JavaScriptCore-7601.1.46.3.tar.gz

[apple/javascriptcore.git] / runtime / RegExp.cpp

diff --git a/runtime/RegExp.cpp b/runtime/RegExp.cpp
index 59f1f87f7747570ef4b2e1842d6bda3bedd4d5ac..af1f2fa1da6398bad6af3263d311e39e11c54d16 100644 (file)
--- a/runtime/RegExp.cpp
+++ b/runtime/RegExp.cpp
@@ -2,6 +2,7 @@
  *  Copyright (C) 1999-2001, 2004 Harri Porten (porten@kde.org)
  *  Copyright (c) 2007, 2008 Apple Inc. All rights reserved.
  *  Copyright (C) 2009 Torch Mobile, Inc.
  *  Copyright (C) 1999-2001, 2004 Harri Porten (porten@kde.org)
  *  Copyright (c) 2007, 2008 Apple Inc. All rights reserved.
  *  Copyright (C) 2009 Torch Mobile, Inc.

+ *  Copyright (C) 2010 Peter Varga (pvarga@inf.u-szeged.hu), University of Szeged

  *
  *  This library is free software; you can redistribute it and/or
  *  modify it under the terms of the GNU Lesser General Public
  *
  *  This library is free software; you can redistribute it and/or
  *  modify it under the terms of the GNU Lesser General Public


@@ -21,223 +22,572 @@
 
 #include "config.h"
 #include "RegExp.h"
 
 #include "config.h"
 #include "RegExp.h"

+

 #include "Lexer.h"
 #include "Lexer.h"

+#include "JSCInlines.h"
+#include "RegExpCache.h"
+#include "Yarr.h"
+#include "YarrJIT.h"
+#include <wtf/Assertions.h>
+
+#define REGEXP_FUNC_TEST_DATA_GEN 0
+
+#if REGEXP_FUNC_TEST_DATA_GEN

 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>

-#include <wtf/Assertions.h>
-#include <wtf/OwnArrayPtr.h>
+#endif

 
 

+namespace JSC {

 
 

-#if ENABLE(YARR)
+const ClassInfo RegExp::s_info = { "RegExp", 0, 0, CREATE_METHOD_TABLE(RegExp) };

 
 

-#include "yarr/RegexCompiler.h"
-#if ENABLE(YARR_JIT)
-#include "yarr/RegexJIT.h"
-#else
-#include "yarr/RegexInterpreter.h"
-#endif
+RegExpFlags regExpFlags(const String& string)
+{
+    RegExpFlags flags = NoFlags;
+
+    for (unsigned i = 0; i < string.length(); ++i) {
+        switch (string[i]) {
+        case 'g':
+            if (flags & FlagGlobal)
+                return InvalidFlags;
+            flags = static_cast<RegExpFlags>(flags | FlagGlobal);
+            break;
+
+        case 'i':
+            if (flags & FlagIgnoreCase)
+                return InvalidFlags;
+            flags = static_cast<RegExpFlags>(flags | FlagIgnoreCase);
+            break;
+
+        case 'm':
+            if (flags & FlagMultiline)
+                return InvalidFlags;
+            flags = static_cast<RegExpFlags>(flags | FlagMultiline);
+            break;
+
+        default:
+            return InvalidFlags;
+        }
+    }

 
 

-#else
+    return flags;
+}

 
 

-#include <pcre/pcre.h>
+#if REGEXP_FUNC_TEST_DATA_GEN
+class RegExpFunctionalTestCollector {
+    // This class is not thread safe.
+protected:
+    static const char* const s_fileName;

 
 

-#endif
+public:
+    static RegExpFunctionalTestCollector* get();

 
 

-namespace JSC {
+    ~RegExpFunctionalTestCollector();

 
 

