Patch for BUG#30472: libmysql doesn't reset charset,

insert_id after succ. mysql_change_user() call.

See also WL 4066.
  
This bug reveals two problems:
  - the problem on the client side which was described originally;
  - the problem in protocol / the server side: connection context
    on client and server should be like after mysql_real_connect()
    and be consistent. The server however just resets character
    set variables to the global defaults.

The fix seems to be as follows:
  - extend the protocol so that the client be able to send
    character set information in COM_CHANGE_USER command;
  - change the server so that it understands client character set
    in the command;
  - change the client:
    - reset character set to the default value (which has been
      read from the configuration);
    - send character set in COM_CHANGE_USER command.

sql/sql_parse.cc

@@ -876,7 +876,13 @@ bool dispatch_command(enum enum_server_command command, THD *thd,
       my_message(ER_UNKNOWN_COM_ERROR, ER(ER_UNKNOWN_COM_ERROR), MYF(0));
       break;
     }
-    db_length= packet_end - db - 1; /* do not count the trailing '\0'  */
+    db_length= strlen(db);
+
+    uint cs_number= 0;
+
+    if (db + db_length < packet_end)
+      cs_number= (uchar) *(db + db_length + 1);
+
     /* Convert database name to utf8 */
     db_buff[copy_and_convert(db_buff, sizeof(db_buff)-1,
                              system_charset_info, db, db_length,
@@ -921,6 +927,12 @@ bool dispatch_command(enum enum_server_command command, THD *thd,
 #endif /* NO_EMBEDDED_ACCESS_CHECKS */
       x_free((uchar*) save_db);
       x_free((uchar*)  save_security_ctx.user);
+
+      if (cs_number)
+      {
+        thd_init_client_charset(thd, cs_number);
+        thd->update_charset();
+      }
     }
     break;
   }