Date: 2015-12-01 18:45:54 +0100
From: Frédéric Jolliton <<frederic.jolliton+monetdb>>
To: SQL devs <>
Version: 11.17.21 (Jan2014-SP3)
CC: martin.van.dinther, @njnes

Last updated: 2016-01-15 11:38:04 +0100

Comment 21623

Date: 2015-12-01 18:45:54 +0100
From: Frédéric Jolliton <<frederic.jolliton+monetdb>>

User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:42.0) Gecko/20100101 Firefox/42.0
Build Identifier:

A query with a serie of UNION ALLs fails because of an assert that trigger when the optimizer reaches 10 levels (of iterations?).

It fails even with optimizer set to `minimal_pipe'. This is worrying :/

Reproducible: Always

Steps to Reproduce:

1. Create a simple table:

CREATE TABLE foo (a INT);

2. Execute the following query:

SELECT 1
FROM (SELECT a FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo
UNION ALL
SELECT 0 FROM foo) AS t;

Actual Results:

The log says:

 mserver5: rel_optimizer.c:7675: _rel_optimizer: Assertion `0' failed.
 database 'test' (12345) was killed by signal SIGABRT

Comment 21624

Date: 2015-12-01 22:48:42 +0100
From: @sjoerdmullender

This only seems to be an actual problem when assertions are enabled. And they're not in binary releases.

Comment 21629

Date: 2015-12-02 14:46:55 +0100
From: Frédéric Jolliton <<frederic.jolliton+monetdb>>

Indeed, we're testing with assert enabled. And it is actually fine with a production release.

We made the assumption that an assert was really pointing at a problem, but you seems to imply that it is benign here. In this case, we will disable this specific assert for our tests.

Thanks.

Comment 21644

Date: 2015-12-07 19:07:00 +0100
From: MonetDB Mercurial Repository <>

Changeset 3c24051e28fa made by Niels Nes niels@cwi.nl in the MonetDB repo, refers to this bug.

For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=3c24051e28fa

Changeset description:

fixed bug #3879, after pushing down project (through the union), we merge
multiple projects directly.

Comment 21645

Date: 2015-12-07 20:33:07 +0100
From: @njnes

improved the optimizer which solve this recursion.

Comment 21657

Date: 2015-12-10 16:52:04 +0100
From: MonetDB Mercurial Repository <>

Changeset f59d2713050c made by Martin van Dinther martin.van.dinther@monetdbsolutions.com in the MonetDB repo, refers to this bug.

For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=f59d2713050c

Changeset description:

Corrected wrong bug number. Should be Bug #3879 instead of 3897

Comment 21658

Date: 2015-12-10 16:55:39 +0100
From: Martin van Dinther <<martin.van.dinther>>

Hi Niels,
Found another test case where the assertion:
mserver5: ../../../dev/sql/server/rel_optimizer.c:7680: _rel_optimizer: Assertion `0' failed.
is still occuring.

See test script: sql/test/BugTracker-2015/Tests/many_union_all.Bug-3879.sql

Comment 21660

Date: 2015-12-12 12:36:26 +0100
From: MonetDB Mercurial Repository <>

Changeset bc157090c301 made by Niels Nes niels@cwi.nl in the MonetDB repo, refers to this bug.

For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=bc157090c301

Changeset description:

fixed many_union_all bug #3879

Comment 21661

Date: 2015-12-12 13:13:28 +0100
From: @njnes

fixed, we now cleanup the extra projects right after a project push down union.