-inline RegExp::RegExp(JSGlobalData* globalData, const UString& pattern)
-    : m_pattern(pattern)
-    , m_flagBits(0)
-    , m_constructionError(0)
-    , m_numSubpatterns(0)
-    , m_lastMatchStart(-1)
+    void outputOneTest(RegExp*, String, int, int*, int);
+    void clearRegExp(RegExp* regExp)
+    {
+        if (regExp == m_lastRegExp)
+            m_lastRegExp = 0;
+    }
+
+private:
+    RegExpFunctionalTestCollector();
+
+    void outputEscapedString(const String&, bool escapeSlash = false);
+
+    static RegExpFunctionalTestCollector* s_instance;
+    FILE* m_file;
+    RegExp* m_lastRegExp;
+};
+
+const char* const RegExpFunctionalTestCollector::s_fileName = "/tmp/RegExpTestsData";
+RegExpFunctionalTestCollector* RegExpFunctionalTestCollector::s_instance = 0;
+
+RegExpFunctionalTestCollector* RegExpFunctionalTestCollector::get()

 {
 {

-    compile(globalData);
+    if (!s_instance)
+        s_instance = new RegExpFunctionalTestCollector();
+
+    return s_instance;

 }
 
 }
 

-inline RegExp::RegExp(JSGlobalData* globalData, const UString& pattern, const UString& flags)
-    : m_pattern(pattern)
-    , m_flagBits(0)
-    , m_constructionError(0)
-    , m_numSubpatterns(0)
-    , m_lastMatchStart(-1)
+void RegExpFunctionalTestCollector::outputOneTest(RegExp* regExp, const String& s, int startOffset, int* ovector, int result)

 {
 {

-    // NOTE: The global flag is handled on a case-by-case basis by functions like
-    // String::match and RegExpObject::match.
-    if (flags.find('g') != UString::NotFound)
-        m_flagBits |= Global;
-    if (flags.find('i') != UString::NotFound)
-        m_flagBits |= IgnoreCase;
-    if (flags.find('m') != UString::NotFound)
-        m_flagBits |= Multiline;
+    if ((!m_lastRegExp) || (m_lastRegExp != regExp)) {
+        m_lastRegExp = regExp;
+        fputc('/', m_file);
+        outputEscapedString(regExp->pattern(), true);
+        fputc('/', m_file);
+        if (regExp->global())
+            fputc('g', m_file);
+        if (regExp->ignoreCase())
+            fputc('i', m_file);
+        if (regExp->multiline())
+            fputc('m', m_file);
+        fprintf(m_file, "\n");
+    }

 
 

-    compile(globalData);
+    fprintf(m_file, " \"");
+    outputEscapedString(s);
+    fprintf(m_file, "\", %d, %d, (", startOffset, result);
+    for (unsigned i = 0; i <= regExp->numSubpatterns(); i++) {
+        int subpatternBegin = ovector[i * 2];
+        int subpatternEnd = ovector[i * 2 + 1];
+        if (subpatternBegin == -1)
+            subpatternEnd = -1;
+        fprintf(m_file, "%d, %d", subpatternBegin, subpatternEnd);
+        if (i < regExp->numSubpatterns())
+            fputs(", ", m_file);
+    }
+
+    fprintf(m_file, ")\n");
+    fflush(m_file);

 }
 
 }
 

-#if !ENABLE(YARR)
-RegExp::~RegExp()
+RegExpFunctionalTestCollector::RegExpFunctionalTestCollector()

 {
 {

-    jsRegExpFree(m_regExp);
+    m_file = fopen(s_fileName, "r+");
+    if  (!m_file)
+        m_file = fopen(s_fileName, "w+");
+
+    fseek(m_file, 0L, SEEK_END);

 }
 }

-#endif

 
 

-PassRefPtr<RegExp> RegExp::create(JSGlobalData* globalData, const UString& pattern)
+RegExpFunctionalTestCollector::~RegExpFunctionalTestCollector()

 {
 {

-    return adoptRef(new RegExp(globalData, pattern));
+    fclose(m_file);
+    s_instance = 0;

