCore/html/parser/HTMLDocumentParserFastPath.cpp: (WebCore::tryFastParsingHTMLFragment): * Source/WebCore/html/parser/HTMLDocumentParserFastPath.h: Change type to StringView (WebCore::requires): * Source/WebCore/xml/DOMParser.cpp: (WebCore::DOMParser::parseFromString): Canonical link: https://commits.webkit.org/269457@main
WebKit · Oct 18, 2023 · 55a2864 · 55a2864
1 parent ab95121
commit 55a2864
Show file tree

Hide file tree

Showing 5 changed files with 26 additions and 4 deletions.
diff --git a/LayoutTests/fast/dom/document-contentType-DOMParser-expected.txt b/LayoutTests/fast/dom/document-contentType-DOMParser-expected.txt
@@ -6,6 +6,13 @@ PASS new DOMParser().parseFromString(xslContent, "text/xsl").contentType threw e
 PASS new DOMParser().parseFromString(xmlContent, "text/dummy+xml").contentType threw exception TypeError: Type error.
 PASS new DOMParser().parseFromString(xmlContent, "text/XML").contentType threw exception TypeError: Type error.
 PASS new DOMParser().parseFromString(htmlContent, "TEXT/html").contentType threw exception TypeError: Type error.
+PASS parsedContent.body.firstChild.nodeName is "DIV"
+PASS parsedContent.body.childNodes.length is 2
+PASS parsedContent.body.childNodes[1].nodeName is "#text"
+PASS div.firstChild.nodeName is "#text"
+PASS div.childNodes.length is 3
+PASS div.childNodes[2].nodeName is "#text"
+PASS new DOMParser().parseFromString(htmlContentWithJustSpaces, "text/html").body.childNodes.length is 0
 PASS successfullyParsed is true
 
 TEST COMPLETE

diff --git a/LayoutTests/fast/dom/document-contentType-DOMParser.html b/LayoutTests/fast/dom/document-contentType-DOMParser.html
@@ -6,7 +6,7 @@
 </head>
 <body>
  <script>
-  
+
 var htmlContent =
                 "<html>" +
                     "<head>" +
@@ -80,6 +80,21 @@
 shouldThrow('new DOMParser().parseFromString(xmlContent, "text/XML").contentType', "'TypeError: Type error'");
 shouldThrow('new DOMParser().parseFromString(htmlContent, "TEXT/html").contentType', "'TypeError: Type error'");
 
+var htmlContentWithSpace = "\n        <div class=\"bpx-player-container\"></div>        \n";
+var parsedContent = new DOMParser().parseFromString(htmlContentWithSpace, "text/html");
+shouldBeEqualToString('parsedContent.body.firstChild.nodeName', 'DIV');
+shouldBe('parsedContent.body.childNodes.length', '2');
+shouldBeEqualToString('parsedContent.body.childNodes[1].nodeName', '#text');
+
+var div = document.createElement("div");
+div.innerHTML = htmlContentWithSpace;
+shouldBeEqualToString('div.firstChild.nodeName', '#text');
+shouldBe('div.childNodes.length', '3');
+shouldBeEqualToString('div.childNodes[2].nodeName', '#text');
+
+var htmlContentWithJustSpaces = "\n                \n";
+shouldBe('new DOMParser().parseFromString(htmlContentWithJustSpaces, "text/html").body.childNodes.length', '0');
+
 </script>
 <script src="../../resources/js-test-post.js"></script>
 </body>
diff --git a/Source/WebCore/html/parser/HTMLDocumentParserFastPath.cpp b/Source/WebCore/html/parser/HTMLDocumentParserFastPath.cpp
@@ -895,7 +895,7 @@ static bool tryFastParsingHTMLFragmentImpl(const std::span<const CharacterType>&
     return parser.parse(contextElement);
 }
 
-bool tryFastParsingHTMLFragment(const String& source, Document& document, ContainerNode& destinationParent, Element& contextElement, OptionSet<ParserContentPolicy> policy)
+bool tryFastParsingHTMLFragment(StringView source, Document& document, ContainerNode& destinationParent, Element& contextElement, OptionSet<ParserContentPolicy> policy)
 {
     if (!canUseFastPath(contextElement, policy))
         return false;

diff --git a/Source/WebCore/html/parser/HTMLDocumentParserFastPath.h b/Source/WebCore/html/parser/HTMLDocumentParserFastPath.h
@@ -41,7 +41,7 @@ class Element;
 
 enum class ParserContentPolicy : uint8_t;
 
-WEBCORE_EXPORT bool tryFastParsingHTMLFragment(const String& source, Document&, ContainerNode&, Element& contextElement, OptionSet<ParserContentPolicy>);
+WEBCORE_EXPORT bool tryFastParsingHTMLFragment(StringView source, Document&, ContainerNode&, Element& contextElement, OptionSet<ParserContentPolicy>);
 
 } // namespace WebCore
 
diff --git a/Source/WebCore/xml/DOMParser.cpp b/Source/WebCore/xml/DOMParser.cpp
@@ -60,7 +60,7 @@ ExceptionOr<Ref<Document>> DOMParser::parseFromString(const String& string, cons
     bool usedFastPath = false;
     if (contentType == "text/html"_s) {
         auto body = HTMLBodyElement::create(document);
-        usedFastPath = tryFastParsingHTMLFragment(string, document, body, body, parsingOptions);
+        usedFastPath = tryFastParsingHTMLFragment(StringView { string }.substring(string.find(isNotASCIIWhitespace<UChar>)), document, body, body, parsingOptions);
         if (LIKELY(usedFastPath)) {
             auto html = HTMLHtmlElement::create(document);
             document->appendChild(html);