Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Duplicate
-
10.1.33, 10.2.15, 10.3.7
-
None
-
IMHO not important (OpenSuSE 42.3)
Description
10.1 commit 31a19683f5289f08439b848da23a28f49ff0accc
– 10.1.33-MariaDB-debug
10.2 commit 87af52d7dd733e71fc7a9e39b882a4fd44f41fec
– 10.2.15-MariaDB-debug
10.3 commit 8fce4065e542ad64438d45f60421241dd383815f
– 10.3.7-MariaDB-debug
CREATE TABLE test . t1 ( c1 INT, c2 INT AS (c1) VIRTUAL ); |
INSERT INTO test . t1 (c1) VALUES(1); |
COMMIT ; |
ALTER TABLE test . t1 ENGINE = InnoDB PARTITION BY KEY(c2) PARTITIONS 4 ; |
SHOW CREATE TABLE t1; |
Table Create Table |
t1 CREATE TABLE `t1` ( |
`c1` int(11) DEFAULT NULL, |
`c2` int(11) AS (c1) VIRTUAL |
) ENGINE=InnoDB DEFAULT CHARSET=latin1 |
/*!50100 PARTITION BY KEY (c2)
|
PARTITIONS 4 */
|
CHECK TABLE t1 EXTENDED; |
Table Op Msg_type Msg_text
|
test.t1 check error Found a misplaced row
|
test.t1 check error Partition p0 returned error
|
test.t1 check error Upgrade required. Please do "REPAIR TABLE `t1`" or dump/reload to fix it!
|
The same test works well for c2 INT AS (c1) PERSISTENT.
Attachments
Issue Links
- is duplicated by
-
-
- Closed
-
Activity
Given scenario have nothing to do with InnoDB. I am able to repeat it with myisam, aria engine.
Given alter also does copy and virtual column doesn't have any index associated with it.
The bug was found when running an RQG test generating roughly the same DDL/DML but concurrent.
There the server asserted with
mysqld: <tree>/sql/ha_partition.cc:4489: virtual int ha_partition::delete_row(const uchar*): Assertion `part_id == m_last_part' failed.
mysys/stacktrace.c:269(my_print_stacktrace)[0x557a07418fe9]
sql/signal_handler.cc:168(handle_fatal_signal)[0x557a06c6c413]
/lib64/libpthread.so.0(+0x10c10)[0x7fc3c5a3ac10]
/lib64/libc.so.6(gsignal+0x37)[0x7fc3c385bf67]
/lib64/libc.so.6(abort+0x13a)[0x7fc3c385d33a]
/lib64/libc.so.6(+0x2dd66)[0x7fc3c3854d66]
/lib64/libc.so.6(+0x2de12)[0x7fc3c3854e12]
sql/ha_partition.cc:4491(ha_partition::delete_row(unsigned char const*))[0x557a073d3bdc]
sql/handler.cc:6285(handler::ha_delete_row(unsigned char const*))[0x557a06c7d607]
sql/sql_delete.cc:253(TABLE::delete_row())[0x557a06e117bb]
sql/sql_delete.cc:748(mysql_delete(THD*, TABLE_LIST*, Item*, SQL_I_List<st_order>, unsigned long long, unsigned long long, select_result))[0x557a06e0eb63]
sql/sql_parse.cc:4922(mysql_execute_command(THD*))[0x557a0698a14a]
sql/sql_parse.cc:8001(mysql_parse(THD*, char*, unsigned int, Parser_state*, bool, bool))[0x557a06993b8c]
sql/sql_parse.cc:1848(dispatch_command(enum_server_command, THD*, char*, unsigned int, bool, bool))[0x557a069813b0]
sql/sql_parse.cc:1391(do_command(THD*))[0x557a0697fdfc]
sql/sql_connect.cc:1402(do_handle_one_connection(CONNECT*))[0x557a06ae3635]
sql/sql_connect.cc:1309(handle_one_connection)[0x557a06ae33c2]
/lib64/libpthread.so.0(+0x8724)[0x7fc3c5a32724]
/lib64/libc.so.6(clone+0x6d)[0x7fc3c3913e8d]
Query (0x7fc328012c10): DELETE FROM test.t1 /* E_R Thread2 QNO 60 CON_ID 16 */
There is no separate bug report for that assert because I assume the base reason for the assert
is the same like for the current bug. So please show up when having a fix for the current bug
and than I will rerun my RQG test.
RQG grammar:
query:
DELETE FROM test . t1 ; INSERT INTO test . t1 (c1) VALUES(1),(2),(3),(4),(5),(6) ; COMMIT ;
thread_connect:
SET SESSION SQL_MODE = 'traditional' ; SET AUTOCOMMIT = 0 ; SET SESSION lock_wait_timeout = 100 ; SELECT SLEEP(5);
thread1:
ALTER TABLE test . t1 ENGINE = InnoDB PARTITION BY KEY(c2) PARTITIONS 4 ;
thread1_init:
SET SESSION lock_wait_timeout = 20 ; SET SESSION innodb_lock_wait_timeout = 10 ; CREATE TABLE test . t1 ( c1 INT, c2 INT AS (c1) VIRTUAL );