Fix input and output to handle UTF16 surrogate pairs.

src/unicode-inl.h

Index: src/unicode-inl.h
===================================================================
--- src/unicode-inl.h	(revision 10944)
+++ src/unicode-inl.h	(working copy)
@@ -78,7 +78,7 @@
 }
 
 
-unsigned Utf8::Encode(char* str, uchar c) {
+unsigned Utf8::Encode(char* str, uchar c, int previous) {
   static const int kMask = ~(1 << 6);
   if (c <= kMaxOneByteChar) {
     str[0] = c;
@@ -88,6 +88,13 @@
     str[1] = 0x80 | (c & kMask);
     return 2;
   } else if (c <= kMaxThreeByteChar) {
+    if (Utf16::IsTrailSurrogate(c) &&
+        previous != kNoPreviousCharacter &&

[rossberg, 2012/03/07 13:32:47]

Isn't that implied by Utf16::IsLeadSurrogate(previous)?

[Erik Corry, 2012/03/11 19:29:22]

No, but it should be.  Fixed.

+        Utf16::IsLeadSurrogate(previous)) {
+      return Encode(str - 3,
+                    Utf16::CombineSurrogatePair(previous, c),
+                    Utf8::kNoPreviousCharacter) - 3;

[rossberg, 2012/03/07 13:32:47]

kSizeOfUnmatchedSurrogate instead of 3 perhaps (here and above).

[Erik Corry, 2012/03/11 19:29:22]

Done.

+    }
     str[0] = 0xE0 | (c >> 12);
     str[1] = 0x80 | ((c >> 6) & kMask);
     str[2] = 0x80 | (c & kMask);
@@ -113,12 +120,17 @@
   return CalculateValue(bytes, length, cursor);
 }
 
-unsigned Utf8::Length(uchar c) {
+unsigned Utf8::Length(uchar c, int previous) {
   if (c <= kMaxOneByteChar) {
     return 1;
   } else if (c <= kMaxTwoByteChar) {
     return 2;
   } else if (c <= kMaxThreeByteChar) {
+    if (Utf16::IsTrailSurrogate(c) &&
+        previous != kNoPreviousCharacter &&

[rossberg, 2012/03/07 13:32:47]

See above.

[Erik Corry, 2012/03/11 19:29:22]

Done.

+        Utf16::IsLeadSurrogate(previous)) {
+      return 1;

[rossberg, 2012/03/07 13:32:47]

This is 4 - 3 already counted, I suppose.

[Erik Corry, 2012/03/11 19:29:22]

Fixed

+    }
     return 3;
   } else {
     return 4;