Fix crash due RegExpAtom method called on RegExpCharacterClass object.

src/jsregexp.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 915 matching lines @@
 }
 
 
 void RegExpText::AppendToText(RegExpText* text, Zone* zone) {
   for (int i = 0; i < elements()->length(); i++)
     text->AddElement(elements()->at(i), zone);
 }
 
 
 TextElement TextElement::Atom(RegExpAtom* atom) {
-  TextElement result = TextElement(ATOM);
-  result.data.u_atom = atom;
-  return result;
+  return TextElement(ATOM, atom);
 }
 
 
-TextElement TextElement::CharClass(
-      RegExpCharacterClass* char_class) {
-  TextElement result = TextElement(CHAR_CLASS);
-  result.data.u_char_class = char_class;
-  return result;
+TextElement TextElement::CharClass(RegExpCharacterClass* char_class) {
+  return TextElement(CHAR_CLASS, char_class);
 }
 
 
-int TextElement::length() {
-  if (text_type == ATOM) {
-    return data.u_atom->length();
-  } else {
-    ASSERT(text_type == CHAR_CLASS);
-    return 1;
+int TextElement::length() const {
+  switch (text_type()) {
+    case ATOM:
+      return atom()->length();
+
+    case CHAR_CLASS:
+      return 1;
   }
+  UNREACHABLE();
+  return 0;
 }
 
 
 DispatchTable* ChoiceNode::GetTable(bool ignore_case) {
   if (table_ == NULL) {
     table_ = new(zone()) DispatchTable(zone());
     DispatchTableConstructor cons(table_, ignore_case, zone());
     cons.BuildTable(this);
   }
   return table_;
@@ skipping 1587 matching lines @@
   ASSERT(characters_filled_in < details->characters());
   int characters = details->characters();
   int char_mask;
   if (compiler->ascii()) {
     char_mask = String::kMaxOneByteCharCode;
   } else {
     char_mask = String::kMaxUtf16CodeUnit;
   }
   for (int k = 0; k < elms_->length(); k++) {
     TextElement elm = elms_->at(k);
-    if (elm.text_type == TextElement::ATOM) {
-      Vector<const uc16> quarks = elm.data.u_atom->data();
+    if (elm.text_type() == TextElement::ATOM) {
+      Vector<const uc16> quarks = elm.atom()->data();
       for (int i = 0; i < characters && i < quarks.length(); i++) {
         QuickCheckDetails::Position* pos =
             details->positions(characters_filled_in);
         uc16 c = quarks[i];
         if (c > char_mask) {
           // If we expect a non-ASCII character from an ASCII string,
           // there is no way we can match. Not even case independent
           // matching can turn an ASCII character into non-ASCII or
           // vice versa.
           details->set_cannot_match();
@@ skipping 41 matching lines @@
         }
         characters_filled_in++;
         ASSERT(characters_filled_in <= details->characters());
         if (characters_filled_in == details->characters()) {
           return;
         }
       }
     } else {
       QuickCheckDetails::Position* pos =
           details->positions(characters_filled_in);
-      RegExpCharacterClass* tree = elm.data.u_char_class;
+      RegExpCharacterClass* tree = elm.char_class();
       ZoneList<CharacterRange>* ranges = tree->ranges(zone());
       if (tree->is_negated()) {
         // A quick check uses multi-character mask and compare.  There is no
         // useful way to incorporate a negative char class into this scheme
         // so we just conservatively create a mask and value that will always
         // succeed.
         pos->mask = 0;
         pos->value = 0;
       } else {
         int first_range = 0;
@@ skipping 169 matching lines @@
 
