net: Do not sniff XHTML content

net/base/mime_sniffer.cc

Index: net/base/mime_sniffer.cc
diff --git a/net/base/mime_sniffer.cc b/net/base/mime_sniffer.cc
index 608f870cfed21ec047be1ea799c5c1aa898ba299..45f91b1e4eafa82f3d6777307b4c900eee48b95c 100644
--- a/net/base/mime_sniffer.cc
+++ b/net/base/mime_sniffer.cc
@@ -552,6 +552,11 @@ static const MagicNumber kMagicXML[] = {
   MAGIC_STRING("application/rss+xml", "<rss"),  // UTF-8
 };
 
+static const MagicNumber kMagicXMLForApplicationXML[] = {
+    MAGIC_STRING("application/atom+xml", "<feed"),
+    MAGIC_STRING("application/rss+xml", "<rss"),  // UTF-8
+};
+
 // Returns true and sets result if the content appears to contain XHTML or a
 // feed.
 // Clears have_enough_content if more data could possibly change the result.
@@ -562,6 +567,7 @@ static const MagicNumber kMagicXML[] = {
 // of ASCII -- do we care?
 static bool SniffXML(const char* content,
                      size_t size,
+                     const std::string& type_hint,
                      bool* have_enough_content,
                      std::string* result) {
   // We allow at most 300 bytes of content before we expect the opening tag.
@@ -600,9 +606,15 @@ static bool SniffXML(const char* content,
       continue;
     }
 
-    if (CheckForMagicNumbers(pos, end - pos, kMagicXML, arraysize(kMagicXML),
-                             result))
-      return true;
+    if (type_hint == "application/xml") {
+      if (CheckForMagicNumbers(pos, end - pos, kMagicXMLForApplicationXML,
+                               arraysize(kMagicXMLForApplicationXML), result))
+        return true;
+    } else {
+      if (CheckForMagicNumbers(pos, end - pos, kMagicXML, arraysize(kMagicXML),
+                               result))
+        return true;
+    }
 
     // TODO(evanm): handle RSS 1.0, which is an RDF format and more difficult
     // to identify.
@@ -822,7 +834,8 @@ bool SniffMimeType(const char* content,
     // We're not interested in sniffing these types for images and the like.
     // Instead, we're looking explicitly for a feed.  If we don't find one
     // we're done and return early.
-    if (SniffXML(content, content_size, &have_enough_content, result))
+    if (SniffXML(content, content_size, type_hint, &have_enough_content,
+                 result))
       return true;
     return have_enough_content;
   }