Date: 2010-12-12 19:51:53 +0100
From: @mlkersten
To: MonetDB5 devs <>
Version: 11.9.7 (Apr2012-SP2) [obsolete]

Last updated: 2012-07-17 14:53:01 +0200

Comment 15337

Date: 2010-12-12 19:51:53 +0100
From: @mlkersten

User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.6; en-US; rv:1.9.2.13) Gecko/20101203 Firefox/3.6.13
Build Identifier:

The attach00.mal test generates
!FATAL: BATpropcheck: BAT tmp_365(245) has inconsistent descriptor 65536 (200000)

The reason seems that the BAT descriptor in BBP.dir contains the initial
capacity 256 after BAT.save, while the file itself has only size equal
to the number of entries stored (GDKsave does not look at capacity, only
the free pointer)

A retrofitting patch in attach could avoid it, but it may be indicative for
a more serious error. Being triggered in other circumstances as well.

print *b->T
$14 = {id = 0x7ffff68471db "t", width = 4, type = 6 '\006', shift = 2 '\002', sorted = 0 '\000', varsized = 0 '\000', key = 0 '\000', dense = 0 '\000', nonil = 1 '\001', nil = 0 '\000', unused = 0 '\000',
align = 1001234, nosorted_rev = 0, nokey = {2, 3}, nosorted = 3, nodense = 0, seq = 0, heap = {maxsize = 8, free = 8, size = 8, base = 0xd534d8 "\001", filename = 0xac66c8 "03/365.tail",
storage = 0 '\000', copied = 0 '\000', hashash = 0 '\000', forcemap = 0 '\000', newstorage = 0 '\000', dirty = 0 '\000', parentid = 0}, vheap = 0x0, hash = 0x0, props = 0x0}
(gdb) print *b->U
$15 = {deleted = 0, first = 0, inserted = 0, count = 21, capacity = 256}

Upon reload this triggers the fatal.

Reproducible: Always

Comment 15884

Date: 2011-07-04 15:26:14 +0200
From: @sjoerdmullender

Changeset 1ed011aa3aec made by Sjoerd Mullender sjoerd@acm.org in the MonetDB repo, refers to this bug.

For complete details, see http//devmonetdborg/hg/MonetDB?cmd=changeset;node=1ed011aa3aec

Changeset description:

Added checks to BBPimportEntry.
This should alleviate bug #2751: it now fails with an error message
instead of causing a crash.

Comment 15885

Date: 2011-07-04 15:29:25 +0200
From: @sjoerdmullender

Probably the main cause of errors for this bug was that there weren't enough consistency checks. For instance, a call to lstat wasn't checked (it failed) but the "result" (the struct stat buffer) was used anyway.
I think the test should be reconsidered. As is, the BAT being attached was not saved to disk at the time of the attach. This now causes the attach to fail, but that may not be the purpose of the test.

Comment 16019

Date: 2011-07-29 10:58:42 +0200
From: @sjoerdmullender

Apr2011-SP2 has been released.

Comment 16074

Date: 2011-08-05 14:00:45 +0200
From: @sjoerdmullender

This is now a MonetDB5 problem: the test is not correct.

Comment 16264

Date: 2011-09-16 15:09:13 +0200
From: @sjoerdmullender

The Aug2011 version has been released.

Comment 17492

Date: 2012-07-17 14:53:01 +0200
From: @grobian

this sounds like fixed to me