 
 RegExpNode* TextNode::FilterASCII(int depth, bool ignore_case) {
   if (info()->replacement_calculated) return replacement();
   if (depth < 0) return this;
   ASSERT(!info()->visited);
   VisitMarker marker(info());
   int element_count = elms_->length();
   for (int i = 0; i < element_count; i++) {
     TextElement elm = elms_->at(i);
-    if (elm.text_type == TextElement::ATOM) {
-      Vector<const uc16> quarks = elm.data.u_atom->data();
+    if (elm.text_type() == TextElement::ATOM) {
+      Vector<const uc16> quarks = elm.atom()->data();
       for (int j = 0; j < quarks.length(); j++) {
         uint16_t c = quarks[j];
         if (c <= String::kMaxOneByteCharCode) continue;
         if (!ignore_case) return set_replacement(NULL);
         // Here, we need to check for characters whose upper and lower cases
         // are outside the Latin-1 range.
         uint16_t converted = unibrow::Latin1::ConvertNonLatin1ToLatin1(c);
         // Character is outside Latin-1 completely
         if (converted == 0) return set_replacement(NULL);
         // Convert quark to Latin-1 in place.
         uint16_t* copy = const_cast<uint16_t*>(quarks.start());
         copy[j] = converted;
       }
     } else {
-      ASSERT(elm.text_type == TextElement::CHAR_CLASS);
-      RegExpCharacterClass* cc = elm.data.u_char_class;
+      ASSERT(elm.text_type() == TextElement::CHAR_CLASS);
+      RegExpCharacterClass* cc = elm.char_class();
       ZoneList<CharacterRange>* ranges = cc->ranges(zone());
       if (!CharacterRange::IsCanonical(ranges)) {
         CharacterRange::Canonicalize(ranges);
       }
       // Now they are in order so we only need to look at the first.
       int range_count = ranges->length();
       if (cc->is_negated()) {
         if (range_count != 0 &&
             ranges->at(0).from() == 0 &&
             ranges->at(0).to() >= String::kMaxOneByteCharCode) {
@@ skipping 404 matching lines @@
                             bool first_element_checked,
                             int* checked_up_to) {
   Isolate* isolate = Isolate::Current();
   RegExpMacroAssembler* assembler = compiler->macro_assembler();
   bool ascii = compiler->ascii();
   Label* backtrack = trace->backtrack();
   QuickCheckDetails* quick_check = trace->quick_check_performed();
   int element_count = elms_->length();
   for (int i = preloaded ? 0 : element_count - 1; i >= 0; i--) {
     TextElement elm = elms_->at(i);
-    int cp_offset = trace->cp_offset() + elm.cp_offset;
-    if (elm.text_type == TextElement::ATOM) {
-      Vector<const uc16> quarks = elm.data.u_atom->data();
+    int cp_offset = trace->cp_offset() + elm.cp_offset();
+    if (elm.text_type() == TextElement::ATOM) {
+      Vector<const uc16> quarks = elm.atom()->data();
       for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) {
         if (first_element_checked && i == 0 && j == 0) continue;
-        if (DeterminedAlready(quick_check, elm.cp_offset + j)) continue;
+        if (DeterminedAlready(quick_check, elm.cp_offset() + j)) continue;
         EmitCharacterFunction* emit_function = NULL;
         switch (pass) {
           case NON_ASCII_MATCH:
             ASSERT(ascii);
             if (quarks[j] > String::kMaxOneByteCharCode) {
               assembler->GoTo(backtrack);
               return;
             }
             break;
           case NON_LETTER_CHARACTER_MATCH:
@@ skipping 13 matching lines @@
                                              compiler,
                                              quarks[j],
                                              backtrack,
                                              cp_offset + j,
                                              *checked_up_to < cp_offset + j,
                                              preloaded);
           if (bound_checked) UpdateBoundsCheck(cp_offset + j, checked_up_to);
         }
       }
     } else {
-      ASSERT_EQ(elm.text_type, TextElement::CHAR_CLASS);
+      ASSERT_EQ(TextElement::CHAR_CLASS, elm.text_type());
       if (pass == CHARACTER_CLASS_MATCH) {
         if (first_element_checked && i == 0) continue;
-        if (DeterminedAlready(quick_check, elm.cp_offset)) continue;
-        RegExpCharacterClass* cc = elm.data.u_char_class;
+        if (DeterminedAlready(quick_check, elm.cp_offset())) continue;
+        RegExpCharacterClass* cc = elm.char_class();
         EmitCharClass(assembler,
                       cc,
                       ascii,
                       backtrack,
                       cp_offset,
                       *checked_up_to < cp_offset,
                       preloaded,
                       zone());
         UpdateBoundsCheck(cp_offset, checked_up_to);
       }
     }
   }
 }
 