 }
 
 }
 

-PassRefPtr<RegExp> RegExp::create(JSGlobalData* globalData, const UString& pattern, const UString& flags)
+void RegExpFunctionalTestCollector::outputEscapedString(const String& s, bool escapeSlash)
+{
+    int len = s.length();
+    
+    for (int i = 0; i < len; ++i) {
+        UChar c = s[i];
+
+        switch (c) {
+        case '\0':
+            fputs("\\0", m_file);
+            break;
+        case '\a':
+            fputs("\\a", m_file);
+            break;
+        case '\b':
+            fputs("\\b", m_file);
+            break;
+        case '\f':
+            fputs("\\f", m_file);
+            break;
+        case '\n':
+            fputs("\\n", m_file);
+            break;
+        case '\r':
+            fputs("\\r", m_file);
+            break;
+        case '\t':
+            fputs("\\t", m_file);
+            break;
+        case '\v':
+            fputs("\\v", m_file);
+            break;
+        case '/':
+            if (escapeSlash)
+                fputs("\\/", m_file);
+            else
+                fputs("/", m_file);
+            break;
+        case '\"':
+            fputs("\\\"", m_file);
+            break;
+        case '\\':
+            fputs("\\\\", m_file);
+            break;
+        case '\?':
+            fputs("\?", m_file);
+            break;
+        default:
+            if (c > 0x7f)
+                fprintf(m_file, "\\u%04x", c);
+            else 
+                fputc(c, m_file);
+            break;
+        }
+    }
+}
+#endif
+
+RegExp::RegExp(VM& vm, const String& patternString, RegExpFlags flags)
+    : JSCell(vm, vm.regExpStructure.get())
+    , m_state(NotCompiled)
+    , m_patternString(patternString)
+    , m_flags(flags)
+    , m_constructionError(0)
+    , m_numSubpatterns(0)
+#if ENABLE(REGEXP_TRACING)
+    , m_rtMatchOnlyTotalSubjectStringLen(0.0)
+    , m_rtMatchTotalSubjectStringLen(0.0)
+    , m_rtMatchOnlyCallCount(0)
+    , m_rtMatchOnlyFoundCount(0)
+    , m_rtMatchCallCount(0)
+    , m_rtMatchFoundCount(0)
+#endif

 {
 {

-    return adoptRef(new RegExp(globalData, pattern, flags));

 }
 
 }
 

-#if ENABLE(YARR)
+void RegExp::finishCreation(VM& vm)
+{
+    Base::finishCreation(vm);
+    Yarr::YarrPattern pattern(m_patternString, ignoreCase(), multiline(), &m_constructionError);
+    if (m_constructionError)
+        m_state = ParseError;
+    else
+        m_numSubpatterns = pattern.m_numSubpatterns;
+}

 
 

-void RegExp::compile(JSGlobalData* globalData)
+void RegExp::destroy(JSCell* cell)

 {
 {

-#if ENABLE(YARR_JIT)
-    Yarr::jitCompileRegex(globalData, m_regExpJITCode, m_pattern, m_numSubpatterns, m_constructionError, ignoreCase(), multiline());
-#else
-    UNUSED_PARAM(globalData);
-    m_regExpBytecode.set(Yarr::byteCompileRegex(m_pattern, m_numSubpatterns, m_constructionError, ignoreCase(), multiline()));
+    RegExp* thisObject = static_cast<RegExp*>(cell);
+#if REGEXP_FUNC_TEST_DATA_GEN
+    RegExpFunctionalTestCollector::get()->clearRegExp(this);

 #endif
 #endif

+    thisObject->RegExp::~RegExp();
+}
+
+RegExp* RegExp::createWithoutCaching(VM& vm, const String& patternString, RegExpFlags flags)
+{
+    RegExp* regExp = new (NotNull, allocateCell<RegExp>(vm.heap)) RegExp(vm, patternString, flags);
+    regExp->finishCreation(vm);
+    return regExp;

