>I think I remember Keith saying something to the effect that Gremlins
>generates a sequence of random events, and those events are apt to change
>with different PC OSs and POSE ROMs. I never felt that Gremlins 0 would
>produce the same sequence of events on my 95 machine as it does on my 98
>machine, I could be wrong (This is where Keith could chime in and prove or
>disprove me).
The random number *sequence* will always been the same given the same starting
seed, regardless of the host platform or the Palm OS version. However, the
purpose to which that number is applied may differ.
The chief reason for, say, Gremlin #0 to produce different results between
different runs is that you're using different versions of Poser. The Gremlins
mechanism is often changing. For instance, a Gremlin used to try to click on
invisible form objects. That attempt would usually result in no effect,
effectively making that attempt a no-op. When we changed that behavior, that
point in the sequence of events would result in a click on a visible button,
effectively making that attempt successful. At that point, history would
diverge. For instance, in the first scenario, nothing would happen; in the
second instance, an alert might appear. The Gremlins engine would then take
different routes, resulting in a different sequence of Gremlin events.
Another reason for seeing different results might be because of host timing
differences in the areas of serial or network communications. If these external
sources provide information at different times between runs, then applications
will react to them in different contexts, again causing history to diverge.
Other than that, Gremlins should be pretty deterministic. It's been a while
since I've done this test, but I used to be able to run 10,000 Gremlins over the
Memopad on my Mac, Windows, and Linux computers, and they'd all end up in the
same place.
-- Keith Rollin
-- Palm OS Emulator engineer
