[regexp] parse RegExpUnicodeEscapeSequence according to spec.

src/regexp/regexp-parser.cc

Index: src/regexp/regexp-parser.cc
diff --git a/src/regexp/regexp-parser.cc b/src/regexp/regexp-parser.cc
index 1e6a0dacca638881cabf0470bf2798c0c53bc0cb..3091b6d49c027937ab67773871884dd1d3ae5864 100644
--- a/src/regexp/regexp-parser.cc
+++ b/src/regexp/regexp-parser.cc
@@ -461,7 +461,7 @@ RegExpTree* RegExpParser::ParseDisjunction() {
             Advance(2);
             uc32 value;
             if (ParseUnicodeEscape(&value)) {
-              builder->AddUnicodeCharacter(value);
+              builder->AddEscapedUnicodeCharacter(value);
             } else if (!unicode()) {
               builder->AddCharacter('u');
             } else {
@@ -780,7 +780,7 @@ bool RegExpParser::ParseHexEscape(int length, uc32* value) {
   return true;
 }
 
-
+// This parses RegExpUnicodeEscapeSequence as described in ECMA262.
 bool RegExpParser::ParseUnicodeEscape(uc32* value) {
   // Accept both \uxxxx and \u{xxxxxx} (if harmony unicode escapes are
   // allowed). In the latter case, the number of hex digits between { } is
@@ -798,7 +798,24 @@ bool RegExpParser::ParseUnicodeEscape(uc32* value) {
     return false;
   }
   // \u but no {, or \u{...} escapes not allowed.
-  return ParseHexEscape(4, value);
+  bool result = ParseHexEscape(4, value);
+  if (result && unicode() && unibrow::Utf16::IsLeadSurrogate(*value) &&
+      current() == '\\') {
+    // Attempt to read trail surrogate.
+    int start = position();
+    if (Next() == 'u') {
+      Advance(2);
+      uc32 trail;
+      if (ParseHexEscape(4, &trail) &&
+          unibrow::Utf16::IsTrailSurrogate(trail)) {
+        *value = unibrow::Utf16::CombineSurrogatePair(static_cast<uc16>(*value),
+                                                      static_cast<uc16>(trail));
+        return true;
+      }
+    }
+    Reset(start);
+  }
+  return result;
 }
 
 
@@ -879,14 +896,18 @@ uc32 RegExpParser::ParseClassCharacterEscape() {
     case '5':
     case '6':
     case '7':
-      // For compatibility, we interpret a decimal escape that isn't
-      // a back reference (and therefore either \0 or not valid according
-      // to the specification) as a 1..3 digit octal character code.
       if (unicode()) {
+        // \0 is interpreted as \u0000 if it is not followed by another digit.
+        if (current() == '0') {
+          Advance();
+          if (current() < '0' || current() > '9') return 0;

[Yang, 2016/02/09 19:18:31]

This now matches how we parse \<digit> outside of a character class.

+        }
         // With /u, decimal escape is not interpreted as octal character code.
         ReportError(CStrVector("Invalid class escape"));
         return 0;
       }
+      // For backward compatibility, we interpret escaped digit from 0 to 7 as
+      // a 1..3 digit octal character code.
       return ParseOctalLiteral();
     case 'x': {
       Advance();
@@ -916,9 +937,9 @@ uc32 RegExpParser::ParseClassCharacterEscape() {
     }
     default: {
       uc32 result = current();
-      // With /u, no identity escapes except for syntax characters are
+      // With /u, no identity escapes except for syntax characters and '-' are
       // allowed. Otherwise, all identity escapes are allowed.
-      if (!unicode() || IsSyntaxCharacterOrSlash(result)) {
+      if (!unicode() || IsSyntaxCharacterOrSlash(result) || result == '-') {
         Advance();
         return result;
       }
@@ -954,22 +975,6 @@ CharacterRange RegExpParser::ParseClassAtom(uc16* char_class) {
     Advance();
   }
 
-  if (unicode() && unibrow::Utf16::IsLeadSurrogate(first)) {
-    // Combine with possibly following trail surrogate.
-    int start = position();
-    uc32 second = current();
-    if (second == '\\') {
-      second = ParseClassCharacterEscape(CHECK_FAILED);
-    } else {
-      Advance();
-    }
-    if (unibrow::Utf16::IsTrailSurrogate(second)) {
-      first = unibrow::Utf16::CombineSurrogatePair(first, second);
-    } else {
-      Reset(start);
-    }
-  }
-
   return CharacterRange::Singleton(first);
 }
 
@@ -1198,6 +1203,13 @@ void RegExpBuilder::AddUnicodeCharacter(uc32 c) {
   }
 }
 
+void RegExpBuilder::AddEscapedUnicodeCharacter(uc32 character) {
+  // A lead or trail surrogate parsed via escape sequence will not
+  // pair up with any preceding lead or following trail surrogate.
+  FlushPendingSurrogate();
+  AddUnicodeCharacter(character);
+  FlushPendingSurrogate();
+}
 
 void RegExpBuilder::AddEmpty() { pending_empty_ = true; }