Mark URLs with empty schemes as invalid.

url/url_canon_etc.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 // Canonicalizers for random bits that aren't big enough for their own files.
 
 #include <string.h>
 
 #include "url/url_canon.h"
 #include "url/url_canon_internal.h"
@@ skipping 70 matching lines @@
 }
 
 template<typename CHAR, typename UCHAR>
 bool DoScheme(const CHAR* spec,
               const Component& scheme,
               CanonOutput* output,
               Component* out_scheme) {
   if (scheme.len <= 0) {
     // Scheme is unspecified or empty, convert to empty by appending a colon.
     *out_scheme = Component(output->length(), 0);
     output->push_back(':');

[Peter Kasting, 2016/09/29 04:54:12]

What happens if we don't bother to insert a colon?  Canonicalization becomes
non-idempotent?

-    return true;
+    return false;
   }
 
   // The output scheme starts from the current position.
   out_scheme->begin = output->length();
 
   // Danger: it's important that this code does not strip any characters;
   // it only emits the canonical version (be it valid or escaped) for each
   // of the input characters. Stripping would put it out of sync with
   // FindAndCompareScheme, which could cause some security checks on
   // schemes to be incorrect.
@@ skipping 255 matching lines @@
 }
 
 void CanonicalizeRef(const base::char16* spec,
                      const Component& ref,
                      CanonOutput* output,
                      Component* out_ref) {
   DoCanonicalizeRef<base::char16, base::char16>(spec, ref, output, out_ref);
 }
 
 }  // namespace url