Escape "--" in the page URL at page serialization

third_party/WebKit/Source/core/page/PageSerializer.cpp

Index: third_party/WebKit/Source/core/page/PageSerializer.cpp
diff --git a/third_party/WebKit/Source/core/page/PageSerializer.cpp b/third_party/WebKit/Source/core/page/PageSerializer.cpp
index 96d4a4220cefb9f71e00401f2d79b9e7dd80b2b3..c41f552a0e6dd0085f231a2303818ee8dd06bc7b 100644
--- a/third_party/WebKit/Source/core/page/PageSerializer.cpp
+++ b/third_party/WebKit/Source/core/page/PageSerializer.cpp
@@ -253,9 +253,7 @@ void LinkChangeSerializerMarkupAccumulator::appendElement(StringBuilder& result,
         // Add MOTW (Mark of the Web) declaration before html tag.
         // See http://msdn2.microsoft.com/en-us/library/ms537628(VS.85).aspx.
         result.append('\n');
-        MarkupFormatter::appendComment(result, String::format(" saved from url=(%04d)%s ",
-            static_cast<int>(document().url().string().utf8().length()),
-            document().url().string().utf8().data()));
+        MarkupFormatter::appendComment(result, PageSerializer::markOfTheWebDeclaration(document().url()));
         result.append('\n');
     }
 
@@ -577,4 +575,26 @@ PageSerializer::Delegate* PageSerializer::delegate()
     return m_delegate.get();
 }
 
+// Returns MOTW (Mark of the Web) declaration before html tag which is in
+// HTML comment, e.g. "<!-- saved from url=(%04zu)%s -->"
+// See http://msdn2.microsoft.com/en-us/library/ms537628(VS.85).aspx.
+String PageSerializer::markOfTheWebDeclaration(const KURL& url)
+{
+    StringBuilder builder;
+    bool emitsMinus = false;
+    CString orignalUrl = url.string().utf8();
+    for (const char* string = orignalUrl.data(); *string; ++string) {
+        const char ch = *string;
+        if (ch == '-' && emitsMinus) {
+            builder.append("%2D");
+            emitsMinus = false;
+            continue;
+        }
+        emitsMinus = ch == '-';
+        builder.append(ch);

[tkent, 2015/09/30 09:01:46]

This line adds a part of UTF8-encoded Unicode character as a single character.
This doesn't work for non-ASCII characters.
Can |url| contain non-ASCII characters?

[yosin_UTC9, 2015/09/30 10:14:59]

No. |url| should be ASCII characters.
I changed |utf8()| to |ascii()|.

+    }
+    CString escapedUrl = builder.toString().utf8();
+    return String::format("saved from url=(%04zu)%s", escapedUrl.length(), escapedUrl.data());

[yosin_UTC9, 2015/09/30 10:14:59]

It seems Windows doesn't like %04zu. I'll use %04d as original code.

+}
+
 } // namespace blink