Merge webkit.org at r58033 : Initial merge by git

Change-Id: If006c38561af287c50cd578d251629b51e4d8cd1

8 files changed, 78 insertions, 164 deletions

diff --git a/JavaScriptCore/yarr/RegexCompiler.cpp b/JavaScriptCore/yarr/RegexCompiler.cpp
index 9cd3d12..9fbe213 100644
--- a/JavaScriptCore/yarr/RegexCompiler.cpp
+++ b/JavaScriptCore/yarr/RegexCompiler.cpp
@@ -36,6 +36,8 @@ using namespace WTF;
 
 namespace JSC { namespace Yarr {
 
+#include "RegExpJitTables.h"
+
 class CharacterClassConstructor {
 public:
     CharacterClassConstructor(bool isCaseInsensitive = false)
@@ -141,7 +143,7 @@ public:
 
     CharacterClass* charClass()
     {
-        CharacterClass* characterClass = new CharacterClass();
+        CharacterClass* characterClass = new CharacterClass(0);
 
         characterClass->m_matches.append(m_matches);
         characterClass->m_ranges.append(m_ranges);
@@ -233,105 +235,6 @@ private:
     Vector<CharacterRange> m_rangesUnicode;
 };
 
-
-CharacterClass* newlineCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_matches.append('\n');
-    characterClass->m_matches.append('\r');
-    characterClass->m_matchesUnicode.append(0x2028);
-    characterClass->m_matchesUnicode.append(0x2029);
-    
-    return characterClass;
-}
-
-CharacterClass* digitsCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_ranges.append(CharacterRange('0', '9'));
-    
-    return characterClass;
-}
-
-CharacterClass* spacesCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_matches.append(' ');
-    characterClass->m_ranges.append(CharacterRange('\t', '\r'));
-    characterClass->m_matchesUnicode.append(0x00a0);
-    characterClass->m_matchesUnicode.append(0x1680);
-    characterClass->m_matchesUnicode.append(0x180e);
-    characterClass->m_matchesUnicode.append(0x2028);
-    characterClass->m_matchesUnicode.append(0x2029);
-    characterClass->m_matchesUnicode.append(0x202f);
-    characterClass->m_matchesUnicode.append(0x205f);
-    characterClass->m_matchesUnicode.append(0x3000);
-    characterClass->m_rangesUnicode.append(CharacterRange(0x2000, 0x200a));
-    
-    return characterClass;
-}
-
-CharacterClass* wordcharCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_matches.append('_');
-    characterClass->m_ranges.append(CharacterRange('0', '9'));
-    characterClass->m_ranges.append(CharacterRange('A', 'Z'));
-    characterClass->m_ranges.append(CharacterRange('a', 'z'));
-    
-    return characterClass;
-}
-
-CharacterClass* nondigitsCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_ranges.append(CharacterRange(0, '0' - 1));
-    characterClass->m_ranges.append(CharacterRange('9' + 1, 0x7f));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x80, 0xffff));
-    
-    return characterClass;
-}
-
-CharacterClass* nonspacesCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_ranges.append(CharacterRange(0, '\t' - 1));
-    characterClass->m_ranges.append(CharacterRange('\r' + 1, ' ' - 1));
-    characterClass->m_ranges.append(CharacterRange(' ' + 1, 0x7f));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x0080, 0x009f));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x00a1, 0x167f));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x1681, 0x180d));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x180f, 0x1fff));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x200b, 0x2027));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x202a, 0x202e));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x2030, 0x205e));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x2060, 0x2fff));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x3001, 0xffff));
-    
-    return characterClass;
-}
-
-CharacterClass* nonwordcharCreate()
-{
-    CharacterClass* characterClass = new CharacterClass();
-
-    characterClass->m_matches.append('`');
-    characterClass->m_ranges.append(CharacterRange(0, '0' - 1));
-    characterClass->m_ranges.append(CharacterRange('9' + 1, 'A' - 1));
-    characterClass->m_ranges.append(CharacterRange('Z' + 1, '_' - 1));
-    characterClass->m_ranges.append(CharacterRange('z' + 1, 0x7f));
-    characterClass->m_rangesUnicode.append(CharacterRange(0x80, 0xffff));
-
-    return characterClass;
-}
-
-
 class RegexPatternConstructor {
 public:
     RegexPatternConstructor(RegexPattern& pattern)
@@ -469,6 +372,7 @@ public:
     void atomBackReference(unsigned subpatternId)
     {
         ASSERT(subpatternId);
+        m_pattern.m_shouldFallBack = true;
         m_pattern.m_maxBackReference = std::max(m_pattern.m_maxBackReference, subpatternId);
 
