libjson 7.6.1 utf8 bug修复的patch

libjson.patch

Common subdirectories: libjson_7.6.1/_internal/Source/JSONDefs and libjson_7.6.1.mine/_internal/Source/JSONDefs
diff -uN libjson_7.6.1/_internal/Source/JSONWorker.cpp libjson_7.6.1.mine/_internal/Source/JSONWorker.cpp
--- libjson_7.6.1/_internal/Source/JSONWorker.cpp	2012-05-30 17:14:36.000000000 +0800
+++ libjson_7.6.1.mine/_internal/Source/JSONWorker.cpp	2013-04-19 07:05:06.000000000 +0800
@@ -275,12 +275,12 @@
 
     void JSONWorker::UTF(const json_char * & pos, json_string & result, const json_char * const end) json_nothrow {
 		JSON_ASSERT_SAFE(((long)end - (long)pos) > 4, JSON_TEXT("UTF will go out of bounds"), return;);
-	   json_uchar first = UTF8(pos, end);
+	   json_uchar first = UTF8(pos, result, end);
 	   if (json_unlikely((first > 0xD800) && (first < 0xDBFF) &&
 		  (*(pos + 1) == '\\') && (*(pos + 2) == 'u'))){
 			 const json_char * original_pos = pos;  //if the 2nd character is not correct I need to roll back the iterator
 			 pos += 2;
-			 json_uchar second = UTF8(pos, end);
+			 json_uchar second = UTF8(pos, result, end);
 			 //surrogate pair, not two characters
 			 if (json_unlikely((second > 0xDC00) && (second < 0xDFFF))){
 				result += SurrogatePair(first, second);
@@ -293,7 +293,10 @@
     }
 #endif
 
-json_uchar JSONWorker::UTF8(const json_char * & pos, const json_char * const end) json_nothrow {
+// 利用\uxxxx目前最多4个字节,将转义后的编码写到原先的5个字节(包括'u')
+// 对JSON_UNICODE仍然返回json_uchar(保持兼容), 并且不对result修改
+// 否则返回'\0',并丢弃结果
+json_uchar JSONWorker::UTF8(const json_char * & pos, json_string & result, const json_char * const end) json_nothrow {
 	JSON_ASSERT_SAFE(((long)end - (long)pos) > 4, JSON_TEXT("UTF will go out of bounds"), return JSON_TEXT('\0'););
     #ifdef JSON_UNICODE
 	   ++pos;
@@ -301,10 +304,29 @@
 	   ++pos;
 	   return temp | Hex(pos);
     #else
+       /*
 	   JSON_ASSERT(*(pos + 1) == JSON_TEXT('0'), JSON_TEXT("wide utf character (hihi)"));
 	   JSON_ASSERT(*(pos + 2) == JSON_TEXT('0'), JSON_TEXT("wide utf character (hilo)"));
 	   pos += 3;
 	   return Hex(pos);
+       */
+       pos++;
+       int len = 1;
+       if (json_likely(pos[0] > '0' || pos[1] >= '8')) {
+           len = 3;
+       } else if (pos[1] > 0 || pos[2] >= '8') {
+           len = 2;
+       }
+       json_uchar first = Hex(pos);
+       pos++;
+       json_uchar second = Hex(pos);
+       uint16_t code = ((uint16_t)first << 8) | second;   // pos位置为最后一个字节,下一次循环自增即可
+       result += (json_uchar)(0xFF << (8 - len)) | (code >> ((len - 1) * 6)); 
+       while (--len > 0) {
+           result += (json_uchar)(0x80 | ((code >> ((len - 1) * 6) & 0x3F)));
+       }
+
+       return '\0';  // 丢弃的结果
     #endif
 }
 
@@ -381,7 +403,7 @@
 		  #ifdef JSON_UNICODE
 			 UTF(pos, res, end);
 		  #else
-			 res += UTF8(pos, end);
+			 UTF8(pos, res, end);
 		  #endif
 		  break;
 	   #ifndef JSON_STRICT
@@ -491,6 +513,7 @@
 		  END_MEM_SCOPE
 		  json_uchar hi = ((p & 0x00F0) >> 4) + 48;
 	   #else
+          // 这里还是假设前两位是00
 		  res += JSON_TEXT("00");
 		  json_uchar hi = (p >> 4) + 48;
 	   #endif
diff -uN libjson_7.6.1/_internal/Source/JSONWorker.h libjson_7.6.1.mine/_internal/Source/JSONWorker.h
--- libjson_7.6.1/_internal/Source/JSONWorker.h	2012-03-20 23:18:54.000000000 +0800
+++ libjson_7.6.1.mine/_internal/Source/JSONWorker.h	2013-04-19 06:05:41.000000000 +0800
@@ -43,7 +43,7 @@
 JSON_PRIVATE
     #ifdef JSON_READ_PRIORITY
 	   static json_char Hex(const json_char * & pos) json_nothrow;
-	   static json_uchar UTF8(const json_char * & pos, const json_char * const end) json_nothrow;
+	   static json_uchar UTF8(const json_char * & pos, json_string & result, const json_char * const end) json_nothrow;
     #endif
     #ifdef JSON_ESCAPE_WRITES
 	   static json_string toUTF8(json_uchar p) json_nothrow;

我这个patch只修复了最多3字节utf8的情况..其他的情况中文情况比较少见..就没有处理了..其实应该加一下的..