Reduce the perf overhead of our dangling markup experiments.

Given that Document::completeURL is a hotspot, this patch revamps the
approach to capture removed whitespace up in //url (where we're already
scanning through the string), and limits the scan for '<' to cases in
which we've actually removed whitespace.

BUG=680970, 682300
R=jochen@chromium.org

Review-Url: https://codereview.chromium.org/2643613002
Cr-Commit-Position: refs/heads/master@{#444734}

third_party/WebKit/Source/core/dom/Document.cpp

@@ -4945,34 +4945,27 @@ void Document::setEncodingData(const DocumentEncodingData& newData) {
 }
 
 KURL Document::completeURL(const String& url) const {
-  String trimmed = url.stripWhiteSpace();
   KURL completed = completeURLWithOverride(url, m_baseURL);
 
-  bool newline = trimmed.contains('\n') || trimmed.contains('\r');
-  bool lessThan = trimmed.contains('<');
-  if ((newline || lessThan) && completed.protocolIsInHTTPFamily()) {
-    if (newline) {
+  if (completed.whitespaceRemoved()) {
+    if (completed.protocolIsInHTTPFamily()) {
       UseCounter::count(*this,
                         UseCounter::DocumentCompleteURLHTTPContainingNewline);
-    }
-    if (lessThan) {
-      UseCounter::count(*this,
-                        UseCounter::DocumentCompleteURLHTTPContainingLessThan);
-    }
-    if (newline && lessThan) {
+      bool lessThan = url.contains('<');
+      if (lessThan) {
+        UseCounter::count(
+            *this,
+            UseCounter::DocumentCompleteURLHTTPContainingNewlineAndLessThan);
+
+        if (RuntimeEnabledFeatures::restrictCompleteURLCharacterSetEnabled())
+          return KURL();
+      }
+    } else {
       UseCounter::count(
-          *this,
-          UseCounter::DocumentCompleteURLHTTPContainingNewlineAndLessThan);
-
-      if (RuntimeEnabledFeatures::restrictCompleteURLCharacterSetEnabled())
-        return KURL();
+          *this, UseCounter::DocumentCompleteURLNonHTTPContainingNewline);
     }
-  } else if (newline || lessThan) {
-    UseCounter::count(
-        *this,
-        UseCounter::DocumentCompleteURLNonHTTPContainingNewlineOrLessThan);
   }
-  return completeURLWithOverride(url, m_baseURL);
+  return completed;
 }
 
 KURL Document::completeURLWithOverride(const String& url,

third_party/WebKit/Source/core/frame/UseCounter.h

@@ -1427,9 +1427,8 @@ class CORE_EXPORT UseCounter {
     V8PromiseConstructorReturnedUndefined = 1766,
     FormSubmittedWithUnclosedFormControl = 1767,
     DocumentCompleteURLHTTPContainingNewline = 1768,
-    DocumentCompleteURLHTTPContainingLessThan = 1769,
     DocumentCompleteURLHTTPContainingNewlineAndLessThan = 1770,
-    DocumentCompleteURLNonHTTPContainingNewlineOrLessThan = 1771,
+    DocumentCompleteURLNonHTTPContainingNewline = 1771,
     CSSSelectorInternalMediaControlsTextTrackList = 1772,
     CSSSelectorInternalMediaControlsTextTrackListItem = 1773,
     CSSSelectorInternalMediaControlsTextTrackListItemInput = 1774,

third_party/WebKit/Source/platform/weborigin/KURL.h

@@ -191,6 +191,8 @@ class PLATFORM_EXPORT KURL {
 
   bool isSafeToSendToAnotherThread() const;
 
+  bool whitespaceRemoved() const { return m_parsed.whitespace_removed; }
+
  private:
   void init(const KURL& base,
             const String& relative,

third_party/WebKit/Source/platform/weborigin/KURLTest.cpp

@@ -302,6 +302,37 @@ TEST(KURLTest, EncodeWithURLEscapeSequences) {
   EXPECT_EQ(encodeWithURLEscapeSequences(precomposed), "%C4%99");
 }
 