 }
 
 }
 

-int RegExp::match(const UString& s, int startOffset, Vector<int, 32>* ovector)
+RegExp* RegExp::create(VM& vm, const String& patternString, RegExpFlags flags)

 {
 {

-    if (startOffset < 0)
-        startOffset = 0;
-    if (ovector)
-        ovector->clear();
+    return vm.regExpCache()->lookupOrCreate(patternString, flags);
+}

 
 

-    if (static_cast<unsigned>(startOffset) > s.size() || s.isNull()) {
-        m_lastMatchString = UString();
-        m_lastMatchStart = -1;
-        m_lastOVector.shrink(0);
-        return -1;
+void RegExp::compile(VM* vm, Yarr::YarrCharSize charSize)
+{
+    Yarr::YarrPattern pattern(m_patternString, ignoreCase(), multiline(), &m_constructionError);
+    if (m_constructionError) {
+        RELEASE_ASSERT_NOT_REACHED();
+#if COMPILER_QUIRK(CONSIDERS_UNREACHABLE_CODE)
+        m_state = ParseError;
+        return;
+#endif

     }
     }

-    
-    // Perform check to see if this match call is the same as the last match invocation
-    // and if it is return the prior result.
-    if ((startOffset == m_lastMatchStart) && (s.rep() == m_lastMatchString.rep())) {
-        if (ovector)
-            *ovector = m_lastOVector;
-        
-        if (m_lastOVector.isEmpty())
-            return -1;
+    ASSERT(m_numSubpatterns == pattern.m_numSubpatterns);

 
 

-        return m_lastOVector.at(0);
+    if (!hasCode()) {
+        ASSERT(m_state == NotCompiled);
+        vm->regExpCache()->addToStrongCache(this);
+        m_state = ByteCode;

     }
 
 #if ENABLE(YARR_JIT)
     }
 
 #if ENABLE(YARR_JIT)

-    if (!!m_regExpJITCode) {
-#else
-    if (m_regExpBytecode) {
-#endif
-        int offsetVectorSize = (m_numSubpatterns + 1) * 3; // FIXME: should be 2 - but adding temporary fallback to pcre.
-        int* offsetVector;
-        Vector<int, 32> nonReturnedOvector;
-        if (ovector) {
-            ovector->resize(offsetVectorSize);
-            offsetVector = ovector->data();
-        } else {
-            nonReturnedOvector.resize(offsetVectorSize);
-            offsetVector = nonReturnedOvector.data();
+    if (!pattern.m_containsBackreferences && !pattern.containsUnsignedLengthPattern() && vm->canUseRegExpJIT()) {
+        Yarr::jitCompile(pattern, charSize, vm, m_regExpJITCode);
+        if (!m_regExpJITCode.isFallBack()) {
+            m_state = JITCode;
+            return;

         }
         }

+    }
+#else
+    UNUSED_PARAM(charSize);
+#endif

 
 

-        ASSERT(offsetVector);
-        for (int j = 0; j < offsetVectorSize; ++j)
-            offsetVector[j] = -1;
-
+    m_state = ByteCode;
+    m_regExpBytecode = Yarr::byteCompile(pattern, &vm->m_regExpAllocator);
+}

 
 

+void RegExp::compileIfNecessary(VM& vm, Yarr::YarrCharSize charSize)
+{
+    if (hasCode()) {

 #if ENABLE(YARR_JIT)
 #if ENABLE(YARR_JIT)

-        int result = Yarr::executeRegex(m_regExpJITCode, s.data(), startOffset, s.size(), offsetVector, offsetVectorSize);
+        if (m_state != JITCode)
+            return;
+        if ((charSize == Yarr::Char8) && (m_regExpJITCode.has8BitCode()))
+            return;
+        if ((charSize == Yarr::Char16) && (m_regExpJITCode.has16BitCode()))
+            return;

 #else
 #else

-        int result = Yarr::interpretRegex(m_regExpBytecode.get(), s.data(), startOffset, s.size(), offsetVector);
+        return;

