STL Time Machine Report #9 - Fri 27 Feb 1998 (1998-02-27)

We completed the third and final cycle of Time Machine testing for my
application this week.  A party is scheduled to celebrate sign-off next Monday.  For the last two months we have been testing a group of five applications.  Another group of applications begins testing next week.  
Two months from now the final group testing, including our Unix 
computers.  An additional two month window is being held open for 
retesting of internal applications.  There is also the possibility that non-mission critical applications can be scheduled for testing in the 
fourth group.  

In our final pass we had some irritating problems that showed up on file
compares.  We were worried that we had found a real Year 2000 bug in our
file refresh process, but this was determined to be an existing
production problem, not date related, and one we will fix as soon as
possible.  The bug has been in production for two years, and it had
never been noticed before.  As someone else said, "Test Early, Test
Often".  

When we wrote our test plans we had to take a completely different
approach from our normal regression testing.  Since we needed to
specifically test our most date sensitive functions, we wrote test plans
that were very different from our customary approaches.  In four cycles
of testing we found three Year 2000 problems (all minor) in our first
dry-run cycle.  In the remaining three cycles we found at least four
non-date-related production defects that had never been discovered
before.  

We still have a great deal of testing to do with other applications, 
but I will be doing mostly non-Y2K work for the immediate future, 
production support and new releases.  I've already got a pretty
good backlog of OS/VS COBOL to COBOL II conversions due by May 1.  

OS/VS COBOL

I really must apologize to William Klein here.  I did not get a chance
to try testing the ILBOWAT routine.  I have no documentation on it, and
I suspect it has never been used by our applications programmers. 
Because of various interruptions in my schedule, I did not get to repeat
my test of the OS/VS COBOL compiler and runtime library on 12/31/2000,
but I did run it on 01/01/2001.  All dates were correct (but limited to
two digit years) except for DATE-COMPILED/WHEN-COMPILED, which showed up
as JAN  1, 1901.  This seems to be a pretty consistent error with OS/VS
COBOL.

This is a 1982 COBOL compiler.  IBM does not support the compiler, but
will guarantee the compiled programs as long as you re-link them and
execute with LE/MVS.  OS/VS COBOL has so many limitations it really is a
dead weight that will hold you back.  If you cannot convert to COBOL for
MVS and VM, you should at least convert to VS COBOL II, release 4.0. 
IBM will support that compiler until March of 2001, but the VS COBOL II
runtime libraries will likely be off IBM support by 01/01/2001.  They
will guarantee COBOL II with LE/MVS.

If you're already running VS COBOL II, it's much easier to convert to
LE/MVS and to COBOL for MVS and VM. 

Previous issues of the STL Time Machine Reports can be found at:

http://home.att.net/~arnold.trembley/tmr.htm

--
Arnold Trembley     
http://home.att.net/~arnold.trembley/
"Y2K?  Because Centuries Happen!"