+TEST(KURLTest, RemoveWhitespace) {
+  struct {
+    const char* input;
+    const char* expected;
+  } cases[] = {
+      {"ht\ntps://example.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"ht\ttps://example.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"ht\rtps://example.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://exa\nmple.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://exa\tmple.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://exa\rmple.com/yay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/y\nay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/y\tay?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/y\ray?boo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?b\noo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?b\too#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?b\roo#foo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?boo#f\noo", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?boo#f\too", "https://example.com/yay?boo#foo"},
+      {"https://example.com/yay?boo#f\roo", "https://example.com/yay?boo#foo"},
+  };
+
+  for (const auto& test : cases) {
+    const KURL input(ParsedURLString, test.input);
+    const KURL expected(ParsedURLString, test.expected);
+    EXPECT_EQ(input, expected);
+    EXPECT_TRUE(input.whitespaceRemoved());
+    EXPECT_FALSE(expected.whitespaceRemoved());
+  }
+}
+
 TEST(KURLTest, ResolveEmpty) {
   KURL emptyBase;
 

tools/metrics/histograms/histograms.xml

@@ -89402,8 +89402,7 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   <int value="1769" label="DocumentCompleteURLHTTPContainingLessThan"/>
   <int value="1770"
       label="DocumentCompleteURLHTTPContainingNewlineAndLessThan"/>
-  <int value="1771"
-      label="DocumentCompleteURLNonHTTPContainingNewlineOrLessThan"/>
+  <int value="1771" label="DocumentCompleteURLNonHTTPContainingNewline"/>
   <int value="1772" label="CSSSelectorInternalMediaControlsTextTrackList"/>
   <int value="1773" label="CSSSelectorInternalMediaControlsTextTrackListItem"/>
   <int value="1774"

url/third_party/mozilla/url_parse.cc

@@ -690,8 +690,7 @@ bool DoExtractQueryKeyValue(const CHAR* spec,
 
 }  // namespace
 
-Parsed::Parsed() : inner_parsed_(NULL) {
-}
+Parsed::Parsed() : whitespace_removed(false), inner_parsed_(NULL) {}
 
 Parsed::Parsed(const Parsed& other) :
     scheme(other.scheme),

url/third_party/mozilla/url_parse.h

@@ -177,6 +177,9 @@ struct URL_EXPORT Parsed {
   // the string with the scheme stripped off.
   Component GetContent() const;
 
+  // True if whitespace was removed from the URL during parsing.
+  bool whitespace_removed;
+
   // This is used for nested URL types, currently only filesystem.  If you
   // parse a filesystem URL, the resulting Parsed will have a nested
   // inner_parsed_ to hold the parsed inner URL's component information.

url/url_util.cc

@@ -199,8 +199,13 @@ bool DoCanonicalize(const CHAR* spec,
   // Remove any whitespace from the middle of the relative URL if necessary.
   // Possibly this will result in copying to the new buffer.
   RawCanonOutputT<CHAR> whitespace_buffer;
-  if (whitespace_policy == REMOVE_WHITESPACE)
-    spec = RemoveURLWhitespace(spec, spec_len, &whitespace_buffer, &spec_len);
+  if (whitespace_policy == REMOVE_WHITESPACE) {
+    int original_len = spec_len;
+    spec =
+        RemoveURLWhitespace(spec, original_len, &whitespace_buffer, &spec_len);
+    if (spec_len != original_len)
+      output_parsed->whitespace_removed = true;
+  }
 
   Parsed parsed_input;
 #ifdef WIN32
@@ -280,6 +285,9 @@ bool DoResolveRelative(const char* base_spec,
   const CHAR* relative = RemoveURLWhitespace(in_relative, in_relative_length,
                                              &whitespace_buffer,
                                              &relative_length);
+  if (in_relative_length != relative_length)
+    output_parsed->whitespace_removed = true;
+
   bool base_is_authority_based = false;
   bool base_is_hierarchical = false;
   if (base_spec &&