Details
-
Bug
-
Status: Stalled (View Workflow)
-
Critical
-
Resolution: Unresolved
-
10.0.13, 10.1(EOL), 10.2(EOL), 10.3(EOL), 10.4(EOL), 10.5, 10.6, 10.7(EOL), 10.8(EOL), 10.9(EOL), 10.10(EOL)
-
None
-
10.0.20, 10.0.24
Description
CREATE TABLE t1 (a VARCHAR(3)) ENGINE=MyISAM; |
INSERT INTO t1 VALUES ('foo'),('bar'); |
|
|
CREATE TABLE t2 (i INT, b VARCHAR(3)) ENGINE=MyISAM; |
CREATE ALGORITHM=MERGE VIEW v2 AS SELECT * FROM t2; |
INSERT INTO t2 VALUES (4,'foo'),(5,'baz'); |
|
|
PREPARE stmt FROM "SELECT * FROM t1 WHERE NOT EXISTS ( SELECT i FROM v2 WHERE b = a )"; |
EXECUTE stmt; |
EXECUTE stmt; |
Actual result:
EXECUTE stmt; |
a
|
bar
|
EXECUTE stmt; |
a
|
foo
|
bar
|
Expected result:
EXECUTE stmt; |
a
|
bar
|
EXECUTE stmt; |
a
|
bar
|
revision-id: knielsen@knielsen-hq.org-20140902120701-txesdmkczel84qgf
|
revno: 4369
|
branch-nick: 10.0
|
Attachments
Issue Links
- is blocked by
-
-
- Stalled
-
Activity
The problem is that Item_in_subselect::setup_mat_engine() uses select_engine->join->fields_list which on the second execution chaged due to transformation rollback!!!
revision-id: 016f44b15455d2ca0018902cdc78d1cb29a14423 (mariadb-10.0.23-30-g016f44b)
parent(s): 089b65fd5e8ecd1971cf4c6c372e20575ff4ca9e
committer: Oleksandr Byelkin
timestamp: 2016-02-14 16:52:56 +0100
message:
MDEV-6704: Wrong result (extra rows) on 2nd execution of PS with exists_to_in, MERGE view
The problem was that after first execution permanent exists-to-in transformation became broken back by rolling back temporary changes of view resolving.
Added method to rewrite temporary changes.
—
Attempted to do a review pass, discussed with sanja. Current approach works but in my opinion change_item_tree() is evil and should not be extended.
I'd like to come up with some other solution. I need to think about this.
Second time used other engine, so probably passing the same JOIN_TAB address to it is just a coincidence or (using wrong JOIN (small probability, because it should be freed)).