clustered string primary with dup prefixed uk, insert ignore ... on duplicate key update return wrong result

[vivid392845427]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

CREATE TABLE `tbl_16` (
  `col_70` varbinary(444) NOT NULL DEFAULT 'bezhs',
  PRIMARY KEY (`col_70`) clustered,
  UNIQUE KEY `idx_22` (`col_70`(1))
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_bin;
 insert into tbl_16 values('lldcxiyfjrqzgj');
 insert ignore into tbl_16 values ( 'lalozlkdosasfklmflo' ) on duplicate key update col_70 = 'lyhohxtby';

2. What did you expect to see? (Required)

mysql> insert ignore into tbl_16 values ( 'lalozlkdosasfklmflo' ) on duplicate key update col_70 = 'lyhohxtby';
Query OK, 2 rows affected (0.00 sec)

mysql> select * from tbl_16;
+-----------+
| col_70    |
+-----------+
| lyhohxtby |
+-----------+
1 row in set (0.00 sec)

3. What did you see instead (Required)

only occur in 5.0-release, master is the same as expected

mysql>  insert ignore into tbl_16 values ( 'lalozlkdosasfklmflo' ) on duplicate key update col_70 = 'lyhohxtby';
Query OK, 0 rows affected, 1 warning (0.00 sec)

mysql> select * from tbl_16;
+----------------+
| col_70         |
+----------------+
| lldcxiyfjrqzgj |
+----------------+
1 row in set (0.00 sec)

4. What is your TiDB version? (Required)

Release Version: v5.0.2-40-g64b7ecd
Edition: Community
Git Commit Hash: 64b7ecd7938a69c34169cd4edc3055544f51e423
Git Branch: HEAD
UTC Build Time: 2021-06-25 18:24:59
GoVersion: go1.13.7
Race Enabled: false
TiKV Min Version: v3.0.0-60965b006877ca7234adaced7890d7b029ed1306
Check Table Before Drop: false 

[lysu]

this question trigger condition is :

• clustered index string primary key - primary key (col_70)
• a prefixed secondary index with the same column but has prefix length - UNIQUE KEY idx_22 (col_70(1))
• insert ignore into on duplicate update change the column value but not change prefixed part

so , it will trigger handle change logic, but 'handle changed' doesn't make 'index value changed'...so it cause problem 😢

[lysu]

the 5.1 and master use new memdb's stage handle feature, so it can avoid this problem

but 3.0, 4.0, 5.0 ti use old version memdb, we cannot backport master logic and need another fix...

[github-actions]

Please check whether the issue should be labeled with 'affects-x.y' or 'fixes-x.y.z', and then remove 'needs-more-info' label.