         if (subpatternId > m_pattern.m_numSubpatterns) {
@@ -544,6 +448,9 @@ public:
             return;
         }
 
+        if (max > 1 && term.type == PatternTerm::TypeParenthesesSubpattern)
+            m_pattern.m_shouldFallBack = true;
+
         if (min == 0)
             term.quantify(max, greedy   ? QuantifierGreedy : QuantifierNonGreedy);
         else if (min == max)
diff --git a/JavaScriptCore/yarr/RegexCompiler.h b/JavaScriptCore/yarr/RegexCompiler.h
index 3ed2be9..9d2443a 100644
--- a/JavaScriptCore/yarr/RegexCompiler.h
+++ b/JavaScriptCore/yarr/RegexCompiler.h
@@ -26,13 +26,11 @@
 #ifndef RegexCompiler_h
 #define RegexCompiler_h
 
-#include <wtf/Platform.h>
-
 #if ENABLE(YARR)
 
-#include <wtf/unicode/Unicode.h>
 #include "RegexParser.h"
 #include "RegexPattern.h"
+#include <wtf/unicode/Unicode.h>
 
 namespace JSC { namespace Yarr {
 
diff --git a/JavaScriptCore/yarr/RegexInterpreter.cpp b/JavaScriptCore/yarr/RegexInterpreter.cpp
index d088086..c2cb1c2 100644
--- a/JavaScriptCore/yarr/RegexInterpreter.cpp
+++ b/JavaScriptCore/yarr/RegexInterpreter.cpp
@@ -280,20 +280,6 @@ public:
         return false;
     }
 
-    bool tryConsumeCharacter(int testChar)
-    {
-        if (input.atEnd())
-            return false;
-
-        int ch = input.read();
-
-        if (pattern->m_ignoreCase ? ((Unicode::toLower(testChar) == ch) || (Unicode::toUpper(testChar) == ch)) : (testChar == ch)) {
-            input.next();
-            return true;
-        }
-        return false;
-    }
-
     bool checkCharacter(int testChar, int inputPosition)
     {
         return testChar == input.readChecked(inputPosition);
@@ -305,23 +291,6 @@ public:
         return (loChar == ch) || (hiChar == ch);
     }
 
-    bool tryConsumeCharacterClass(CharacterClass* characterClass, bool invert)
-    {
-        if (input.atEnd())
-            return false;
-
-        bool match = testCharacterClass(characterClass, input.read());
-
-        if (invert)
-            match = !match;
-
-        if (match) {
-            input.next();
-            return true;
-        }
-        return false;
-    }
-
     bool checkCharacterClass(CharacterClass* characterClass, bool invert, int inputPosition)
     {
         bool match = testCharacterClass(characterClass, input.readChecked(inputPosition));
diff --git a/JavaScriptCore/yarr/RegexInterpreter.h b/JavaScriptCore/yarr/RegexInterpreter.h
index 48c9a5e..e3c3122 100644
--- a/JavaScriptCore/yarr/RegexInterpreter.h
+++ b/JavaScriptCore/yarr/RegexInterpreter.h
@@ -26,13 +26,11 @@
 #ifndef RegexInterpreter_h
 #define RegexInterpreter_h
 
-#include <wtf/Platform.h>
-
 #if ENABLE(YARR)
 
