[regexp] back refs must not start/end in the middle of a surrogate pair

src/regexp/regexp-macro-assembler.cc

Index: src/regexp/regexp-macro-assembler.cc
diff --git a/src/regexp/regexp-macro-assembler.cc b/src/regexp/regexp-macro-assembler.cc
index caf8b51fe548d6a9e68365654110473ddbdea9b5..89e162a802ef08a4dbb452e1be51c8aedfbd5f6f 100644
--- a/src/regexp/regexp-macro-assembler.cc
+++ b/src/regexp/regexp-macro-assembler.cc
@@ -23,6 +23,21 @@ RegExpMacroAssembler::~RegExpMacroAssembler() {
 }
 
 
+void RegExpMacroAssembler::CheckNotInSurrogatePair(int cp_offset,
+                                                   Label* on_failure) {
+  Label ok;
+  // Check that current character is not a trail surrogate.
+  LoadCurrentCharacter(cp_offset, &ok, true, 1);

[erikcorry, 2016/01/26 16:25:06]

True, 1 are default args and not needed.

[Yang, 2016/01/27 09:06:06]

Done.

+  CheckCharacterNotInRange(kTrailSurrogateStart, kTrailSurrogateEnd, &ok);
+  // Check that previous character is not a lead surrogate.
+  LoadCurrentCharacter(cp_offset - 1, &ok, true, 1);
+  CheckCharacterNotInRange(kLeadSurrogateStart, kLeadSurrogateEnd, &ok);

[erikcorry, 2016/01/26 16:25:06]

CheckCharacterInRange(on_failure) ?

[Yang, 2016/01/27 09:06:06]

Done.

+  // Looks like we are right in between a lead and a trail surrogate.
+  GoTo(on_failure);
+  Bind(&ok);
+}
+
+
 #ifndef V8_INTERPRETED_REGEXP  // Avoid unused code, e.g., on ARM.
 
 NativeRegExpMacroAssembler::NativeRegExpMacroAssembler(Isolate* isolate,