| Bug #26869 | TABLE_LIST::table_name_length inconsistent with TABLE_LIST::table_name | ||
|---|---|---|---|
| Submitted: | 6 Mar 2007 11:48 | Modified: | 18 Jun 2007 18:56 |
| Reporter: | Shan Lu | ||
| Status: | Closed | ||
| Category: | Server: General | Severity: | S3 (Non-critical) |
| Version: | 5.1, mysql-5.2.-falcon-alpha | OS: | Any |
| Assigned to: | Dmitri Lenev | Target Version: | |
[6 Mar 2007 14:24]
Sergei Golubchik
will be fixed by the patch for bug#25578
[23 May 2007 17:44]
Dmitri Lenev
This problem is fixed by the same patch as bug #23667. Patch for 5.1: http://lists.mysql.com/commits/27205
[1 Jun 2007 21:21]
Bugs System
Pushed into 5.1.20-beta
[18 Jun 2007 18:56]
Paul DuBois
No user-visible change. No changelog entry needed.
Description: Based on my understanding, it is better to keep TABLE_LIST::table_name_length consistent with TABLE_LIST::table_name, because one might use the former to read or write the latter string. In following mysql_create_like_table function, src_tables_list's name-related fields: db, db_length, table_name and alias are all set. However, table_name_length is not set. The table_name_length field is actually 0 at this point due to the earlier bzero function call. I just feel this very dangerous: string table_name is non-null, while the variable describing its length, i.e. table_name_length, is 0. I think it is at least good habit to add the assignment to src_tables_list.table_name_length. In file sql/sql_table.cc 4600 bool mysql_create_like_table(THD* thd, TABLE_LIST* table, 4601 HA_CREATE_INFO *lex_create_info, 4602 Table_ident *table_ident) 4603 { ... 4687 bzero((gptr)&src_tables_list, sizeof(src_tables_list)); 4688 src_tables_list.db= src_db; 4689 src_tables_list.db_length= table_ident->db.length; 4690 src_tables_list.lock_type= TL_READ; 4691 src_tables_list.table_name= src_table; 4692 src_tables_list.alias= src_table; 4693 4694 if (simple_open_n_lock_tables(thd, &src_tables_list)) 4695 DBUG_RETURN(TRUE); How to repeat: no test case. Find this by source code reading.