 #endif
 #endif

+    }
+
+    compile(&vm, charSize);
+}

 
 

-        if (result < 0) {
-#ifndef NDEBUG
-            // TODO: define up a symbol, rather than magic -1
-            if (result != -1)
-                fprintf(stderr, "jsRegExpExecute failed with result %d\n", result);
+int RegExp::match(VM& vm, const String& s, unsigned startOffset, Vector<int, 32>& ovector)
+{
+#if ENABLE(REGEXP_TRACING)
+    m_rtMatchCallCount++;
+    m_rtMatchTotalSubjectStringLen += (double)(s.length() - startOffset);

 #endif
 #endif

-            if (ovector)
-                ovector->clear();
-        }
-        
-        m_lastMatchString = s;
-        m_lastMatchStart = startOffset;

 
 

-        if (ovector)
-            m_lastOVector = *ovector;
+    ASSERT(m_state != ParseError);
+    compileIfNecessary(vm, s.is8Bit() ? Yarr::Char8 : Yarr::Char16);
+
+    int offsetVectorSize = (m_numSubpatterns + 1) * 2;
+    ovector.resize(offsetVectorSize);
+    int* offsetVector = ovector.data();
+
+    int result;
+#if ENABLE(YARR_JIT)
+    if (m_state == JITCode) {
+        if (s.is8Bit())
+            result = m_regExpJITCode.execute(s.characters8(), startOffset, s.length(), offsetVector).start;

         else
         else

-            m_lastOVector = nonReturnedOvector;
+            result = m_regExpJITCode.execute(s.characters16(), startOffset, s.length(), offsetVector).start;
+#if ENABLE(YARR_JIT_DEBUG)
+        matchCompareWithInterpreter(s, startOffset, offsetVector, result);
+#endif
+    } else
+#endif
+        result = Yarr::interpret(m_regExpBytecode.get(), s, startOffset, reinterpret_cast<unsigned*>(offsetVector));
+
+    // FIXME: The YARR engine should handle unsigned or size_t length matches.
+    // The YARR Interpreter is "unsigned" clean, while the YARR JIT hasn't been addressed.
+    // The offset vector handling needs to change as well.
+    // Right now we convert a match where the offsets overflowed into match failure.
+    // There are two places in WebCore that call the interpreter directly that need to
+    // have their offsets changed to int as well. They are yarr/RegularExpression.cpp
+    // and inspector/ContentSearchUtilities.cpp
+    if (s.length() > INT_MAX) {
+        bool overflowed = false;
+
+        if (result < -1)
+            overflowed = true;
+
+        for (unsigned i = 0; i <= m_numSubpatterns; i++) {
+            if ((offsetVector[i*2] < -1) || ((offsetVector[i*2] >= 0) && (offsetVector[i*2+1] < -1))) {
+                overflowed = true;
+                offsetVector[i*2] = -1;
+                offsetVector[i*2+1] = -1;
+            }
+        }

 
 

-        return result;
+        if (overflowed)
+            result = -1;

     }
 
     }
 

-    m_lastMatchString = UString();
-    m_lastMatchStart = -1;
-    m_lastOVector.shrink(0);
+    ASSERT(result >= -1);

 
 

-    return -1;
+#if REGEXP_FUNC_TEST_DATA_GEN
+    RegExpFunctionalTestCollector::get()->outputOneTest(this, s, startOffset, offsetVector, result);
+#endif
+
+#if ENABLE(REGEXP_TRACING)
+    if (result != -1)
+        m_rtMatchFoundCount++;
+#endif
+
+    return result;

