Date: 2014-10-03 12:35:21 +0200
From: Richard Hughes <<richard.monetdb>>
To: SQL devs <>
Version: 11.17.21 (Jan2014-SP3)
CC: @njnes

Last updated: 2014-10-31 14:14:24 +0100

Comment 20232

Date: 2014-10-03 12:35:21 +0200
From: Richard Hughes <<richard.monetdb>>

Created attachment 299
test script

Use attached test script to reproduce, starting from a blank database.

Build is Oct2014 307281054d25 plus a few patches which Niels has given me.

The important bit of that test script is to have a concurrent transaction open when doing the delete-all, thus ensuring that store_nr_active > 1 and we don't go through the fast path. In that case 'count' is updated and correct (=1 in the attached test) but 'ibase' isn't (=0 in the attached test).

I have no idea how to go about fixing this. Should ibase have been updated in the new copy at commit time? Should the subsequent insert force a partial WAL flush before doing its thing? Help!

Attached file: delete-concurrency-test.py (application/octet-stream, 730 bytes)
Description: test script

Comment 20281

Date: 2014-10-11 18:01:23 +0200
From: MonetDB Mercurial Repository <>

Changeset 9117e7d7bd0f 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=9117e7d7bd0f

Changeset description:

fix for bug #3593. When we move the insert bats into the bid slot we should
set the ibase.

Comment 20282

Date: 2014-10-11 18:01:51 +0200
From: @njnes

fixed by setting the ibase also when we take over the ibat

Comment 20368

Date: 2014-10-31 14:14:24 +0100
From: @sjoerdmullender

Oct2014 has been released.