 
 int TextNode::Length() {
   TextElement elm = elms_->last();
-  ASSERT(elm.cp_offset >= 0);
-  if (elm.text_type == TextElement::ATOM) {
-    return elm.cp_offset + elm.data.u_atom->data().length();
-  } else {
-    return elm.cp_offset + 1;
-  }
+  ASSERT(elm.cp_offset() >= 0);
+  return elm.cp_offset() + elm.length();
 }
 
 
 bool TextNode::SkipPass(int int_pass, bool ignore_case) {
   TextEmitPassType pass = static_cast<TextEmitPassType>(int_pass);
   if (ignore_case) {
     return pass == SIMPLE_CHARACTER_MATCH;
   } else {
     return pass == NON_LETTER_CHARACTER_MATCH || pass == CASE_CHARACTER_MATCH;
   }
@@ skipping 81 matching lines @@
     cp_offset_ = 0;
   }
   bound_checked_up_to_ = Max(0, bound_checked_up_to_ - by);
 }
 
 
 void TextNode::MakeCaseIndependent(bool is_ascii) {
   int element_count = elms_->length();
   for (int i = 0; i < element_count; i++) {
     TextElement elm = elms_->at(i);
-    if (elm.text_type == TextElement::CHAR_CLASS) {
-      RegExpCharacterClass* cc = elm.data.u_char_class;
+    if (elm.text_type() == TextElement::CHAR_CLASS) {
+      RegExpCharacterClass* cc = elm.char_class();
       // None of the standard character classes is different in the case
       // independent case and it slows us down if we don't know that.
       if (cc->is_standard(zone())) continue;
       ZoneList<CharacterRange>* ranges = cc->ranges(zone());
       int range_count = ranges->length();
       for (int j = 0; j < range_count; j++) {
         ranges->at(j).AddCaseEquivalents(ranges, is_ascii, zone());
       }
     }
   }
 }
 
 
 int TextNode::GreedyLoopTextLength() {
   TextElement elm = elms_->at(elms_->length() - 1);
-  if (elm.text_type == TextElement::CHAR_CLASS) {
-    return elm.cp_offset + 1;
-  } else {
-    return elm.cp_offset + elm.data.u_atom->data().length();
-  }
+  return elm.cp_offset() + elm.length();
 }
 
 
 RegExpNode* TextNode::GetSuccessorOfOmnivorousTextNode(
     RegExpCompiler* compiler) {
   if (elms_->length() != 1) return NULL;
   TextElement elm = elms_->at(0);
-  if (elm.text_type != TextElement::CHAR_CLASS) return NULL;
-  RegExpCharacterClass* node = elm.data.u_char_class;
+  if (elm.text_type() != TextElement::CHAR_CLASS) return NULL;
+  RegExpCharacterClass* node = elm.char_class();
   ZoneList<CharacterRange>* ranges = node->ranges(zone());
   if (!CharacterRange::IsCanonical(ranges)) {
     CharacterRange::Canonicalize(ranges);
   }
   if (node->is_negated()) {
     return ranges->length() == 0 ? on_success() : NULL;
   }
   if (ranges->length() != 1) return NULL;
   uint32_t max_char;
   if (compiler->ascii()) {
@@ skipping 1053 matching lines @@
   }
 }
 
 
 void DotPrinter::VisitText(TextNode* that) {
   Zone* zone = that->zone();
   stream()->Add("  n%p [label=\"", that);
   for (int i = 0; i < that->elements()->length(); i++) {
     if (i > 0) stream()->Add(" ");
     TextElement elm = that->elements()->at(i);
-    switch (elm.text_type) {
+    switch (elm.text_type()) {
       case TextElement::ATOM: {
-        stream()->Add("'%w'", elm.data.u_atom->data());
+        stream()->Add("'%w'", elm.atom()->data());
         break;
       }
       case TextElement::CHAR_CLASS: {
-        RegExpCharacterClass* node = elm.data.u_char_class;
+        RegExpCharacterClass* node = elm.char_class();
         stream()->Add("[");
         if (node->is_negated())
           stream()->Add("^");
         for (int j = 0; j < node->ranges(zone)->length(); j++) {
           CharacterRange range = node->ranges(zone)->at(j);
           stream()->Add("%k-%k", range.from(), range.to());
         }
         stream()->Add("]");
         break;
       }
@@ skipping 1161 matching lines @@
 }
 
