Re:
> I have just finished a Y2K project at a clients site using Time Machine
> for (extensive) testing. We had testing and productive environment(s) on
^^^^^^^^^
<plug>
Nope. Sorry. It wasn't. :)
We've seen problems found in at least 2 products (M...... and NB.....)
*AFTER* they were tested with one of the other products ... the problems
surfaced only when tested with HourGlass 2000 for the HP 3000.
Why?
Simple...you *MUST* affect file dates when you test. If you don't,
you just aren't testing 100%. In the cases I mention, it was found
that the M and NB products were using the spooler AIF routines to
control spoolfiles. And, guess what, the AIF routines had a bug
with respect to 1999. That bug wasn't found with the other products,
because they use AIF:OS to intercept the simple intrinsics ... they don't
use our technique of intercepting *everything*. As a result, when
a spoolfile was created from a session whose "date" as 2000-01-01,
the spoofile date was 1998-09-09 (or whatever) ... not the simulated date.
When run with HourGlass, the spoolfile date was 2000-01-01, which
exposed the AIF bug.
Now, the problems in the AIF that M and NB found could have been
discovered if they'd set their system clock forward. But, like many
testers, they couldn't afford to do that kind of testing on a separate
machine, so they used a date/time simulator. Unfortunately for them,
only *one* date/time simulator intercepts (and, therefore, simulates)
all date/time calls. HourGlass 2000 for the HP 3000.
</plug>
BTW, if you aren't on MPE/iX 6.0, and if you use any products that
interact with the spooler, you'll want patch MPEKXE9 (or a later version)
to fix the AIF:OS spooler problem.
--
Stan Sieler [log in to unmask]
http://www.allegro.com/sieler.html
|