DB2 FOR LUW / DB2FORLUW / A50 - DB2

The problem can happen in DPF environment only. Queries that 
aggregate on a provable 0 row stream without group by 
clause can return incorrect result when DB2 optimizer uses the 
"0 row" property to optimize query plans, such as removing 
duplicate-eliminating sort. 
 
For example: 
 
select max(IBMREQD) from sysibm.sysdummy1 where 0=1 
union 
select max(IBMREQD) from sysibm.sysdummy1 where 0=1 
; 
 
 
In this query, each union leg should return 1 row of "NULL" 
value. Since this is a distinct UNION, the final result should 
still be 1 row of "NULL" value. 
 
But because DB2 optimizer thinks that each union leg returns 0 
row, it removes the duplicate-eliminating sort that should have 
been done after "UNION ALL". So the final returned result is 2 
rows of "NULL" value, which is incorrect.

**************************************************************** 
* USERS AFFECTED:                                              * 
* DPF (Database Partitioning Feature)                          * 
**************************************************************** 
* PROBLEM DESCRIPTION:                                         * 
* See Error Description                                        * 
**************************************************************** 
* RECOMMENDATION:                                              * 
* Upgrade to DB2 10.5 Fix Pack 5                               * 
****************************************************************

First fixed in DB2 10.5 Fix Pack 5

not known / see Local fix

Date  - problem reported    :
Date  - problem closed      :
Date  - last modified       :

07.11.2014
15.12.2014
15.12.2014