 
 void TextNode::CalculateOffsets() {
   int element_count = elements()->length();
   // Set up the offsets of the elements relative to the start.  This is a fixed
   // quantity since a TextNode can only contain fixed-width things.
   int cp_offset = 0;
   for (int i = 0; i < element_count; i++) {
     TextElement& elm = elements()->at(i);
-    elm.cp_offset = cp_offset;
-    if (elm.text_type == TextElement::ATOM) {
-      cp_offset += elm.data.u_atom->data().length();
-    } else {
-      cp_offset++;
-    }
+    elm.set_cp_offset(cp_offset);
+    cp_offset += elm.length();
   }
 }
 
 
 void Analysis::VisitText(TextNode* that) {
   if (ignore_case_) {
     that->MakeCaseIndependent(is_ascii_);
   }
   EnsureAnalyzed(that->on_success());
   if (!has_failed()) {
@@ skipping 95 matching lines @@
                             bool not_at_start) {
   if (initial_offset >= bm->length()) return;
   int offset = initial_offset;
   int max_char = bm->max_char();
   for (int i = 0; i < elements()->length(); i++) {
     if (offset >= bm->length()) {
       if (initial_offset == 0) set_bm_info(not_at_start, bm);
       return;
     }
     TextElement text = elements()->at(i);
-    if (text.text_type == TextElement::ATOM) {
-      RegExpAtom* atom = text.data.u_atom;
+    if (text.text_type() == TextElement::ATOM) {
+      RegExpAtom* atom = text.atom();
       for (int j = 0; j < atom->length(); j++, offset++) {
         if (offset >= bm->length()) {
           if (initial_offset == 0) set_bm_info(not_at_start, bm);
           return;
         }
         uc16 character = atom->data()[j];
         if (bm->compiler()->ignore_case()) {
           unibrow::uchar chars[unibrow::Ecma262UnCanonicalize::kMaxWidth];
           int length = GetCaseIndependentLetters(
               ISOLATE,
               character,
               bm->max_char() == String::kMaxOneByteCharCode,
               chars);
           for (int j = 0; j < length; j++) {
             bm->Set(offset, chars[j]);
           }
         } else {
           if (character <= max_char) bm->Set(offset, character);
         }
       }
     } else {
-      ASSERT(text.text_type == TextElement::CHAR_CLASS);
-      RegExpCharacterClass* char_class = text.data.u_char_class;
+      ASSERT_EQ(TextElement::CHAR_CLASS, text.text_type());
+      RegExpCharacterClass* char_class = text.char_class();
       ZoneList<CharacterRange>* ranges = char_class->ranges(zone());
       if (char_class->is_negated()) {
         bm->SetAll(offset);
       } else {
         for (int k = 0; k < ranges->length(); k++) {
           CharacterRange& range = ranges->at(k);
           if (range.from() > max_char) continue;
           int to = Min(max_char, static_cast<int>(range.to()));
           bm->SetInterval(offset, Interval(range.from(), to));
         }
@@ skipping 91 matching lines @@
         last = range.to() + 1;
       }
     }
   }
   AddRange(CharacterRange(last, String::kMaxUtf16CodeUnit));
 }
 
 
 void DispatchTableConstructor::VisitText(TextNode* that) {
   TextElement elm = that->elements()->at(0);
-  switch (elm.text_type) {
+  switch (elm.text_type()) {
     case TextElement::ATOM: {
-      uc16 c = elm.data.u_atom->data()[0];
+      uc16 c = elm.atom()->data()[0];
       AddRange(CharacterRange(c, c));
       break;
     }
     case TextElement::CHAR_CLASS: {
-      RegExpCharacterClass* tree = elm.data.u_char_class;
+      RegExpCharacterClass* tree = elm.char_class();
       ZoneList<CharacterRange>* ranges = tree->ranges(that->zone());
       if (tree->is_negated()) {
         AddInverse(ranges);
       } else {
         for (int i = 0; i < ranges->length(); i++)
           AddRange(ranges->at(i));
       }
       break;
     }
     default: {
@@ skipping 131 matching lines @@
   }
 
   return compiler.Assemble(&macro_assembler,
                            node,
                            data->capture_count,
                            pattern);
 }
 
 
 }}  // namespace v8::internal