Mark URLs with empty schemes as invalid.

components/url_formatter/url_fixer.cc

 // Copyright (c) 2012 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.
 
 #include "components/url_formatter/url_fixer.h"
 
 #include <stddef.h>
 
 #include <algorithm>
 
@@ skipping 462 matching lines @@
 
   // We need to add a scheme in order for ParseStandardURL to be happy.
   // Find the first non-whitespace character.
   std::string::iterator first_nonwhite = text->begin();
   while ((first_nonwhite != text->end()) &&
          base::IsUnicodeWhitespace(*first_nonwhite))
     ++first_nonwhite;
 
   // Construct the text to parse by inserting the scheme.
   std::string inserted_text(scheme);
-  inserted_text.append(url::kStandardSchemeSeparator);
+  if (first_nonwhite == text->end() || *first_nonwhite != ':')
+    inserted_text.append(url::kStandardSchemeSeparator);  // Add "://".

[Peter Kasting, 2016/10/05 00:56:30]

Nit: After your explanation, I understand why this works.  GURL fixing up
"foo:bar" into "foo://bar" was the missing piece.

This seems a little unclear, so it'd be nice to do one of the following:

(1) Skip over the colon in the input:

  inserted_text.append(url::kStandardSchemeSeparator);
  // Assume a leading colon was meant to be a scheme separator (which we can
  // thus skip, since we just added this).
  if (first_nonwhite != text->end() && *first_nonwhite == ':')
    ++first_nonwhite;

(2) Add a comment about GURL's behavior here:

  // Assume a leading colon was meant to be a scheme separator (which GURL will
  // fix up for us into the full "://").  Otherwise add the separator ourselves.
  if (first_nonwhite == text->end() || *first_nonwhite != ':')
    inserted_text.append(url::kStandardSchemeSeparator);

[brettw, 2016/10/05 05:10:12]

I did the second.

   std::string text_to_parse(text->begin(), first_nonwhite);
   text_to_parse.append(inserted_text);
   text_to_parse.append(first_nonwhite, text->end());
 
   // Have the GURL parser do the heavy lifting for us.
   url::ParseStandardURL(text_to_parse.data(),
                         static_cast<int>(text_to_parse.length()), parts);
 
   // Offset the results of the parse to match the original text.
   const int offset = -static_cast<int>(inserted_text.length());
@@ skipping 181 matching lines @@
 }
 
 bool IsEquivalentScheme(const std::string& scheme1,
                         const std::string& scheme2) {
   return scheme1 == scheme2 ||
          (scheme1 == url::kAboutScheme && scheme2 == kChromeUIScheme) ||
          (scheme1 == kChromeUIScheme && scheme2 == url::kAboutScheme);
 }
 
 }  // namespace url_formatter