[regexp] Use ZoneVector in parser and compiler

For a variety of reasons related to OOM handling and custom
allocators, SpiderMonkey wants to be able to see all memory
allocations. To enforce this, we have a static analysis that verifies
that we don't link in malloc/new/etc in unexpected places. One
consequence of this is that we can't use STL containers without a
custom allocator, because they call operator new internally.

This is mostly not an issue in irregexp, which makes heavy use of zone
allocation. The main exceptions are a handful of uses of std::vector
in regexp-compiler.* and regexp-parser.*. If these vectors are
converted to ZoneVectors, then our static analysis is satisfied.

R=jgruber@chromium.org

Bug: v8:10303
Change-Id: I8b14a2eb54d3b20959e3fbe878f77effae124a2c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2091402Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#66674}

src/regexp/regexp-compiler.cc

@@ -252,7 +252,7 @@ RegExpCompiler::CompilationResult RegExpCompiler::Assemble(
 #endif
     macro_assembler_ = macro_assembler;
 
-  std::vector<RegExpNode*> work_list;
+  ZoneVector<RegExpNode*> work_list(zone());
   work_list_ = &work_list;
   Label fail;
   macro_assembler_->PushBacktrack(&fail);

src/regexp/regexp-compiler.h

@@ -576,7 +576,7 @@ class RegExpCompiler {
   int next_register_;
   int unicode_lookaround_stack_register_;
   int unicode_lookaround_position_register_;
-  std::vector<RegExpNode*>* work_list_;
+  ZoneVector<RegExpNode*>* work_list_;
   int recursion_depth_;
   RegExpMacroAssembler* macro_assembler_;
   bool one_byte_;

src/regexp/regexp-parser.cc

@@ -347,7 +347,8 @@ RegExpTree* RegExpParser::ParseDisjunction() {
             if (unicode()) {
               ZoneList<CharacterRange>* ranges =
                   new (zone()) ZoneList<CharacterRange>(2, zone());
-              std::vector<char> name_1, name_2;
+              ZoneVector<char> name_1(zone());
+              ZoneVector<char> name_2(zone());
               if (ParsePropertyClassName(&name_1, &name_2)) {
                 if (AddPropertyClassRange(ranges, p == 'P', name_1, name_2)) {
                   RegExpCharacterClass* cc = new (zone())
@@ -1385,8 +1386,8 @@ bool IsUnicodePropertyValueCharacter(char c) {
 
 }  // anonymous namespace
 
-bool RegExpParser::ParsePropertyClassName(std::vector<char>* name_1,
-                                          std::vector<char>* name_2) {
+bool RegExpParser::ParsePropertyClassName(ZoneVector<char>* name_1,
+                                          ZoneVector<char>* name_2) {
   DCHECK(name_1->empty());
   DCHECK(name_2->empty());
   // Parse the property class as follows:
@@ -1425,8 +1426,8 @@ bool RegExpParser::ParsePropertyClassName(std::vector<char>* name_1,
 
 bool RegExpParser::AddPropertyClassRange(ZoneList<CharacterRange>* add_to,
                                          bool negate,
-                                         const std::vector<char>& name_1,
-                                         const std::vector<char>& name_2) {
+                                         const ZoneVector<char>& name_1,
+                                         const ZoneVector<char>& name_2) {
   if (name_2.empty()) {
     // First attempt to interpret as general category property value name.
     const char* name = name_1.data();
@@ -1463,7 +1464,7 @@ bool RegExpParser::AddPropertyClassRange(ZoneList<CharacterRange>* add_to,
   }
 }
 
-RegExpTree* RegExpParser::GetPropertySequence(const std::vector<char>& name_1) {
+RegExpTree* RegExpParser::GetPropertySequence(const ZoneVector<char>& name_1) {
   if (!FLAG_harmony_regexp_sequence) return nullptr;
   const char* name = name_1.data();
   const uc32* sequence_list = nullptr;
@@ -1529,19 +1530,19 @@ RegExpTree* RegExpParser::GetPropertySequence(const std::vector<char>& name_1) {
 
 #else  // V8_INTL_SUPPORT
 
-bool RegExpParser::ParsePropertyClassName(std::vector<char>* name_1,
-                                          std::vector<char>* name_2) {
+bool RegExpParser::ParsePropertyClassName(ZoneVector<char>* name_1,
+                                          ZoneVector<char>* name_2) {
   return false;
 }
 
 bool RegExpParser::AddPropertyClassRange(ZoneList<CharacterRange>* add_to,
                                          bool negate,
-                                         const std::vector<char>& name_1,
-                                         const std::vector<char>& name_2) {
+                                         const ZoneVector<char>& name_1,
+                                         const ZoneVector<char>& name_2) {
   return false;
 }
 
-RegExpTree* RegExpParser::GetPropertySequence(const std::vector<char>& name) {
+RegExpTree* RegExpParser::GetPropertySequence(const ZoneVector<char>& name) {
   return nullptr;
 }
 
@@ -1710,7 +1711,8 @@ void RegExpParser::ParseClassEscape(ZoneList<CharacterRange>* ranges,
         if (unicode()) {
           bool negate = Next() == 'P';
           Advance(2);
-          std::vector<char> name_1, name_2;
+          ZoneVector<char> name_1(zone);
+          ZoneVector<char> name_2(zone);
           if (!ParsePropertyClassName(&name_1, &name_2) ||
               !AddPropertyClassRange(ranges, negate, name_1, name_2)) {
             ReportError(CStrVector("Invalid property name in character class"));

src/regexp/regexp-parser.h

@@ -177,13 +177,13 @@ class V8_EXPORT_PRIVATE RegExpParser {
   bool ParseUnicodeEscape(uc32* value);
   bool ParseUnlimitedLengthHexNumber(int max_value, uc32* value);
 
-  bool ParsePropertyClassName(std::vector<char>* name_1,
-                              std::vector<char>* name_2);
+  bool ParsePropertyClassName(ZoneVector<char>* name_1,
+                              ZoneVector<char>* name_2);
   bool AddPropertyClassRange(ZoneList<CharacterRange>* add_to, bool negate,
-                             const std::vector<char>& name_1,
-                             const std::vector<char>& name_2);
+                             const ZoneVector<char>& name_1,
+                             const ZoneVector<char>& name_2);
 
-  RegExpTree* GetPropertySequence(const std::vector<char>& name_1);
+  RegExpTree* GetPropertySequence(const ZoneVector<char>& name_1);
   RegExpTree* ParseCharacterClass(const RegExpBuilder* state);
 
   uc32 ParseOctalLiteral();