Date: 2015-01-14 23:41:36 +0100
From: sorear
To: MonetDB5 devs <>
Version: 11.19.7 (Oct2014-SP1)
CC: @mlkersten

Last updated: 2015-01-29 14:07:28 +0100

Comment 20556

Date: 2015-01-14 23:41:36 +0100
From: sorear

User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_7_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/39.0.2171.95 Safari/537.36
Build Identifier:

There was a crash in RECYCLEdrop on a mserver5 instance today, and I noticed that it's checking that recycleBlk is not null, taking a lock, and then assuming that recycleBlk is still not null - it could have been nulled before the lock-take succeeds.

There appear to be other locking issues in the recycler as the code in the reproduction section crashes in many places.

Reproducible: Always

Steps to Reproduce:

Start and release a database (assumed "test" below), and:

for x in 1 2 ; do perl -MDBI -e 'my $db = DBI->connect("dbi:monetdb:database=test;host=localhost;port=50000", "monetdb", "monetdb"); for (1 .. 1000) { $db->do("SET OPTIMIZER=''recycler_pipe''"); $db->do("SELECT * FROM sys.schemas LIMIT 1"); $db->do("START TRANSACTION"); $db->do("COMMIT") }' & done

Actual Results:

Segfaults almost immediately, but in an unpredictable place.

Expected Results:

No segfaults.

Comment 20558

Date: 2015-01-15 15:01:36 +0100
From: @mlkersten

Thank you for reporting. The Recycler optimizer is alpha-release code and any real-life use case would help its further maturing.
Martin

Comment 20563

Date: 2015-01-20 14:18:36 +0100
From: @mlkersten

Applied the patch in RECYCLEDROP.

Comment 20565

Date: 2015-01-20 18:14:23 +0100
From: MonetDB Mercurial Repository <>

Changeset 22c1bf98c903 made by Martin Kersten mk@cwi.nl in the MonetDB repo, refers to this bug.

For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=22c1bf98c903

Changeset description:

Bug #3649 reported fix.
Don't assume variables won't change in a concurrent situation.

Comment 20597

Date: 2015-01-29 14:07:28 +0100
From: @sjoerdmullender

Oct2014-SP2 has been released.