Details
-
Type:
Bug
-
Status: Closed
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: 10.0.4
-
Fix Version/s: 10.0.6
-
Component/s: None
-
Labels:None
Description
SET optimizer_switch = 'outer_join_with_cache=on'; CREATE TABLE t1 (c1 VARCHAR(6)) ENGINE=MyISAM; INSERT INTO t1 VALUES ('s'),('t'); CREATE TABLE t2 (c2 VARCHAR(1)) ENGINE=MyISAM; INSERT INTO t2 VALUES ('a'),('x'); SET join_cache_level = 1; SELECT t2.* FROM t1 RIGHT JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1; SET join_cache_level = 0; SELECT t2.* FROM t1 RIGHT JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1;
Result:
SET join_cache_level = 1; SELECT t2.* FROM t1 RIGHT JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1; c2 SET join_cache_level = 0; SELECT t2.* FROM t1 RIGHT JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1; c2 x
The 2nd result is the correct one.
The failure happens on current 10.0-base (revno 3733) and 10.0 (revno 3856). I found the revision on 10.0-base when it started happening, it was a merge from 5.5:
revno: 3645 [merge] revision-id: sergii@pisem.net-20130606155128-5mytep9v42626tfs committer: Sergei Golubchik <sergii@pisem.net> branch nick: 10.0-base timestamp: Thu 2013-06-06 17:51:28 +0200 message: 5.5 merge
But on some reason I could not reproduce it on 5.5, even after rolling back to the revision which was merged into 10.0-base.
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
The bug happens because sort_table_cond is extracted incorrectly due to the fact that curr_join->tmp_having needs an update of used tables.
The bug can be fixed with the following patch:
=== modified file 'sql/sql_select.cc'
— sql/sql_select.cc 2013-10-17 05:45:31 +0000
+++ sql/sql_select.cc 2013-10-21 22:23:16 +0000
@@ -2903,6 +2903,7 @@ void JOIN::exec_inner()
JOIN_TAB *curr_table= &curr_join->join_tab[curr_join->const_tables];
table_map used_tables= (curr_join->const_table_map |
curr_table->table->map);
+ curr_join->tmp_having->update_used_tables();
Item* sort_table_cond= make_cond_for_table(thd, curr_join->tmp_having,
used_tables,
It looks like the bug can manifest itself only starting with10.0-base.
The fix for this bug was pushed into 10.0-base
The bug is reproducible with a regular join as well:
MariaDB [test]> INSERT INTO t1 values ('a'), ('x');
Query OK, 2 rows affected (0.00 sec)
MariaDB [test]> SELECT * FROM t1 JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1;
Empty set (0.00 sec)
MariaDB [test]> SET join_cache_level = 0;
Query OK, 0 rows affected (0.00 sec)
MariaDB [test]> SELECT * FROM t1 JOIN t2 ON c1 = c2 HAVING c2 > 'a' ORDER BY c2 LIMIT 1;
-----
-----+-----
-----+-----
-----+1 row in set (0.00 sec)