Introduce ENABLE_LATIN_1 compile flag

src/jsregexp.cc

Index: src/jsregexp.cc
diff --git a/src/jsregexp.cc b/src/jsregexp.cc
index 813208c9590e651749321a818351488629da4f1f..384af62071af5dab457270d6f88c6854fbd9c8a3 100644
--- a/src/jsregexp.cc
+++ b/src/jsregexp.cc
@@ -1681,7 +1681,7 @@ static int GetCaseIndependentLetters(Isolate* isolate,
     letters[0] = character;
     length = 1;
   }
-  if (!ascii_subject || character <= String::kMaxAsciiCharCode) {
+  if (!ascii_subject || character <= String::kMaxOneByteCharCode) {
     return length;
   }
   // The standard requires that non-ASCII characters cannot have ASCII
@@ -1732,7 +1732,7 @@ static inline bool EmitAtomNonLetter(Isolate* isolate,
   bool checked = false;
   // We handle the length > 1 case in a later pass.
   if (length == 1) {
-    if (ascii && c > String::kMaxAsciiCharCodeU) {
+    if (ascii && c > String::kMaxOneByteCharCodeU) {
       // Can't match - see above.
       return false;  // Bounds not checked.
     }
@@ -1753,7 +1753,7 @@ static bool ShortCutEmitCharacterPair(RegExpMacroAssembler* macro_assembler,
                                       Label* on_failure) {
   uc16 char_mask;
   if (ascii) {
-    char_mask = String::kMaxAsciiCharCode;
+    char_mask = String::kMaxOneByteCharCode;
   } else {
     char_mask = String::kMaxUtf16CodeUnit;
   }
@@ -2007,7 +2007,7 @@ static void SplitSearchSpace(ZoneList<int>* ranges,
   // range with a single not-taken branch, speeding up this important
   // character range (even non-ASCII charset-based text has spaces and
   // punctuation).
-  if (*border - 1 > String::kMaxAsciiCharCode &&  // ASCII case.
+  if (*border - 1 > String::kMaxOneByteCharCode &&  // ASCII case.

[Yang, 2013/01/07 16:15:07]

Change comment.

       end_index - start_index > (*new_start_index - start_index) * 2 &&
       last - first > kSize * 2 &&
       binary_chop_index > *new_start_index &&
@@ -2211,7 +2211,7 @@ static void EmitCharClass(RegExpMacroAssembler* macro_assembler,
 
   int max_char;
   if (ascii) {
-    max_char = String::kMaxAsciiCharCode;
+    max_char = String::kMaxOneByteCharCode;
   } else {
     max_char = String::kMaxUtf16CodeUnit;
   }
@@ -2513,7 +2513,7 @@ bool QuickCheckDetails::Rationalize(bool asc) {
   bool found_useful_op = false;
   uint32_t char_mask;
   if (asc) {
-    char_mask = String::kMaxAsciiCharCode;
+    char_mask = String::kMaxOneByteCharCode;
   } else {
     char_mask = String::kMaxUtf16CodeUnit;
   }
@@ -2522,7 +2522,7 @@ bool QuickCheckDetails::Rationalize(bool asc) {
   int char_shift = 0;
   for (int i = 0; i < characters_; i++) {
     Position* pos = &positions_[i];
-    if ((pos->mask & String::kMaxAsciiCharCode) != 0) {
+    if ((pos->mask & String::kMaxOneByteCharCode) != 0) {
       found_useful_op = true;
     }
     mask_ |= (pos->mask & char_mask) << char_shift;
@@ -2565,7 +2565,7 @@ bool RegExpNode::EmitQuickCheck(RegExpCompiler* compiler,
     // load so the value is already masked down.
     uint32_t char_mask;
     if (compiler->ascii()) {

[Yang, 2013/01/07 16:15:07]

Maybe onebyte() instead of ascii()?

