From 4498419438746bf94fc6a296ccc2eb61a57e2738 Mon Sep 17 00:00:00 2001
From: jgruber <jgruber@chromium.org>
Date: Fri, 7 Apr 2017 00:52:10 -0700
Subject: [PATCH] [regexp] Add tests for recent changes in Annex B

See https://github.com/tc39/ecma262/pull/303.

BUG=v8:5937,v8:6201

Review-Url: https://codereview.chromium.org/2793313002
Cr-Commit-Position: refs/heads/master@{#44467}
---
 src/regexp/regexp-parser.cc | 14 ++++++++-----
 test/mjsunit/regexp.js      | 39 +++++++++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 5 deletions(-)

diff --git a/src/regexp/regexp-parser.cc b/src/regexp/regexp-parser.cc
index 656e88cff8..84044cfc36 100644
--- a/src/regexp/regexp-parser.cc
+++ b/src/regexp/regexp-parser.cc
@@ -499,9 +499,9 @@ RegExpTree* RegExpParser::ParseDisjunction() {
             uc32 letter = controlLetter & ~('a' ^ 'A');
             if (letter < 'A' || 'Z' < letter) {
               // controlLetter is not in range 'A'-'Z' or 'a'-'z'.
-              // This is outside the specification. We match JSC in
-              // reading the backslash as a literal character instead
-              // of as starting an escape.
+              // Read the backslash as a literal character instead of as
+              // starting an escape.
+              // ES#prod-annexB-ExtendedPatternCharacter
               if (unicode()) {
                 // With /u, invalid escapes are not treated as identity escapes.
                 return ReportError(CStrVector("Invalid unicode escape"));
@@ -1046,6 +1046,7 @@ uc32 RegExpParser::ParseOctalLiteral() {
   DCHECK(('0' <= current() && current() <= '7') || current() == kEndMarker);
   // For compatibility with some other browsers (not all), we parse
   // up to three octal digits with a value below 256.
+  // ES#prod-annexB-LegacyOctalEscapeSequence
   uc32 value = current() - '0';
   Advance();
   if ('0' <= current() && current() <= '7') {
@@ -1334,8 +1335,9 @@ uc32 RegExpParser::ParseClassCharacterEscape() {
     case 'c': {
       uc32 controlLetter = Next();
       uc32 letter = controlLetter & ~('A' ^ 'a');
-      // For compatibility with JSC, inside a character class. We also accept
-      // digits and underscore as control characters, unless with /u.
+      // Inside a character class, we also accept digits and underscore as
+      // control characters, unless with /u. See Annex B:
+      // ES#prod-annexB-ClassControlLetter
       if (letter >= 'A' && letter <= 'Z') {
         Advance(2);
         // Control letters mapped to ASCII control characters in the range
@@ -1354,6 +1356,7 @@ uc32 RegExpParser::ParseClassCharacterEscape() {
       }
       // We match JSC in reading the backslash as a literal
       // character instead of as starting an escape.
+      // TODO(v8:6201): Not yet covered by the spec.
       return '\\';
     }
     case '0':
@@ -1373,6 +1376,7 @@ uc32 RegExpParser::ParseClassCharacterEscape() {
       // 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.
+      // ES#prod-annexB-LegacyOctalEscapeSequence
       if (unicode()) {
         // With /u, decimal escape is not interpreted as octal character code.
         ReportError(CStrVector("Invalid class escape"));
diff --git a/test/mjsunit/regexp.js b/test/mjsunit/regexp.js
index c9a1fd21b3..6fb5660c08 100644
--- a/test/mjsunit/regexp.js
+++ b/test/mjsunit/regexp.js
@@ -764,3 +764,42 @@ re.lastIndex = NaN;
 assertEquals(NaN, re.lastIndex);
 "abc".search(re);
 assertEquals(NaN, re.lastIndex);
+
+// Annex B changes: https://github.com/tc39/ecma262/pull/303
+
+assertThrows("/{1}/", SyntaxError);
+assertTrue(/^{*$/.test("{{{"));
+assertTrue(/^}*$/.test("}}}"));
+assertTrue(/]/.test("]"));
+assertTrue(/^\c%$/.test("\\c%"));   // We go into ExtendedPatternCharacter.
+assertTrue(/^\d%$/.test("2%"));     // ... CharacterClassEscape.
+assertTrue(/^\e%$/.test("e%"));     // ... IdentityEscape.
+assertTrue(/^\ca$/.test("\u{1}"));  // ... ControlLetter.
+assertTrue(/^\cA$/.test("\u{1}"));  // ... ControlLetter.
+assertTrue(/^\c9$/.test("\\c9"));   // ... ExtendedPatternCharacter.
+assertTrue(/^\c$/.test("\\c"));   // ... ExtendedPatternCharacter.
+assertTrue(/^[\c%]*$/.test("\\c%"));  // TODO(v8:6201): Not covered by the spec.
+assertTrue(/^[\c:]*$/.test("\\c:"));  // TODO(v8:6201): Not covered by the spec.
+assertTrue(/^[\c0]*$/.test("\u{10}"));  // ... ClassControlLetter.
+assertTrue(/^[\c1]*$/.test("\u{11}"));  // ('0' % 32 == 0x10)
+assertTrue(/^[\c2]*$/.test("\u{12}"));
+assertTrue(/^[\c3]*$/.test("\u{13}"));
+assertTrue(/^[\c4]*$/.test("\u{14}"));
+assertTrue(/^[\c5]*$/.test("\u{15}"));
+assertTrue(/^[\c6]*$/.test("\u{16}"));
+assertTrue(/^[\c7]*$/.test("\u{17}"));
+assertTrue(/^[\c8]*$/.test("\u{18}"));
+assertTrue(/^[\c9]*$/.test("\u{19}"));
+assertTrue(/^[\c_]*$/.test("\u{1F}"));
+assertTrue(/^[\c11]*$/.test("\u{11}1"));
+assertTrue(/^[\8]*$/.test("8"));  // ... ClassEscape ~~> IdentityEscape.
+assertTrue(/^[\7]*$/.test("\u{7}"));  // ... ClassEscape
+                                      // ~~> LegacyOctalEscapeSequence.
+assertTrue(/^[\11]*$/.test("\u{9}"));
+assertTrue(/^[\111]*$/.test("\u{49}"));
+assertTrue(/^[\222]*$/.test("\u{92}"));
+assertTrue(/^[\333]*$/.test("\u{DB}"));
+assertTrue(/^[\444]*$/.test("\u{24}4"));
+assertTrue(/^[\d-X]*$/.test("234-X-432"));  // CharacterRangeOrUnion.
+assertTrue(/^[\d-X-Z]*$/.test("234-XZ-432"));
+assertFalse(/^[\d-X-Z]*$/.test("234-XYZ-432"));