Details
-
Bug
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Fixed
-
10.0(EOL)
Description
InnoDB and XtraDB crash on the following test case:
|
a.opt |
--max-allowed-packet=128M
|
--innodb-log-file-size=210M
|
|
a.test |
source include/have_innodb.inc;
|
create table foo (id varchar(37) not null, content longblob) engine=INNODB;
|
insert into foo (id, content) values('xyz', '');
|
update foo set content=repeat('a', 43941888) where id='xyz';
|
Note that if the blob gets larger (45000000) or innodb-log-file-size gets smaller (209M) then there is no crash, instead there's an error InnoDB: The total blob data length (44999232) is greater than 10% of the total redo log size (440401920). Please increase total redo log size.
Attachments
Issue Links
Activity
How important this really is, I do not follow why someone would run their database using debug build ?
commit 99a2c061d7bfc181b04a0adc14ce036fab821d85
Author: Jan Lindström <jan.lindstrom@mariadb.com>
Date: Tue Mar 17 20:35:05 2015 +0200
MDEV-7754: innodb assert "array->n_elems < array->max_elems" on a huge blob
update
Problem was that static array was used for storing thread mutex sync levels.
Fixed by using std::vector instead.
Does not contain test case to avoid too big memory/disk space usage
on buildbot VMs.
2015-03-13 15:53:32 7f8920357700 InnoDB: Assertion failure in thread 140226927621888 in file sync0sync.cc line 1325
InnoDB: Failing assertion: array->n_elems < array->max_elems
with 10.0 on commit 184f718fef0101a7559364cb97e22ee568e64c12