-#include <wtf/unicode/Unicode.h>
 #include "RegexParser.h"
 #include "RegexPattern.h"
+#include <wtf/unicode/Unicode.h>
 
 namespace JSC { namespace Yarr {
 
diff --git a/JavaScriptCore/yarr/RegexJIT.cpp b/JavaScriptCore/yarr/RegexJIT.cpp
index fcb8d86..340b53d 100644
--- a/JavaScriptCore/yarr/RegexJIT.cpp
+++ b/JavaScriptCore/yarr/RegexJIT.cpp
@@ -40,7 +40,6 @@ using namespace WTF;
 
 namespace JSC { namespace Yarr {
 
-
 class RegexGenerator : private MacroAssembler {
     friend void jitCompileRegex(JSGlobalData* globalData, RegexCodeBlock& jitObject, const UString& pattern, unsigned& numSubpatterns, const char*& error, bool ignoreCase, bool multiline);
 
@@ -54,6 +53,16 @@ class RegexGenerator : private MacroAssembler {
     static const RegisterID regT1 = ARMRegisters::r6;
 
     static const RegisterID returnRegister = ARMRegisters::r0;
+#elif CPU(MIPS)
+    static const RegisterID input = MIPSRegisters::a0;
+    static const RegisterID index = MIPSRegisters::a1;
+    static const RegisterID length = MIPSRegisters::a2;
+    static const RegisterID output = MIPSRegisters::a3;
+
+    static const RegisterID regT0 = MIPSRegisters::t4;
+    static const RegisterID regT1 = MIPSRegisters::t5;
+
+    static const RegisterID returnRegister = MIPSRegisters::v0;
 #elif CPU(X86)
     static const RegisterID input = X86Registers::eax;
     static const RegisterID index = X86Registers::edx;
@@ -145,6 +154,11 @@ class RegexGenerator : private MacroAssembler {
 
     void matchCharacterClass(RegisterID character, JumpList& matchDest, const CharacterClass* charClass)
     {
+        if (charClass->m_table) {
+            ExtendedAddress tableEntry(character, reinterpret_cast<intptr_t>(charClass->m_table->m_table));
+            matchDest.append(branchTest8(charClass->m_table->m_inverted ? Zero : NonZero, tableEntry));   
+            return;
+        }
         Jump unicodeFail;
         if (charClass->m_matchesUnicode.size() || charClass->m_rangesUnicode.size()) {
             Jump isAscii = branch32(LessThanOrEqual, character, Imm32(0x7f));
@@ -599,9 +613,14 @@ class RegexGenerator : private MacroAssembler {
             ASSERT(!m_pattern.m_ignoreCase || (Unicode::toLower(ch) == Unicode::toUpper(ch)));
             failures.append(jumpIfCharNotEquals(ch, state.inputOffset()));
         }
+
         add32(Imm32(1), countRegister);
         add32(Imm32(1), index);
-        branch32(NotEqual, countRegister, Imm32(term.quantityCount)).linkTo(loop, this);
+        if (term.quantityCount != 0xffffffff)
+            branch32(NotEqual, countRegister, Imm32(term.quantityCount)).linkTo(loop, this);
+        else
+            jump(loop);
+
         failures.append(jump());
 
         Label backtrackBegin(this);
@@ -636,7 +655,8 @@ class RegexGenerator : private MacroAssembler {
         loadFromFrame(term.frameLocation, countRegister);
 
         atEndOfInput().linkTo(hardFail, this);
-        branch32(Equal, countRegister, Imm32(term.quantityCount), hardFail);
+        if (term.quantityCount != 0xffffffff)
+            branch32(Equal, countRegister, Imm32(term.quantityCount), hardFail);
         if (m_pattern.m_ignoreCase && isASCIIAlpha(ch)) {
             readCharacter(state.inputOffset(), character);
             or32(Imm32(32), character);
@@ -722,7 +742,11 @@ class RegexGenerator : private MacroAssembler {
 
         add32(Imm32(1), countRegister);
         add32(Imm32(1), index);
-        branch32(NotEqual, countRegister, Imm32(term.quantityCount)).linkTo(loop, this);
+        if (term.quantityCount != 0xffffffff)
+            branch32(NotEqual, countRegister, Imm32(term.quantityCount)).linkTo(loop, this);
+        else
+            jump(loop);
+
         failures.append(jump());
 
