Text Encoding: Add 'replacement' as label for replacement encoding

The 'replacement' encoding originated as a spec concept to prevent
security attacks via problematic encodings by recognizing the
label but not decoding the stream.

It was initially specified as the only encoding where the name
wasn't one of the labels, requiring special cases in all implementations.
Based on more implementer feedback we'd like to remove the special
case. Delete the special case code in Blink too.

See also: whatwg/encoding#70

Bug: 744405

Change-Id: Ia15ccef1a9d7f35c23af4509a5a9758cbefc2087
Reviewed-on: https://chromium-review.googlesource.com/559973
Reviewed-by: Kent Tamura <[email protected]>
Commit-Queue: Joshua Bell <[email protected]>
Cr-Commit-Position: refs/heads/master@{#487288}

Author: inexorabletash

third_party/WebKit/LayoutTests/FlagExpectations/enable-blink-features=LayoutNG

@@ -7363,7 +7363,6 @@ crbug.com/591099 fast/encoding/bom-in-content.html [ Failure ]
 crbug.com/591099 fast/encoding/bracket-in-script.html [ Failure ]
 crbug.com/591099 fast/encoding/charset-invalid.html [ Failure ]
 crbug.com/591099 fast/encoding/charset-koi8-u.html [ Failure ]
-crbug.com/591099 fast/encoding/charset-replacement.html [ Failure ]
 crbug.com/591099 fast/encoding/charset-xuser-defined.html [ Crash Failure ]
 crbug.com/591099 fast/encoding/css-charset-default.xhtml [ Failure ]
 crbug.com/591099 fast/encoding/css-charset.html [ Failure ]

third_party/WebKit/LayoutTests/fast/encoding/char-decoding.html

@@ -144,7 +144,7 @@
 testDecode('unicodeFFFE', '%D8%69%DE%D6', 'U+D869/U+DED6');
 
 // Replacement encodings should decode non-empty streams as replacement (U+FFFD) then EOF
-["csiso2022kr", "hz-gb-2312", "iso-2022-cn", "iso-2022-cn-ext", "iso-2022-kr"]
+["replacement", "csiso2022kr", "hz-gb-2312", "iso-2022-cn", "iso-2022-cn-ext", "iso-2022-kr"]
 .forEach(function(encoding) {
     testDecode(encoding, "", "");
     testDecode(encoding, "%41%42%43%61%62%63%31%32%33%A0", "U+FFFD");

third_party/WebKit/LayoutTests/fast/encoding/charset-replacement-expected.txt

@@ -16,12 +16,8 @@ TextCodecReplacement::TextCodecReplacement()
 
 void TextCodecReplacement::RegisterEncodingNames(
     EncodingNameRegistrar registrar) {
-  // The 'replacement' label itself should not be referenceable by
-  // resources or script - it's a specification convenience - but much of
-  // the wtf/text API asserts that an encoding name is a label for itself.
-  // This is handled in TextEncoding by marking it as not valid.
+  // Taken from the alias table at·https://encoding.spec.whatwg.org/
   registrar("replacement", "replacement");
-
   registrar("csiso2022kr", "replacement");
   registrar("hz-gb-2312", "replacement");
   registrar("iso-2022-cn", "replacement");

third_party/WebKit/LayoutTests/fast/encoding/charset-replacement.html

@@ -10,6 +10,11 @@
 
 namespace WTF {
 
+// The "replacement" encoding exists to prevent attacks that abuse a mismatch
+// between encodings supported on the server and the client. The encoder is
+// the same as UTF-8; and for a non-empty input the decoder emits U+FFFD and
+// terminates. See: https://encoding.spec.whatwg.org/#replacement and
+// https://encoding.spec.whatwg.org/#output-encodings
 class TextCodecReplacement final : public TextCodecUTF8 {
  public:
   TextCodecReplacement();

third_party/WebKit/Source/platform/wtf/text/TextCodecReplacement.cpp

@@ -20,9 +20,8 @@ namespace {
 const char* g_replacement_alias = "iso-2022-kr";
 
 TEST(TextCodecReplacement, Aliases) {
-  // "replacement" is not a valid alias for itself
-  EXPECT_FALSE(TextEncoding("replacement").IsValid());
-  EXPECT_FALSE(TextEncoding("rEpLaCeMeNt").IsValid());
+  EXPECT_TRUE(TextEncoding("replacement").IsValid());
+  EXPECT_TRUE(TextEncoding("rEpLaCeMeNt").IsValid());
 
   EXPECT_TRUE(TextEncoding(g_replacement_alias).IsValid());
   EXPECT_STREQ("replacement", TextEncoding(g_replacement_alias).GetName());

third_party/WebKit/Source/platform/wtf/text/TextCodecReplacement.h

@@ -44,16 +44,10 @@ static const TextEncoding& UTF7Encoding() {
 
 TextEncoding::TextEncoding(const char* name)
     : name_(AtomicCanonicalTextEncodingName(name)) {
-  // Aliases are valid, but not "replacement" itself.
-  if (name_ && IsReplacementEncoding(name))
-    name_ = 0;
 }
 
 TextEncoding::TextEncoding(const String& name)
     : name_(AtomicCanonicalTextEncodingName(name)) {
-  // Aliases are valid, but not "replacement" itself.
-  if (name_ && IsReplacementEncoding(name))
-    name_ = 0;
 }
 
 String TextEncoding::Decode(const char* data,

third_party/WebKit/Source/platform/wtf/text/TextCodecReplacementTest.cpp

@@ -222,14 +222,6 @@ static void BuildBaseTextCodecMaps() {
   TextCodecUserDefined::RegisterCodecs(AddToTextCodecMap);
 }
 
-bool IsReplacementEncoding(const char* alias) {
-  return alias && !strcasecmp(alias, "replacement");
-}
-
-bool IsReplacementEncoding(const String& alias) {
-  return alias == "replacement";
-}
-
 static void ExtendTextCodecMaps() {
   TextCodecReplacement::RegisterEncodingNames(AddToTextEncodingNameMap);
   TextCodecReplacement::RegisterCodecs(AddToTextCodecMap);

third_party/WebKit/Source/platform/wtf/text/TextEncoding.cpp

@@ -46,8 +46,6 @@ template <typename CharacterType>
 const char* AtomicCanonicalTextEncodingName(const CharacterType*, size_t);
 const char* AtomicCanonicalTextEncodingName(const String&);
 bool NoExtendedTextEncodingNameUsed();
-bool IsReplacementEncoding(const char* alias);
-bool IsReplacementEncoding(const String& alias);
 
 #ifndef NDEBUG
 void DumpTextEncodingNameMap();