Predicting path of a body/shape

[arahlf]

I'm not exactly sure what to take from that link. Is it recommending the use of a constant time step value instead of using what's supplied to love.update? I'm not using any kind of gravity in my test. The only other thing I can figure is that because I'm using the timestep supplied from love.update, the ball collides with the wall (and possibly bites into it a bit?) resulting in inconsistent contact points. Is that possible?

Also, since box2d is non-deterministic, if I have 2 worlds (1 simulation, 1 actual), doesn't that mean the simulation might not actually prove to be true?

Thanks for the replies!

[Taehl]

Box2D is non-deterministic, even with a fixed timestep?

[arahlf]

I guess I'm not sure, I was just going off of what ivan said earlier.

[arahlf]

Actually with the constant time step the collisions seem to be consistent. Going to give the simulation a try w/ that

[ivan]

I'm not exactly sure what to take from that link. Is it recommending the use of a constant time step value instead of using what's supplied to love.update?

The Box2D documation says that you should use a constant timestep.
I think that some joints may become unstable when your timestep varies too much.

Also, since box2d is non-deterministic, if I have 2 worlds (1 simulation, 1 actual), doesn't that mean the simulation might not actually prove to be true?!

You probably won't notice any difference when your simulation constists of a finite number of billiard balls and you're not trying to predict what happens too far ahead.

Box2D is non-deterministic, even with a fixed timestep?

I'm pretty sure that - yes even with a fixed timestep the simulation might differ each time.
I have to look it up though, not sure if this is caused by how the Box2D solver works or if it's a rounding issue.

[bartbes]

In any case you can't trust it to be the same.
(Of course the question arises whether this needs to be completely accurate.)