Details
-
Bug
-
Status: Open (View Workflow)
-
Minor
-
Resolution: Unresolved
-
11.4, 11.8, 12.3, 13.1
-
Can result in data loss
Description
In the test case below, if the UPDATE is binlogged as a statement (or not binlogged at all), it succeeds; but if it is going to be binlogged in row format, it fails with ER_DIVISION_BY_ZERO.
Besides the inconsistency, it easily causes replication failures when the slave runs with log-slave-updates and ROW binlog format. In the test case it is set to RBR explicitly, but there can be also runtime reasons why the master runs in MBR but the slave in RBR.
--source include/have_innodb.inc
|
--source include/have_binlog_format_mixed.inc
|
--source include/master-slave.inc
|
|
|
--connection slave
|
--source include/stop_slave.inc
|
SET GLOBAL binlog_format = ROW; |
--source include/start_slave.inc
|
|
|
--connection master
|
CREATE TABLE t1 (a INT, b INT, c INT, v INT AS (a DIV b) PERSISTENT) ENGINE=InnoDB; |
INSERT IGNORE INTO t1 (a,b,c) VALUES (1,0,0); |
|
|
UPDATE t1 SET c=1; |
|
|
--sync_slave_with_master
|
|
|
--source include/stop_slave.inc
|
SET GLOBAL binlog_format = MIXED; |
--source include/start_slave.inc
|
|
|
--connection master
|
DROP TABLE t1; |
--source include/rpl_end.inc |
|
11.4 151631b8cffcbf6510501262e36d950780ee4bc8 |
Last_Errno 1365
|
Last_Error Error 'Division by 0' on query. Default database: 'test'. Query: 'UPDATE t1 SET c=1' |
The behavior changed after this commit in 11.4.11:
commit a4e4a56720c974b547d4e469a8c54510318bc2c9
|
Author: Kristian Nielsen
|
Date: Tue Feb 3 12:51:19 2026 +0100
|
|
|
MDEV-38734: Virtual columns wrongly included in binlog_row_image=MINIMAL
|
Before the commit, UPDATE fails regardless whether it's run with statement or row binary logging (or without any), so it would not reach the slave to cause a problem.
Attachments
Issue Links
- is caused by
-
-
- Closed
-