-      char_mask = String::kMaxAsciiCharCode;
+      char_mask = String::kMaxOneByteCharCode;
     } else {
       char_mask = String::kMaxUtf16CodeUnit;
     }
@@ -2575,7 +2575,7 @@ bool RegExpNode::EmitQuickCheck(RegExpCompiler* compiler,
     // For 2-character preloads in ASCII mode or 1-character preloads in
     // TWO_BYTE mode we also use a 16 bit load with zero extend.
     if (details->characters() == 2 && compiler->ascii()) {
-      if ((mask & 0x7f7f) == 0x7f7f) need_mask = false;
+      if ((mask & 0xffff) == 0xffff) need_mask = false;
     } else if (details->characters() == 1 && !compiler->ascii()) {
       if ((mask & 0xffff) == 0xffff) need_mask = false;
     } else {
@@ -2617,7 +2617,7 @@ void TextNode::GetQuickCheckDetails(QuickCheckDetails* details,
   int characters = details->characters();
   int char_mask;
   if (compiler->ascii()) {
-    char_mask = String::kMaxAsciiCharCode;
+    char_mask = String::kMaxOneByteCharCode;
   } else {
     char_mask = String::kMaxUtf16CodeUnit;
   }
@@ -2865,7 +2865,7 @@ RegExpNode* TextNode::FilterASCII(int depth) {
         // We don't need special handling for case independence
         // because of the rule that case independence cannot make
         // a non-ASCII character match an ASCII character.
-        if (quarks[j] > String::kMaxAsciiCharCode) {
+        if (quarks[j] > String::kMaxOneByteCharCode) {

[Yang, 2013/01/07 16:15:07]

Does the comment still hold true for the Latin1 char set? Apparently ÿ (\u00ff)
is part of Latin1, but its uppercase  Ÿ is is \u0178...

Maybe we want to keep FilterASCII the way it is, and use it in combination with
kAsciiDataHintTag.

           return set_replacement(NULL);
         }
       }
@@ -2881,12 +2881,12 @@ RegExpNode* TextNode::FilterASCII(int depth) {
       if (cc->is_negated()) {
         if (range_count != 0 &&
             ranges->at(0).from() == 0 &&
-            ranges->at(0).to() >= String::kMaxAsciiCharCode) {
+            ranges->at(0).to() >= String::kMaxOneByteCharCode) {
           return set_replacement(NULL);
         }
       } else {
         if (range_count == 0 ||
-            ranges->at(0).from() > String::kMaxAsciiCharCode) {
+            ranges->at(0).from() > String::kMaxOneByteCharCode) {
           return set_replacement(NULL);
         }
       }
@@ -3299,7 +3299,7 @@ void TextNode::TextEmitPass(RegExpCompiler* compiler,
         switch (pass) {
           case NON_ASCII_MATCH:
             ASSERT(ascii);
-            if (quarks[j] > String::kMaxAsciiCharCode) {
+            if (quarks[j] > String::kMaxOneByteCharCode) {
               assembler->GoTo(backtrack);
               return;
             }
@@ -3498,7 +3498,7 @@ RegExpNode* TextNode::GetSuccessorOfOmnivorousTextNode(
   if (ranges->length() != 1) return NULL;
   uint32_t max_char;
   if (compiler->ascii()) {
-    max_char = String::kMaxAsciiCharCode;
+    max_char = String::kMaxOneByteCharCode;
   } else {
     max_char = String::kMaxUtf16CodeUnit;
   }
@@ -3698,7 +3698,7 @@ BoyerMooreLookahead::BoyerMooreLookahead(
     : length_(length),
       compiler_(compiler) {
   if (compiler->ascii()) {
-    max_char_ = String::kMaxAsciiCharCode;
+    max_char_ = String::kMaxOneByteCharCode;
   } else {
     max_char_ = String::kMaxUtf16CodeUnit;
   }
@@ -5337,8 +5337,8 @@ void CharacterRange::AddCaseEquivalents(ZoneList<CharacterRange>* ranges,
   uc16 bottom = from();
   uc16 top = to();
   if (is_ascii) {
-    if (bottom > String::kMaxAsciiCharCode) return;
-    if (top > String::kMaxAsciiCharCode) top = String::kMaxAsciiCharCode;
+    if (bottom > String::kMaxOneByteCharCode) return;
+    if (top > String::kMaxOneByteCharCode) top = String::kMaxOneByteCharCode;
   }
   unibrow::uchar chars[unibrow::Ecma262UnCanonicalize::kMaxWidth];
   if (top == bottom) {
@@ -5885,7 +5885,7 @@ void TextNode::FillInBMInfo(int initial_offset,
           int length = GetCaseIndependentLetters(
               ISOLATE,
               character,
-              bm->max_char() == String::kMaxAsciiCharCode,
+              bm->max_char() == String::kMaxOneByteCharCode,
               chars);
           for (int j = 0; j < length; j++) {
             bm->Set(offset, chars[j]);