 }
 
 }
 

+void RegExp::compileMatchOnly(VM* vm, Yarr::YarrCharSize charSize)
+{
+    Yarr::YarrPattern pattern(m_patternString, ignoreCase(), multiline(), &m_constructionError);
+    if (m_constructionError) {
+        RELEASE_ASSERT_NOT_REACHED();
+#if COMPILER_QUIRK(CONSIDERS_UNREACHABLE_CODE)
+        m_state = ParseError;
+        return;
+#endif
+    }
+    ASSERT(m_numSubpatterns == pattern.m_numSubpatterns);
+
+    if (!hasCode()) {
+        ASSERT(m_state == NotCompiled);
+        vm->regExpCache()->addToStrongCache(this);
+        m_state = ByteCode;
+    }
+
+#if ENABLE(YARR_JIT)
+    if (!pattern.m_containsBackreferences && !pattern.containsUnsignedLengthPattern() && vm->canUseRegExpJIT()) {
+        Yarr::jitCompile(pattern, charSize, vm, m_regExpJITCode, Yarr::MatchOnly);
+        if (!m_regExpJITCode.isFallBack()) {
+            m_state = JITCode;
+            return;
+        }
+    }

 #else
 #else

+    UNUSED_PARAM(charSize);
+#endif
+
+    m_state = ByteCode;
+    m_regExpBytecode = Yarr::byteCompile(pattern, &vm->m_regExpAllocator);
+}

 
 

-void RegExp::compile(JSGlobalData*)
+void RegExp::compileIfNecessaryMatchOnly(VM& vm, Yarr::YarrCharSize charSize)

 {
 {

-    m_regExp = 0;
-    JSRegExpIgnoreCaseOption ignoreCaseOption = ignoreCase() ? JSRegExpIgnoreCase : JSRegExpDoNotIgnoreCase;
-    JSRegExpMultilineOption multilineOption = multiline() ? JSRegExpMultiline : JSRegExpSingleLine;
-    m_regExp = jsRegExpCompile(reinterpret_cast<const UChar*>(m_pattern.data()), m_pattern.size(), ignoreCaseOption, multilineOption, &m_numSubpatterns, &m_constructionError);
+    if (hasCode()) {
+#if ENABLE(YARR_JIT)
+        if (m_state != JITCode)
+            return;
+        if ((charSize == Yarr::Char8) && (m_regExpJITCode.has8BitCodeMatchOnly()))
+            return;
+        if ((charSize == Yarr::Char16) && (m_regExpJITCode.has16BitCodeMatchOnly()))
+            return;
+#else
+        return;
+#endif
+    }
+
+    compileMatchOnly(&vm, charSize);

 }
 
 }
 

-int RegExp::match(const UString& s, int startOffset, Vector<int, 32>* ovector)
+MatchResult RegExp::match(VM& vm, const String& s, unsigned startOffset)

 {
 {

-    if (startOffset < 0)
-        startOffset = 0;
-    if (ovector)
-        ovector->clear();
+#if ENABLE(REGEXP_TRACING)
+    m_rtMatchOnlyCallCount++;
+    m_rtMatchOnlyTotalSubjectStringLen += (double)(s.length() - startOffset);
+#endif

 
 

-    if (static_cast<unsigned>(startOffset) > s.size() || s.isNull())
-        return -1;
+    ASSERT(m_state != ParseError);
+    compileIfNecessaryMatchOnly(vm, s.is8Bit() ? Yarr::Char8 : Yarr::Char16);

 
 

