Reject some previuosly-escaped chars in hostnames.

url/url_canon_unittest.cc

Index: url/url_canon_unittest.cc
diff --git a/url/url_canon_unittest.cc b/url/url_canon_unittest.cc
index 8ac73bc44b1f39580217e10f19b39e0a29ca6ba9..7e7d619e50518f1289340bc5e4717e3ad896d55e 100644
--- a/url/url_canon_unittest.cc
+++ b/url/url_canon_unittest.cc
@@ -294,10 +294,10 @@ TEST(URLCanonTest, Host) {
   IPAddressCase host_cases[] = {
        // Basic canonicalization, uppercase should be converted to lowercase.
     {"GoOgLe.CoM", L"GoOgLe.CoM", "google.com", Component(0, 10), CanonHostInfo::NEUTRAL, -1, ""},
-      // Spaces and some other characters should be escaped.
-    {"Goo%20 goo%7C|.com", L"Goo%20 goo%7C|.com", "goo%20%20goo%7C%7C.com", Component(0, 22), CanonHostInfo::NEUTRAL, -1, ""},
+      // Spaces and some other characters should fail (used to be allowed).
+    {"Goo%20 goo%7C|.com", L"Goo%20 goo%7C|.com", "goo%20%20goo%7C%7C.com", Component(0, 22), CanonHostInfo::BROKEN, -1, ""},
       // Exciting different types of spaces!
-    {NULL, L"GOO\x00a0\x3000goo.com", "goo%20%20goo.com", Component(0, 16), CanonHostInfo::NEUTRAL, -1, ""},
+    {NULL, L"GOO\x00a0\x3000goo.com", "goo%20%20goo.com", Component(0, 16), CanonHostInfo::BROKEN, -1, ""},
       // Other types of space (no-break, zero-width, zero-width-no-break) are
       // name-prepped away to nothing.
     {NULL, L"GOO\x200b\x2060\xfeffgoo.com", "googoo.com", Component(0, 10), CanonHostInfo::NEUTRAL, -1, ""},
@@ -361,10 +361,9 @@ TEST(URLCanonTest, Host) {
      L"\x915\x94d\x200d\x91c", "xn--11bo0m",
      Component(0, 10), CanonHostInfo::NEUTRAL, -1, ""},
       // Fullwidth exclamation mark is disallowed. UTS 46, table 4, row (b)
-      // However, we do allow this at the moment because we don't use
-      // STD3 rules and canonicalize full-width ASCII to ASCII.
+      // Full-width ASCII is converted to ASCII.
     {"wow\xef\xbc\x81", L"wow\xff01", "wow%21",
-      Component(0, 6), CanonHostInfo::NEUTRAL, -1, ""},
+      Component(0, 6), CanonHostInfo::BROKEN, -1, ""},
       // U+2132 (turned capital F) is disallowed. UTS 46, table 4, row (c)
       // Allowed in IDNA 2003, but the mapping changed after Unicode 3.2
     {"\xe2\x84\xb2oo", L"\x2132oo", "%E2%84%B2oo",
@@ -478,7 +477,7 @@ TEST(URLCanonTest, Host) {
     {"%3g%78%63%30%2e%30%32%35%30%2E.01", L"%3g%78%63%30%2e%30%32%35%30%2E.01", "%253gxc0.0250..01", Component(0, 17), CanonHostInfo::BROKEN, -1, ""},
       // Something that isn't exactly an IP should get treated as a host and
       // spaces escaped.
-    {"192.168.0.1 hello", L"192.168.0.1 hello", "192.168.0.1%20hello", Component(0, 19), CanonHostInfo::NEUTRAL, -1, ""},
+    {"192.168.0.1 hello", L"192.168.0.1 hello", "192.168.0.1%20hello", Component(0, 19), CanonHostInfo::BROKEN, -1, ""},
       // Fullwidth and escaped UTF-8 fullwidth should still be treated as IP.
       // These are "0Xc0.0250.01" in fullwidth.
     {"\xef\xbc\x90%Ef%bc\xb8%ef%Bd%83\xef\xbc\x90%EF%BC%8E\xef\xbc\x90\xef\xbc\x92\xef\xbc\x95\xef\xbc\x90\xef\xbc%8E\xef\xbc\x90\xef\xbc\x91", L"\xff10\xff38\xff43\xff10\xff0e\xff10\xff12\xff15\xff10\xff0e\xff10\xff11", "192.168.0.1", Component(0, 11), CanonHostInfo::IPV4, 3, "C0A80001"},
@@ -490,7 +489,7 @@ TEST(URLCanonTest, Host) {
       // if '(' is escaped after punycode is created we would get xn--%28-8tb
       // (incorrect).
     {"\xd1\x82(", L"\x0442(", "xn--%28-7ed", Component(0, 11),
-      CanonHostInfo::NEUTRAL, -1, ""},
+      CanonHostInfo::BROKEN, -1, ""},
       // Address with all hexidecimal characters with leading number of 1<<32
       // or greater and should return NEUTRAL rather than BROKEN if not all
       // components are numbers.