ICE Kinematics: strangely [solved]

[Hirazi Blue]

I have been doing my first steps in ICE Kinematics and generally I would say it's highly amusing (to use a phrase right out of "production" ), even though Softimage sometimes crashes totally unexpectedly (but that might as well be silly me, doing something wrong).
But one thing has happened in conjunction with the ICE Kinematics twice now, which I cannot explain: after a lot of tweaking I manage to get a ICE constraints driven animation (almost) just right, only to find, that after saving and reopening the same scene the animation is way off again, undoing my hard work at tweaking & more importantly: making tweaking utterly pointless...

Has anyone seen this happen yet? For obvious reasons it would be hard to send a scene to support...

[grahamef]

It sounds like you might be creating cycles. Those can cause indeterminancy as well as crashes.

Can you provide any more details about what you were doing or trying to do?

[Hirazi Blue]

Yes, it might very well be that I'm inadvertently creating some cycles (as I am relatively new at this), but wouldn't one expect a saved file that seemed to work, not to start with the old errors (already taken "care of" before saving) again after reload? Or am I missing something completely?

To be overly clear: this is an issue with reopening saved ICE Kinematics scenes to find something seems to have "demonically" changed between saving and reloading (???)...

[grahamef]

It's because Softimage uses a pull model and stops as soon as it detects a cycle.

Let's say that it wants to update A, and A depends on B, so it needs to evaluate B first. B depends on C, so it needs to evaluate C first. C depends on A, so it determines that there's a cycle and uses the current value of C.

However if it wants to update B, it needs to evaluate C, which means that it needs to evaluate A, which means that it needs to evaluate B, at which point it determines that there's a cycle and uses the current values of A.

So in this situation, you can get different results depending on whether Softimage tries to update A, B, or C first. That's the indeterminancy that I was referring to above.

[Hirazi Blue]

Ah, now I get it...
Silly me...
Thanks for the info...