Add implicit trailing dot domain matching support to URLPattern.

extensions/common/url_pattern.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 "extensions/common/url_pattern.h"
 
 #include <stddef.h>
 
 #include <ostream>
 
@@ skipping 95 matching lines @@
 // called for the patterns inside URLPatternSet. In those cases, we know that
 // the path will have only a single wildcard at the end. This makes figuring
 // out overlap much easier. It seems like there is probably a computer-sciency
 // way to solve the general case, but we don't need that yet.
 std::string StripTrailingWildcard(const std::string& path) {
   size_t wildcard_index = path.find('*');
   size_t path_last = path.size() - 1;
   return wildcard_index == path_last ? path.substr(0, path_last) : path;
 }
 
+// Removes trailing dot from the host if any. |host| must outlive a call to this
+// function.
+base::StringPiece CanonicalizeHostForMatching(const std::string& host) {
+  base::StringPiece host_piece(host);
+  if (host_piece.ends_with("."))
+    host_piece.remove_suffix(1);
+  return host_piece;
+}
+
 }  // namespace
 
 // static
 bool URLPattern::IsValidSchemeForExtensions(const std::string& scheme) {
   for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
     if (scheme == kValidSchemes[i])
       return true;
   }
   return false;
 }
@@ skipping 271 matching lines @@
 
 bool URLPattern::MatchesHost(const std::string& host) const {
   std::string test(url::kHttpScheme);
   test += url::kStandardSchemeSeparator;
   test += host;
   test += "/";
   return MatchesHost(GURL(test));
 }
 
 bool URLPattern::MatchesHost(const GURL& test) const {
+  // Explicit std::string to outlive the call to CanonicalizeHostForMatching.
+  const std::string& test_host_str = test.host();

[Devlin, 2016/10/31 16:13:17]

GURL::host() actually returns a std::string (rather than a const &), so there
this should be const std::string.  (The StringPiece work was important to
prevent an additional copy from the host str.)

[Devlin, 2016/10/31 16:58:47]

Actually, even better yet, we have a GURL::host_piece() method that would work
here if we change CanonicalizeHostForMatching to take a StringPiece.

+  const base::StringPiece test_host(CanonicalizeHostForMatching(test_host_str));
+  const base::StringPiece pattern_host(CanonicalizeHostForMatching(host_));
+
   // If the hosts are exactly equal, we have a match.
-  if (test.host() == host_)
+  if (test_host == pattern_host)
     return true;
 
   // If we're matching subdomains, and we have no host in the match pattern,
   // that means that we're matching all hosts, which means we have a match no
   // matter what the test host is.
-  if (match_subdomains_ && host_.empty())
+  if (match_subdomains_ && pattern_host.empty())
     return true;
 
   // Otherwise, we can only match if our match pattern matches subdomains.
   if (!match_subdomains_)
     return false;
 
   // We don't do subdomain matching against IP addresses, so we can give up now
   // if the test host is an IP address.
   if (test.HostIsIPAddress())
     return false;
 
   // Check if the test host is a subdomain of our host.
-  if (test.host().length() <= (host_.length() + 1))
+  if (test_host.length() <= (pattern_host.length() + 1))
     return false;
 
-  if (test.host().compare(test.host().length() - host_.length(),
-                          host_.length(), host_) != 0)
+  if (test_host.find(pattern_host,

[Devlin, 2016/10/31 16:13:17]

simpler:
if (!test_host.ends_with(pattern_host))
  return false;

+                     test_host.length() - pattern_host.length()) ==
+      base::StringPiece::npos)
     return false;
 
-  return test.host()[test.host().length() - host_.length() - 1] == '.';
+  return test_host[test_host.length() - pattern_host.length() - 1] == '.';
 }
 
 bool URLPattern::ImpliesAllHosts() const {
   // Check if it matches all urls or is a pattern like http://*/*.
   if (match_all_urls_ ||
       (match_subdomains_ && host_.empty() && port_ == "*" && path_ == "/*")) {
     return true;
   }
 
   // If this doesn't even match subdomains, it can't possibly imply all hosts.
@@ skipping 160 matching lines @@
   }
 
   return result;
 }
 
 // static
 const char* URLPattern::GetParseResultString(
     URLPattern::ParseResult parse_result) {
   return kParseResultMessages[parse_result];
 }