Desperate attempt to push part of prepared statements cleanup which was

reviewed in Saint-Petersbourg (including post-review fixes).


include/my_sys.h:
  Added clear_alloc_root (reset alloc root without freeing its memory)
sql/item.h:
  - rename setup_param -> set_parap (function assigns parameter value to item)
sql/mysql_priv.h:
  - all return values are void, because return value is never checked in
      dispatch_command
  - removed unused declaration of setup_param_functions
sql/protocol.h:
  - unused declarations of setup_params_data* removed
sql/sql_class.cc:
  Cleanup:
    - bzero(mem_root) replaced with clear_alloc_root
    - query_id and command members moved back to THD from Statement
  Assignment of mem_root, free_list, query_id and command optimized
  away from set_statement().
sql/sql_class.h:
  - query_id and command moved back to THD from Statement
sql/sql_lex.h:
  - better type for param_list
  - param_count is the same thing as param_list.elements
sql/sql_parse.cc:
  - comments for dispatch_command
sql/sql_prepare.cc:
  Cleanup:
  - added comments to many functions and removed trailing spaces in many 
    lines, some stale comments removed.
  - it's faster to iterate using pointers, than classes
  - Renames: error_in_prepare renamed to get_longdata_error (because it is set
    when there is an error in mysql_send_longdata, rather than in
    mysql_prepare), embedded versions of placeholder assignement functions
    now have prefix emb_, setup_ functions renamed to set_, because they
    perform assignment, not installation, setup_params_data now doesn't
    call insert_params and was renamed to setup_set_param_functions,
  - find_prepared_statement should not send error if called from no-reply
    calls, like mysql_stmt_reset
  - error reporting is checked up, to always report errors and not report
    errors twice. send_prep_stmt can be done mostly in send_prepare_result, 
    rather than in test_* functions.
  - now we don't need to reinit THD->mem_root/free_list in mysql_stmt_execute,
    because it's not reset there.
tests/client_test.c:
  - removed second call to test_subqueries

sql/sql_parse.cc

@@ -1232,10 +1232,34 @@ bool do_command(THD *thd)
 		       command_name[command]));
   }
   net->read_timeout=old_timeout;		// restore it
+  /*
+    packet_length contains length of data, as it was stored in packet
+    header. In case of malformed header, packet_length can be zero.
+    If packet_length is not zero, my_net_read ensures that this number
+    of bytes was actually read from network. Additionally my_net_read
+    sets packet[packet_length]= 0 (thus if packet_length == 0,
+    command == packet[0] == COM_SLEEP).
+    In dispatch_command packet[packet_length] points beyond the end of packet.
+  */
   DBUG_RETURN(dispatch_command(command,thd, packet+1, (uint) packet_length));
 }
 #endif  /* EMBEDDED_LIBRARY */
 
+/*
+   Perform one connection-level (COM_XXXX) command.
+  SYNOPSIS
+    dispatch_command()
+    thd             connection handle
+    command         type of command to perform 
+    packet          data for the command, packet is always null-terminated
+    packet_length   length of packet + 1 (to show that data is
+                    null-terminated) except for COM_SLEEP, where it
+                    can be zero.
+  RETURN VALUE
+    0   ok
+    1   request of thread shutdown, i. e. if command is
+        COM_QUIT/COM_SHUTDOWN
+*/
 
 bool dispatch_command(enum enum_server_command command, THD *thd,
 		      char* packet, uint packet_length)