CASE statement produces GDK error on multithreaded database: BATproject does not match always

[monetdb-team]

Date: 2015-11-06 16:27:01 +0100
From: Kevin Boulain <<kevin.boulain>>
To: SQL devs <>
Version: 11.21.5 (Jul2015)
CC: guillaume.savary, @njnes

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

Comment 21473

Date: 2015-11-06 16:27:01 +0100
From: Kevin Boulain <<kevin.boulain>>

User-Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:39.0) Gecko/20100101 Firefox/39.0
Build Identifier:

On my Ubuntu work machine (Intel i5-4590) the bug happens on a multithreaded database with nthreads=4 (default value) but not with nthreads=1.

Also reproducible with a Debian test server (two Intel Xeon E5-2697 v3) with nthreads=56 (default value).

MonetDB version is Jul2015 SP1 (http://dev.monetdb.org/hg/MonetDB/rev/d1dd6e5b7a5a).

mserver5 informations from the Ubuntu machine:
$ ./bin/mserver5 --version
MonetDB 5 server v11.21.11 "Jul2015-SP1" (64-bit, 64-bit oids, 128-bit integers)
Copyright (c) 1993-July 2008 CWI
Copyright (c) August 2008-2015 MonetDB B.V., all rights reserved
Visit http://www.monetdb.org/ for further information
Found 19.5GiB available memory, 4 available cpu cores
Libraries:
libpcre: 8.35 2014-04-04 (compiled with 8.35)
openssl: OpenSSL 1.0.1f 6 Jan 2014 (compiled with OpenSSL 1.0.1f 6 Jan 2014)
libxml2: 2.9.1 (compiled with 2.9.1)
Compiled by: ether@lethe (x86_64-unknown-linux-gnu)
Compilation: gcc -g -Werror -Wall -Wextra -W -Werror-implicit-function-declaration -Wpointer-arith -Wdeclaration-after-statement -Wundef -Wformat=2 -Wno-format-nonliteral -Winit-self -Winvalid-pch -Wmissing-declarations -Wmissing-format-attribute -Wmissing-prototypes -Wold-style-definition -Wpacked -Wunknown-pragmas -Wvariadic-macros -fstack-protector-all -Wstack-protector -Wpacked-bitfield-compat -Wsync-nand -Wjump-misses-init -Wmissing-include-dirs -Wlogical-op -Wunreachable-code
Linking : /usr/bin/ld -m elf_x86_64

Reproducible: Always

Steps to Reproduce:

Generate a CSV big enough (on my work machine, it starts crashing around 200k lines, but it does not crash if I import 100k line multiple times in the same table. On the server, a table which encounter the problem has 54+ millions lines).
$ seq 1000000 > /tmp/test.csv

On the mclient:
sql>create table test (x int);
sql>COPY INTO "test" FROM '/tmp/test.csv';
sql>select 1 from test where CASE WHEN FALSE THEN TRUE ELSE (x > 0) END;

Actual Results:

GDK reported error.
BATproject: does not match always

Expected Results:

+--------------+
| single_value |
+==============+
| 1 |
| 1 |
| 1 |
| 1 |
...

The given request is quite silly (generated and not optimized) but it is not reproducible with the following one that should be equivalent:
sql>select 1 from test where CASE WHEN (x > 0) THEN TRUE ELSE FALSE END;

Comment 21481

Date: 2015-11-09 08:36:45 +0100
From: @njnes

fixed on 2015 branch. The complex expressions of boolean values are now properly handled.

Comment 21485

Date: 2015-11-09 11:01:00 +0100
From: Guillaume Savary <<guillaume.savary>>

I've updates the Juk2015 branch and test that 3851 (complex boolean expressions) is now fixed but not this one. I still have

GDK reported error.
BATproject: does not match always

Comment 21492

Date: 2015-11-09 15:41:43 +0100
From: @sjoerdmullender

With the current version (changeset 56a7f0f475c4) both SELECT queries that are mentioned in the report work for me.
Can you test again, please?

Comment 21494

Date: 2015-11-09 16:13:18 +0100
From: Guillaume Savary <<guillaume.savary>>

It works as expected now. Thanks!

Comment 21495

Date: 2015-11-09 16:30:15 +0100
From: @sjoerdmullender

Thanks for testing.