ES6 template literals should not use legacy octal strings

Correctly handle SyntaxErrors in escape sequences.

BUG=v8:3736
LOG=Y
R=dslomov@chromium.org, caitpotter88@gmail.com

Review URL: https://codereview.chromium.org/811113002

Cr-Commit-Position: refs/heads/master@{#25891}

src/scanner.cc

@@ -697,13 +697,13 @@ void Scanner::SeekForward(int pos) {
 }
 
 
-template <bool capture_raw>
+template <bool capture_raw, bool in_template_literal>
 bool Scanner::ScanEscape() {
   uc32 c = c0_;
   Advance<capture_raw>();
 
   // Skip escaped newlines.
-  if (c0_ >= 0 && unicode_cache_->IsLineTerminator(c)) {
+  if (!in_template_literal && c0_ >= 0 && unicode_cache_->IsLineTerminator(c)) {
     // Allow CR+LF newlines in multiline string literals.
     if (IsCarriageReturn(c) && IsLineFeed(c0_)) Advance<capture_raw>();
     // Allow LF+CR newlines in multiline string literals.
@@ -725,23 +725,45 @@ bool Scanner::ScanEscape() {
       if (c < 0) return false;
       break;
     }
-    case 'v' : c = '\v'; break;
-    case 'x' : {
+    case 'v':
+      c = '\v';
+      break;
+    case 'x': {
       c = ScanHexNumber<capture_raw>(2);
       if (c < 0) return false;
       break;
     }
-    case '0' :  // fall through
-    case '1' :  // fall through
-    case '2' :  // fall through
-    case '3' :  // fall through
-    case '4' :  // fall through
-    case '5' :  // fall through
-    case '6' :  // fall through
+    case '0':
+      if (in_template_literal) {
+        // \ 0 DecimalDigit is never allowed in templates.
+        if (IsDecimalDigit(c0_)) {
+          Advance<capture_raw>();  // Advance to include the problematic char.
+          return false;
+        }
+
+        // The TV of TemplateCharacter :: \ EscapeSequence is the CV of
+        //     EscapeSequence.
+        // The CV of EscapeSequence :: 0 is the code unit value 0.
+        c = 0;
+        break;
+      }
+    // Fall through.
+    case '1':  // fall through
+    case '2':  // fall through
+    case '3':  // fall through
+    case '4':  // fall through
+    case '5':  // fall through
+    case '6':  // fall through
     case '7':
+      if (!in_template_literal) {
         c = ScanOctalEscape<capture_raw>(c, 2);
         break;
       }
+    // Fall through
+    case '8':
+    case '9':
+      if (in_template_literal) return false;
+  }
 
   // According to ECMA-262, section 7.8.4, characters not covered by the
   // above cases should be illegal, but they are commonly handled as
@@ -787,7 +809,7 @@ Token::Value Scanner::ScanString() {
     uc32 c = c0_;
     Advance();
     if (c == '\\') {
-      if (c0_ < 0 || !ScanEscape<false>()) return Token::ILLEGAL;
+      if (c0_ < 0 || !ScanEscape<false, false>()) return Token::ILLEGAL;
     } else {
       AddLiteralChar(c);
     }
@@ -818,6 +840,7 @@ Token::Value Scanner::ScanTemplateSpan() {
   LiteralScope literal(this);
   StartRawLiteral();
   const bool capture_raw = true;
+  const bool in_template_literal = true;
 
   while (true) {
     uc32 c = c0_;
@@ -844,11 +867,8 @@ Token::Value Scanner::ScanTemplateSpan() {
             AddRawLiteralChar('\n');
           }
         }
-      } else if (c0_ == '0') {
-        Advance<capture_raw>();
-        AddLiteralChar('0');
-      } else {
-        ScanEscape<true>();
+      } else if (!ScanEscape<capture_raw, in_template_literal>()) {
+        return Token::ILLEGAL;
       }
     } else if (c < 0) {
       // Unterminated template literal

src/scanner.h

@@ -677,8 +677,9 @@ class Scanner {
   // Scans an escape-sequence which is part of a string and adds the
   // decoded character to the current literal. Returns true if a pattern
   // is scanned.
-  template <bool capture_raw>
+  template <bool capture_raw, bool in_template_literal>
   bool ScanEscape();
+
   // Decodes a Unicode escape-sequence which is part of an identifier.
   // If the escape sequence cannot be decoded the result is kBadChar.
   uc32 ScanIdentifierUnicodeEscape();

test/mjsunit/harmony/templates.js

@@ -253,7 +253,7 @@ var obj = {
   //   The TRV of CharacterEscapeSequence :: NonEscapeCharacter is the CV of the
   //   NonEscapeCharacter.
   calls = 0;
-  (function(s) { calls++; assertEquals("\u005Cx", s.raw[0]); })`\x`;
+  (function(s) { calls++; assertEquals("\u005Cz", s.raw[0]); })`\z`;
   assertEquals(1, calls);
 
   // The TRV of LineTerminatorSequence :: <LF> is the code unit value 0x000A.
@@ -471,3 +471,25 @@ var obj = {
     // block
   }`jkl`;
 })();
+
+
+(function testLegacyOctal() {
+  assertEquals('\u0000', `\0`);
+  assertEquals('\u0000a', `\0a`);
+  for (var i = 0; i < 10; i++) {
+    var code = "`\\0" + i + "`";
+    assertThrows(code, SyntaxError);
+  }
+
+  assertEquals('\\0', String.raw`\0`);
+})();
+
+
+(function testSyntaxErrorsNonEscapeCharacter() {
+  assertThrows("`\\x`", SyntaxError);
+  assertThrows("`\\u`", SyntaxError);
+  for (var i = 1; i < 10; i++) {
+    var code = "`\\" + i + "`";
+    assertThrows(code, SyntaxError);
+  }
+})();