         Label backtrackBegin(this);
@@ -1078,17 +1102,15 @@ class RegexGenerator : private MacroAssembler {
             break;
 
         case PatternTerm::TypeBackReference:
-            m_generationFailed = true;
+            ASSERT_NOT_REACHED();
             break;
 
         case PatternTerm::TypeForwardReference:
             break;
 
         case PatternTerm::TypeParenthesesSubpattern:
-            if ((term.quantityCount == 1) && !term.parentheses.isCopy)
-                generateParenthesesSingle(state);
-            else
-                m_generationFailed = true;
+            ASSERT((term.quantityCount == 1) && !term.parentheses.isCopy); // must fallback to pcre before this point
+            generateParenthesesSingle(state);
             break;
 
         case PatternTerm::TypeParentheticalAssertion:
@@ -1313,6 +1335,8 @@ class RegexGenerator : private MacroAssembler {
         push(ARMRegisters::r5);
         push(ARMRegisters::r6);
         move(ARMRegisters::r3, output);
+#elif CPU(MIPS)
+        // Do nothing.
 #endif
     }
 
@@ -1330,6 +1354,8 @@ class RegexGenerator : private MacroAssembler {
         pop(ARMRegisters::r6);
         pop(ARMRegisters::r5);
         pop(ARMRegisters::r4);
+#elif CPU(MIPS)
+        // Do nothing
 #endif
         ret();
     }
@@ -1337,7 +1363,6 @@ class RegexGenerator : private MacroAssembler {
 public:
     RegexGenerator(RegexPattern& pattern)
         : m_pattern(pattern)
-        , m_generationFailed(false)
     {
     }
 
@@ -1367,15 +1392,9 @@ public:
         jitObject.set(patchBuffer.finalizeCode());
     }
 
-    bool generationFailed()
-    {
-        return m_generationFailed;
-    }
-
 private:
     RegexPattern& m_pattern;
     Vector<AlternativeBacktrackRecord> m_backtrackRecords;
-    bool m_generationFailed;
 };
 
 void jitCompileRegex(JSGlobalData* globalData, RegexCodeBlock& jitObject, const UString& patternString, unsigned& numSubpatterns, const char*& error, bool ignoreCase, bool multiline)
@@ -1387,13 +1406,13 @@ void jitCompileRegex(JSGlobalData* globalData, RegexCodeBlock& jitObject, const
 
     numSubpatterns = pattern.m_numSubpatterns;
 
-    RegexGenerator generator(pattern);
-    generator.compile(globalData, jitObject);
-
-    if (generator.generationFailed()) {
+    if (pattern.m_shouldFallBack) {
         JSRegExpIgnoreCaseOption ignoreCaseOption = ignoreCase ? JSRegExpIgnoreCase : JSRegExpDoNotIgnoreCase;
         JSRegExpMultilineOption multilineOption = multiline ? JSRegExpMultiline : JSRegExpSingleLine;
         jitObject.setFallback(jsRegExpCompile(reinterpret_cast<const UChar*>(patternString.data()), patternString.size(), ignoreCaseOption, multilineOption, &numSubpatterns, &error));
+    } else {
+        RegexGenerator generator(pattern);
+        generator.compile(globalData, jitObject);
     }
 }
 
diff --git a/JavaScriptCore/yarr/RegexJIT.h b/JavaScriptCore/yarr/RegexJIT.h
index 935b9a3..7f9c16e 100644
--- a/JavaScriptCore/yarr/RegexJIT.h
+++ b/JavaScriptCore/yarr/RegexJIT.h
@@ -26,8 +26,6 @@
 #ifndef RegexJIT_h
 #define RegexJIT_h
 
