Reject some previuosly-escaped chars in hostnames.

url/url_canon_host.cc

Index: url/url_canon_host.cc
diff --git a/url/url_canon_host.cc b/url/url_canon_host.cc
index 76a22369b87648ce3651a94e4ca6aaa460224ac6..433ae1408197e68e90ea6b3b67f464656a248ff4 100644
--- a/url/url_canon_host.cc
+++ b/url/url_canon_host.cc
@@ -10,7 +10,7 @@ namespace url {
 
 namespace {
 
-// For reference, here's what IE supports:
+// For reference, here's what IE6 supported:
 // Key: 0 (disallowed: failure if present in the input)
 //      + (allowed either escaped or unescaped, and unmodified)
 //      U (allowed escaped or unescaped but always unescaped if present in
@@ -38,34 +38,23 @@ namespace {
 // Surprisingly, space is accepted in the input and always escaped.
 
 // This table lists the canonical version of all characters we allow in the
-// input, with 0 indicating it is disallowed. We use the magic kEscapedHostChar
-// value to indicate that this character should be escaped. We are a little more
-// restrictive than IE, but less restrictive than Firefox.
-//
-// Note that we disallow the % character. We will allow it when part of an
-// escape sequence, of course, but this disallows "%25". Even though IE allows
-// it, allowing it would put us in a funny state. If there was an invalid
-// escape sequence like "%zz", we'll add "%25zz" to the output and fail.
-// Allowing percents means we'll succeed a second time, so validity would change
-// based on how many times you run the canonicalizer. We prefer to always report
-// the same vailidity, so reject this.
-const unsigned char kEsc = 0xff;
+// input, with 0 indicating it is disallowed.
 const unsigned char kHostCharLookup[0x80] = {
 // 00-1f: all are invalid
      0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,
      0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,
 //  ' '   !    "    #    $    %    &    '    (    )    *    +    ,    -    .    /
-   kEsc,kEsc,kEsc,kEsc,kEsc,  0, kEsc,kEsc,kEsc,kEsc,kEsc, '+',kEsc, '-', '.',  0,
+     0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  '-', '.',  0,
 //   0    1    2    3    4    5    6    7    8    9    :    ;    <    =    >    ?
-    '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', ':',  0 ,kEsc,kEsc,kEsc,  0 ,
+    '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', ':',  0,   0,   0,   0,   0 ,
 //   @    A    B    C    D    E    F    G    H    I    J    K    L    M    N    O
-   kEsc, 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o',
+     0,  'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o',
 //   P    Q    R    S    T    U    V    W    X    Y    Z    [    \    ]    ^    _
     'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', '[',  0 , ']',  0 , '_',
 //   `    a    b    c    d    e    f    g    h    i    j    k    l    m    n    o
-   kEsc, 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o',
+     0,  'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o',
 //   p    q    r    s    t    u    v    w    x    y    z    {    |    }    ~
-    'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z',kEsc,kEsc,kEsc,  0 ,  0 };
+    'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z',  0,   0,   0,   0,   0 };
 
 const int kTempHostBufferLen = 1024;
 typedef RawCanonOutputT<char, kTempHostBufferLen> StackBuffer;
@@ -142,9 +131,6 @@ bool DoSimpleHost(const INCHAR* host,
         // Invalid character, add it as percent-escaped and mark as failed.
         AppendEscapedChar(source, output);
         success = false;
-      } else if (replacement == kEsc) {
-        // This character is valid but should be escaped.
-        AppendEscapedChar(source, output);
       } else {
         // Common case, the given character is valid in a hostname, the lookup
         // table tells us the canonical representation of that character (lower