-    if (m_regExp) {
-        // Set up the offset vector for the result.
-        // First 2/3 used for result, the last third used by PCRE.
-        int* offsetVector;
-        int offsetVectorSize;
-        int fixedSizeOffsetVector[3];
-        if (!ovector) {
-            offsetVectorSize = 3;
-            offsetVector = fixedSizeOffsetVector;
+#if ENABLE(YARR_JIT)
+    if (m_state == JITCode) {
+        MatchResult result = s.is8Bit() ?
+            m_regExpJITCode.execute(s.characters8(), startOffset, s.length()) :
+            m_regExpJITCode.execute(s.characters16(), startOffset, s.length());
+#if ENABLE(REGEXP_TRACING)
+        if (!result)
+            m_rtMatchOnlyFoundCount++;
+#endif
+        return result;
+    }
+#endif
+
+    int offsetVectorSize = (m_numSubpatterns + 1) * 2;
+    int* offsetVector;
+    Vector<int, 32> nonReturnedOvector;
+    nonReturnedOvector.resize(offsetVectorSize);
+    offsetVector = nonReturnedOvector.data();
+    int r = Yarr::interpret(m_regExpBytecode.get(), s, startOffset, reinterpret_cast<unsigned*>(offsetVector));
+#if REGEXP_FUNC_TEST_DATA_GEN
+    RegExpFunctionalTestCollector::get()->outputOneTest(this, s, startOffset, offsetVector, result);
+#endif
+
+    if (r >= 0) {
+#if ENABLE(REGEXP_TRACING)
+        m_rtMatchOnlyFoundCount++;
+#endif
+        return MatchResult(r, reinterpret_cast<unsigned*>(offsetVector)[1]);
+    }
+
+    return MatchResult::failed();
+}
+
+void RegExp::invalidateCode()
+{
+    if (!hasCode())
+        return;
+    m_state = NotCompiled;
+#if ENABLE(YARR_JIT)
+    m_regExpJITCode.clear();
+#endif
+    m_regExpBytecode = nullptr;
+}
+
+#if ENABLE(YARR_JIT_DEBUG)
+void RegExp::matchCompareWithInterpreter(const String& s, int startOffset, int* offsetVector, int jitResult)
+{
+    int offsetVectorSize = (m_numSubpatterns + 1) * 2;
+    Vector<int, 32> interpreterOvector;
+    interpreterOvector.resize(offsetVectorSize);
+    int* interpreterOffsetVector = interpreterOvector.data();
+    int interpreterResult = 0;
+    int differences = 0;
+
+    // Initialize interpreterOffsetVector with the return value (index 0) and the 
+    // first subpattern start indicies (even index values) set to -1.
+    // No need to init the subpattern end indicies.
+    for (unsigned j = 0, i = 0; i < m_numSubpatterns + 1; j += 2, i++)
+        interpreterOffsetVector[j] = -1;
+
+    interpreterResult = Yarr::interpret(m_regExpBytecode.get(), s, startOffset, interpreterOffsetVector);
+
+    if (jitResult != interpreterResult)
+        differences++;
+
+    for (unsigned j = 2, i = 0; i < m_numSubpatterns; j +=2, i++)
+        if ((offsetVector[j] != interpreterOffsetVector[j])
+            || ((offsetVector[j] >= 0) && (offsetVector[j+1] != interpreterOffsetVector[j+1])))
+            differences++;
+
+    if (differences) {
+        dataLogF("RegExp Discrepency for /%s/\n    string input ", pattern().utf8().data());
+        unsigned segmentLen = s.length() - static_cast<unsigned>(startOffset);
+
+        dataLogF((segmentLen < 150) ? "\"%s\"\n" : "\"%148s...\"\n", s.utf8().data() + startOffset);
+
+        if (jitResult != interpreterResult) {
+            dataLogF("    JIT result = %d, blah interpreted result = %d\n", jitResult, interpreterResult);
+            differences--;

         } else {
         } else {

-            offsetVectorSize = (m_numSubpatterns + 1) * 3;
-            ovector->resize(offsetVectorSize);
-            offsetVector = ovector->data();
+            dataLogF("    Correct result = %d\n", jitResult);

         }
 
         }
 

-        int numMatches = jsRegExpExecute(m_regExp, reinterpret_cast<const UChar*>(s.data()), s.size(), startOffset, offsetVector, offsetVectorSize);
-    
-        if (numMatches < 0) {
-#ifndef NDEBUG
-            if (numMatches != JSRegExpErrorNoMatch)
-                fprintf(stderr, "jsRegExpExecute failed with result %d\n", numMatches);
-#endif
-            if (ovector)
-                ovector->clear();
-            return -1;
+        if (differences) {
+            for (unsigned j = 2, i = 0; i < m_numSubpatterns; j +=2, i++) {
+                if (offsetVector[j] != interpreterOffsetVector[j])
+                    dataLogF("    JIT offset[%d] = %d, interpreted offset[%d] = %d\n", j, offsetVector[j], j, interpreterOffsetVector[j]);
+                if ((offsetVector[j] >= 0) && (offsetVector[j+1] != interpreterOffsetVector[j+1]))
+                    dataLogF("    JIT offset[%d] = %d, interpreted offset[%d] = %d\n", j+1, offsetVector[j+1], j+1, interpreterOffsetVector[j+1]);
+            }

