support CDATA in xml parser

ChangeLog

@@ -1,3 +1,4 @@
+	* supporr CDATA tags in xml parser
 	* use a python python dictionary for settings instead of session_settings object (in python bindings)
 	* optimized metadata transfer (magnet link) startup time (shaved off about 1 second)
 	* optimized swarm startup time (shaved off about 1 second)

include/libtorrent/xml_parse.hpp

@@ -81,6 +81,29 @@ namespace libtorrent
 		
 			// skip '<'
 			++p;
+			if (p != end && p+8 < end && string_begins_no_case("![CDATA[", p))
+			{
+				// CDATA. match '![CDATA['
+				p += 8;
+				start = p;
+				while (p != end && !string_begins_no_case("]]>", p-2)) ++p;
+
+				// parse error
+				if (p == end)
+				{
+					token = xml_parse_error;
+					start = "unexpected end of file";
+					callback(token, start, val_start);
+					break;
+				}
+			
+				token = xml_string;
+				char tmp = p[-2];
+				p[-2] = 0;
+				callback(token, start, val_start);
+				p[-2] = tmp;
+				continue;
+			}
 
 			// parse the name of the tag.
 			for (start = p; p != end && *p != '>' && !isspace(*p); ++p);
@@ -202,7 +225,6 @@ namespace libtorrent
 				*i = save;
 			}
 		}
-	
 	}
 
 }