[MDEV-6780] Regression select @var := behavior from MySQL 5.5 Created: 2014-09-24  Updated: 2014-10-10  Resolved: 2014-09-26

Status: Closed
Project: MariaDB Server
Component/s: OTHER
Affects Version/s: 5.5.37
Fix Version/s: 10.0.6

Type: Bug Priority: Major
Reporter: Stephen Blackstone Assignee: Elena Stepanova
Resolution: Fixed Votes: 0
Labels: None
Environment:

OSX, CentOS



 Description   

There is a difference between MariaDB's behavior and MySQL 5.5 when a query is doing select @var := count(*). MySQL would return the number of rows matched while Maria returns 0.. Its unclear to me which behavior is actually correct.

While I personally wouldn't have used the := syntax here, someone else did and it broke our application a little bit..

------------------------------------------------------

Minimal Example:

drop database if exists `test_example`;
 
create database test_example;
 
use test_example;
 
CREATE TABLE `blarg` (
  `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `v` varchar(10),
  PRIMARY KEY (`id`)
) ENGINE=InnoDB;
 
 
insert into blarg (`v`) values ('AAA');
insert into blarg (`v`) values ('AAA');
insert into blarg (`v`) values ('AAA');
insert into blarg (`v`) values ('BBB');
insert into blarg (`v`) values ('BBB');
insert into blarg (`v`) values ('CCC');
 
MariaDB [test_example]> select count(*), @a := count(*), v from blarg group by v;
+----------+----------------+------+
| count(*) | @a := count(*) | v    |
+----------+----------------+------+
|        3 |              0 | AAA  |
|        2 |              0 | BBB  |
|        1 |              0 | CCC  |
+----------+----------------+------+
3 rows in set (0.00 sec)
 
 
 
mysql> select count(*), @a := count(*), v from blarg group by v;
+----------+----------------+------+
| count(*) | @a := count(*) | v    |
+----------+----------------+------+
|        3 |              3 | AAA  |
|        2 |              2 | BBB  |
|        1 |              1 | CCC  |
+----------+----------------+------+
3 rows in set (0.00 sec)
 
mysql>
 



 Comments   
Comment by Elena Stepanova [ 2014-09-26 ]

It was fixed in 10.0.5. I can't point at the particular revision, as it was fixed by a big merge of 10.0-base (revno 3817 on 10.0 tree).

It doesn't look critical enough to backport the fix to 5.5; please comment if you disagree.

(I've set 'Fix Version' to 10.0.6 because 10.0.5 is not on the list).

Generated at Thu Feb 08 07:14:31 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.