         }
         }

-
-        return offsetVector[0];

     }
     }

-
-    return -1;

 }
 }

+#endif
+
+#if ENABLE(REGEXP_TRACING)
+    void RegExp::printTraceData()
+    {
+        char formattedPattern[41];
+        char rawPattern[41];
+
+        strncpy(rawPattern, pattern().utf8().data(), 40);
+        rawPattern[40]= '\0';
+
+        int pattLen = strlen(rawPattern);

 
 

+        snprintf(formattedPattern, 41, (pattLen <= 38) ? "/%.38s/" : "/%.36s...", rawPattern);
+
+#if ENABLE(YARR_JIT)
+        Yarr::YarrCodeBlock& codeBlock = m_regExpJITCode;
+
+        const size_t jitAddrSize = 20;
+        char jit8BitMatchOnlyAddr[jitAddrSize];
+        char jit16BitMatchOnlyAddr[jitAddrSize];
+        char jit8BitMatchAddr[jitAddrSize];
+        char jit16BitMatchAddr[jitAddrSize];
+        if (m_state == ByteCode) {
+            snprintf(jit8BitMatchOnlyAddr, jitAddrSize, "fallback    ");
+            snprintf(jit16BitMatchOnlyAddr, jitAddrSize, "----      ");
+            snprintf(jit8BitMatchAddr, jitAddrSize, "fallback    ");
+            snprintf(jit16BitMatchAddr, jitAddrSize, "----      ");
+        } else {
+            snprintf(jit8BitMatchOnlyAddr, jitAddrSize, "0x%014lx", reinterpret_cast<unsigned long int>(codeBlock.get8BitMatchOnlyAddr()));
+            snprintf(jit16BitMatchOnlyAddr, jitAddrSize, "0x%014lx", reinterpret_cast<unsigned long int>(codeBlock.get16BitMatchOnlyAddr()));
+            snprintf(jit8BitMatchAddr, jitAddrSize, "0x%014lx", reinterpret_cast<unsigned long int>(codeBlock.get8BitMatchAddr()));
+            snprintf(jit16BitMatchAddr, jitAddrSize, "0x%014lx", reinterpret_cast<unsigned long int>(codeBlock.get16BitMatchAddr()));
+        }
+#else
+        const char* jit8BitMatchOnlyAddr = "JIT Off";
+        const char* jit16BitMatchOnlyAddr = "";
+        const char* jit8BitMatchAddr = "JIT Off";
+        const char* jit16BitMatchAddr = "";
+#endif
+        unsigned averageMatchOnlyStringLen = (unsigned)(m_rtMatchOnlyTotalSubjectStringLen / m_rtMatchOnlyCallCount);
+        unsigned averageMatchStringLen = (unsigned)(m_rtMatchTotalSubjectStringLen / m_rtMatchCallCount);
+
+        printf("%-40.40s %16.16s %16.16s %10d %10d %10u\n", formattedPattern, jit8BitMatchOnlyAddr, jit16BitMatchOnlyAddr, m_rtMatchOnlyCallCount, m_rtMatchOnlyFoundCount, averageMatchOnlyStringLen);
+        printf("                                         %16.16s %16.16s %10d %10d %10u\n", jit8BitMatchAddr, jit16BitMatchAddr, m_rtMatchCallCount, m_rtMatchFoundCount, averageMatchStringLen);
+    }

 #endif
 
 } // namespace JSC
 #endif
 
 } // namespace JSC