[regexp] correctly advance zero length matches for global/unicode.

src/regexp/jsregexp.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/regexp/jsregexp.h"
 
 #include "src/ast/ast.h"
 #include "src/base/platform/platform.h"
 #include "src/compilation-cache.h"
 #include "src/compiler.h"
@@ skipping 620 matching lines @@
   }
   SetLastCaptureCount(array, capture_register_count);
   SetLastSubject(array, *subject);
   SetLastInput(array, *subject);
   return last_match_info;
 }
 
 
 RegExpImpl::GlobalCache::GlobalCache(Handle<JSRegExp> regexp,
                                      Handle<String> subject,
-                                     bool is_global,
                                      Isolate* isolate)
   : register_array_(NULL),
     register_array_size_(0),
     regexp_(regexp),
     subject_(subject) {
 #ifdef V8_INTERPRETED_REGEXP
   bool interpreted = true;
 #else
   bool interpreted = false;
 #endif  // V8_INTERPRETED_REGEXP
 
   if (regexp_->TypeTag() == JSRegExp::ATOM) {
     static const int kAtomRegistersPerMatch = 2;
     registers_per_match_ = kAtomRegistersPerMatch;
     // There is no distinction between interpreted and native for atom regexps.
     interpreted = false;
   } else {
     registers_per_match_ = RegExpImpl::IrregexpPrepare(regexp_, subject_);
     if (registers_per_match_ < 0) {
       num_matches_ = -1;  // Signal exception.
       return;
     }
   }
 
-  if (is_global && !interpreted) {
+  DCHECK_NE(0, regexp->GetFlags() & JSRegExp::kGlobal);
+  if (!interpreted) {
     register_array_size_ =
         Max(registers_per_match_, Isolate::kJSRegexpStaticOffsetsVectorSize);
     max_matches_ = register_array_size_ / registers_per_match_;
   } else {
     // Global loop in interpreted regexp is not implemented.  We choose
     // the size of the offsets vector so that it can only store one match.
     register_array_size_ = registers_per_match_;
     max_matches_ = 1;
   }
 
   if (register_array_size_ > Isolate::kJSRegexpStaticOffsetsVectorSize) {
     register_array_ = NewArray<int32_t>(register_array_size_);
   } else {
     register_array_ = isolate->jsregexp_static_offsets_vector();
   }
 
   // Set state so that fetching the results the first time triggers a call
   // to the compiled regexp.
   current_match_index_ = max_matches_ - 1;
   num_matches_ = max_matches_;
   DCHECK(registers_per_match_ >= 2);  // Each match has at least one capture.
   DCHECK_GE(register_array_size_, registers_per_match_);
   int32_t* last_match =
       &register_array_[current_match_index_ * registers_per_match_];
   last_match[0] = -1;
   last_match[1] = 0;
 }
 
+int RegExpImpl::GlobalCache::AdvanceZeroLength(int last_index) {
+  if ((regexp_->GetFlags() & JSRegExp::kUnicode) != 0 &&
+      last_index + 1 < subject_->length() &&
+      unibrow::Utf16::IsLeadSurrogate(subject_->Get(last_index)) &&
+      unibrow::Utf16::IsTrailSurrogate(subject_->Get(last_index + 1))) {
+    // Advance over the surrogate pair.
+    return last_index + 2;
+  }
+  return last_index + 1;
+}
 
 // -------------------------------------------------------------------
 // Implementation of the Irregexp regular expression engine.
 //
 // The Irregexp regular expression engine is intended to be a complete
 // implementation of ECMAScript regular expressions.  It generates either
 // bytecodes or native code.
 
 //   The Irregexp regexp engine is structured in three steps.
 //   1) The parser generates an abstract syntax tree.  See ast.cc.
@@ skipping 5911 matching lines @@
 RegExpEngine::CompilationResult RegExpEngine::Compile(
     Isolate* isolate, Zone* zone, RegExpCompileData* data,
     JSRegExp::Flags flags, Handle<String> pattern,
     Handle<String> sample_subject, bool is_one_byte) {
   if ((data->capture_count + 1) * 2 - 1 > RegExpMacroAssembler::kMaxRegister) {
     return IrregexpRegExpTooBig(isolate);
   }
   bool ignore_case = flags & JSRegExp::kIgnoreCase;
   bool is_sticky = flags & JSRegExp::kSticky;
   bool is_global = flags & JSRegExp::kGlobal;
+  bool is_unicode = flags & JSRegExp::kUnicode;
   RegExpCompiler compiler(isolate, zone, data->capture_count, flags,
                           is_one_byte);
 
   if (compiler.optimize()) compiler.set_optimize(!TooMuchRegExpCode(pattern));
 
   // Sample some characters from the middle of the string.
   static const int kSampleSize = 128;
 
   sample_subject = String::Flatten(sample_subject);
   int chars_sampled = 0;
@@ skipping 99 matching lines @@
   // Inserted here, instead of in Assembler, because it depends on information
   // in the AST that isn't replicated in the Node structure.
   static const int kMaxBacksearchLimit = 1024;
   if (is_end_anchored &&
       !is_start_anchored &&
       max_length < kMaxBacksearchLimit) {
     macro_assembler.SetCurrentPositionFromEnd(max_length);
   }
 
   if (is_global) {
-    macro_assembler.set_global_mode(
-        (data->tree->min_match() > 0)
-            ? RegExpMacroAssembler::GLOBAL_NO_ZERO_LENGTH_CHECK
-            : RegExpMacroAssembler::GLOBAL);
+    RegExpMacroAssembler::GlobalMode mode = RegExpMacroAssembler::GLOBAL;
+    if (data->tree->min_match() > 0) {
+      mode = RegExpMacroAssembler::GLOBAL_NO_ZERO_LENGTH_CHECK;
+    } else if (is_unicode) {
+      mode = RegExpMacroAssembler::GLOBAL_UNICODE;
+    }
+    macro_assembler.set_global_mode(mode);
   }
 
   return compiler.Assemble(&macro_assembler,
                            node,
                            data->capture_count,
                            pattern);
 }
 
 
 bool RegExpEngine::TooMuchRegExpCode(Handle<String> pattern) {
@@ skipping 102 matching lines @@
 
 
 void RegExpResultsCache::Clear(FixedArray* cache) {
   for (int i = 0; i < kRegExpResultsCacheSize; i++) {
     cache->set(i, Smi::FromInt(0));
   }
 }
 
 }  // namespace internal
 }  // namespace v8