DB2 Service would like to draw your attention to a high-impact pervasive
(HIPER) APAR (defect) that affects DB2 Universal Database Version 8.1 with
FixPak 4.


APAR number:
============
JR19507
http://www-306.ibm.com/cgi-bin/db2www/data/db2/udb/winos2unix/support/aparlib.d2w/display_apar_details?aparno=JR19507

DB2 version affected:
===================
DB2 Universal Database for Linux, Unix, and Windows Version 8.1 with FixPak
4.

Description:
==========
Under very specific conditions, a SELECT statement on a single column using
the DISTINCT keyword may return more results (rows) than expected.

This problem will occur only if *all* of the following conditions are
satisfied:

You are running DB2 Universal Database Version 8.1 with FixPak 4.
Earlier versions and FixPak levels are not affected.
A SELECT is issued against one (and only one) large integer (4 byte)
column. Selecting more than one column, or selecting one column that is
anything other than large integer, will not trigger the problem.
At least two predicates exist in the statement. No predicates, or just
one predicate, will not trigger the problem.
The DISTINCT keyword is used. However, this problem is not strictly
limited to the use of the DISTINCT keyword. If the plan for any SQL
statement chosen by the optimizer calls for establishing the
distinctness of a single 'large' integer while applying two or more
predicates, this problem can be recreated, but may not be in every case.

For example, the following statement satisfies all of the conditions
necessary to recreate the problem (setting aside the caveats about
distinctness):

select distinct c1 from t1 where c1 > 5 and c2 = 'T'

However, the following statements do *not* satisfy the conditions necessary
to trigger the problem:

select distinct c1 from t1
or
select distinct c1 from t1 where c1 > 5

Workaround:
===========
None.

Fix:
===
This problem will be addressed in FixPak 4a which will be available for
download during the week of January 26 at the following web site: