Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
1.2.5
-
None
-
None
-
2020-4, 2020-5, 2020-6, 2020-7
Description
MariaDB [smtp_email_validator]> select count(null) over () from logs where timestamp > now() - interval 1 hour;
|
ERROR 1815 (HY000): Internal error: InetStreamSocket::readToMagic: Remote is closed
|
MariaDB [smtp_email_validator]> show warnings\G
|
*************************** 1. row ***************************
|
Level: Error
|
Code: 1815
|
Message: Internal error: InetStreamSocket::readToMagic: Remote is closed
|
*************************** 2. row ***************************
|
Level: Error
|
Code: 1030
|
Message: Got error 1815 "Unknown error 1815" from storage engine InfiniDB
|
*************************** 3. row ***************************
|
Level: Error
|
Code: 1815
|
Message: Internal error: /data/buildbot/bb-worker/centos7/mariadb-columnstore-engine/dbcon/execplan/clientrotator.cpp: Could not get a connection to a ExeMgr
|
3 rows in set (0.000 sec)
|
If we use an expression as a `count ` argument, it counts all rows including the ones containing NULLs
Attachments
Activity
Looking at colIn == -1 to check for NULL has the incorrect assumption that the -1 is caused by a NULL parameter. What is actually happening is that because the parameter is a constant, the field index is -1. Any constant will do this.
This solution results in any constant parameter being treated as NULL, which is incorrect.
A more correct solution is to look at the value of fConstantColumns[0] for the constant value. Unfortunately, the value inserted here for NULL is 0 (incorrect). Perhaps adding a line in ConstantColumn.cpp after if (fType == ConstantColumn::NULLDATA) to push the actual NULL value in would work.
I have updated change to work for other constants as well as NULLs.
Build verified: 1.4.4-1m 1.5.0-1 source
1.4.4-1
/root/ColumnStore/buildColumnstoreFromGithubSource/server
commit d8ff39957275afc7a870487631cd3b3be5eb8818
Author: Rasmus Johansson <razze@iki.fi>
Date: Mon May 4 10:10:07 2020 +0000
MDEV-22273 jUnit patch: xml test result differs from MTR output in case if retry
/root/ColumnStore/buildColumnstoreFromGithubSource/server/engine
commit 4c275557633edb60905faa500990d55a6834951d
Merge: aa054a9 47f2933
Author: David.Hall <david.hall@mariadb.com>
Date: Tue May 5 12:34:03 2020 -0500
Merge pull request #1179 from pleblanc1976/update-libs3-ref
Updated the s3 lib.
1.5.0-1
/root/ColumnStore/buildColumnstoreFromGithubSource/server
commit 43b7e5d29a5480214cee8317a4625b749ccffbaf
Author: Rasmus Johansson <razze@iki.fi>
Date: Mon May 4 10:10:07 2020 +0000
MDEV-22273 jUnit patch: xml test result differs from MTR output in case if retry
/root/ColumnStore/buildColumnstoreFromGithubSource/server/engine
commit 368c4fac059d5cf4f1596e56ea7b1e729e29ec49
Merge: 15a9efa 4bc408c
Author: David.Hall <david.hall@mariadb.com>
Date: Tue May 5 12:35:14 2020 -0500
Merge pull request #1178 from pleblanc1976/update-libs3-ref-1.5
Updated s3 lib ref
Reproduced the issue in 1.2.5-1
MariaDB [mytest]> select count(null) over () from orders where o_orderdate > '1997-05-17' - interval 1 hour;
ERROR 1815 (HY000): Internal error: InetStreamSocket::readToMagic: Remote is closed
MariaDB [mytest]> select count(null) over () from orders where o_orderdate > '1998-08-02' - interval 1 hour;
---------------------
| count(null) over () |
---------------------
| 0 |
.
.
.
| 0 |
| 0 |
---------------------
581 rows in set (0.070 sec)
MariaDB [mytest]> select count
from orders where o_orderdate = '1998-08-02';
----------
| count |
----------
| 581 |
----------
1 row in set (0.057 sec)
One may be tempted to ask why anyone would write count(NULL). Automated systems do.