-#include <wtf/Platform.h>
-
 #if ENABLE(YARR_JIT)
 
 #include "MacroAssembler.h"
@@ -68,7 +66,7 @@ public:
     JSRegExp* getFallback() { return m_fallback; }
     void setFallback(JSRegExp* fallback) { m_fallback = fallback; }
 
-    bool operator!() { return !m_ref.m_code.executableAddress(); }
+    bool operator!() { return (!m_ref.m_code.executableAddress() && !m_fallback); }
     void set(MacroAssembler::CodeRef ref) { m_ref = ref; }
 
     int execute(const UChar* input, unsigned start, unsigned length, int* output)
diff --git a/JavaScriptCore/yarr/RegexParser.h b/JavaScriptCore/yarr/RegexParser.h
index 64e8463..c946c2e 100644
--- a/JavaScriptCore/yarr/RegexParser.h
+++ b/JavaScriptCore/yarr/RegexParser.h
@@ -26,14 +26,12 @@
 #ifndef RegexParser_h
 #define RegexParser_h
 
-#include <wtf/Platform.h>
-
 #if ENABLE(YARR)
 
 #include <UString.h>
+#include <limits.h>
 #include <wtf/ASCIICType.h>
 #include <wtf/unicode/Unicode.h>
-#include <limits.h>
 
 namespace JSC { namespace Yarr {
 
diff --git a/JavaScriptCore/yarr/RegexPattern.h b/JavaScriptCore/yarr/RegexPattern.h
index dd7512d..3271cc1 100644
--- a/JavaScriptCore/yarr/RegexPattern.h
+++ b/JavaScriptCore/yarr/RegexPattern.h
@@ -26,7 +26,6 @@
 #ifndef RegexPattern_h
 #define RegexPattern_h
 
-#include <wtf/Platform.h>
 
 #if ENABLE(YARR)
 
@@ -57,11 +56,35 @@ struct CharacterRange {
     }
 };
 
+struct CharacterClassTable : RefCounted<CharacterClassTable> {
+    const char* m_table;
+    bool m_inverted;
+    static PassRefPtr<CharacterClassTable> create(const char* table, bool inverted)
+    {
+        return adoptRef(new CharacterClassTable(table, inverted));
+    }
+
+private:
+    CharacterClassTable(const char* table, bool inverted)
+        : m_table(table)
+        , m_inverted(inverted)
+    {
+    }
+};
+
 struct CharacterClass : FastAllocBase {
+    // All CharacterClass instances have to have the full set of matches and ranges,
+    // they may have an optional table for faster lookups (which must match the
+    // specified matches and ranges)
+    CharacterClass(PassRefPtr<CharacterClassTable> table)
+        : m_table(table)
+    {
+    }
     Vector<UChar> m_matches;
     Vector<CharacterRange> m_ranges;
     Vector<UChar> m_matchesUnicode;
     Vector<CharacterRange> m_rangesUnicode;
+    RefPtr<CharacterClassTable> m_table;
 };
 
 enum QuantifierType {
@@ -248,6 +271,7 @@ struct RegexPattern {
         , m_multiline(multiline)
         , m_numSubpatterns(0)
         , m_maxBackReference(0)
+        , m_shouldFallBack(false)
         , newlineCached(0)
         , digitsCached(0)
         , spacesCached(0)
@@ -269,6 +293,8 @@ struct RegexPattern {
         m_numSubpatterns = 0;
         m_maxBackReference = 0;
 
+        m_shouldFallBack = false;
+
         newlineCached = 0;
         digitsCached = 0;
         spacesCached = 0;
@@ -335,6 +361,7 @@ struct RegexPattern {
     bool m_multiline;
     unsigned m_numSubpatterns;
     unsigned m_maxBackReference;
+    bool m_shouldFallBack;
     PatternDisjunction* m_body;
     Vector<PatternDisjunction*, 4> m_disjunctions;
     Vector<CharacterClass*> m_userCharacterClasses;