CannonJS Plugin bug (v2.4)

[santarcade]

Hi guys,

I upgraded to BabylonJS 2.4 and I think I've found a bug on the CannonJS plugin.

It's line 32 of babylon.cannonJSPlugin.ts:

this.world.step(this._fixedTimeStep, this._useDeltaForWorldStep ? delta * 1000 : 0, 3);

If I leave the default _useDeltaForWorldStep set to true, the game runs 1000 faster than normal, if I set it to false it works fine.

By removing the delta*1000 works as expected, resulting in this patch:

this.world.step(this._fixedTimeStep, this._useDeltaForWorldStep ? delta : 0, 3);

I can't do a PR at the moment, so I'm writing here... but I promise I will be able in the future!

Cheers,

Joez

[GameMonetize]

Hey!

 

Do you reproduce the bug with Oimo?

 

[santarcade]

Hi Deltakosh,

I double checked: OimoJSPlugin does not have the useDeltaForWorldStep parameter.
I created a small playground that shows the issue: http://www.babylonjs-playground.com/#2B84TV#4

I dug a little in the code, and in addition, I think that this accumulator done with time delta, could suffer on another bug when setting a different timeStep via setTimeStep(timeStep).

Best,
Joez

[GameMonetize]

Ok I'm trying to summon @RaananW

 

[RaananW]

Summoned, will take care of that later ☺

[RaananW]

Hi Joez,

I am trying to understand the comment regarding "1000" faster. I see the speed difference between the two, but it is certainly not 1000 faster. care to elaborate? maybe make a short screen capture so I can see how you see things?

[santarcade]

Hi RaananW,
you are right: the speed increase factor is not directly 1 thousand.
I wrote some trajectory prediction code to estimate when a body launched with given velocity will hit the ground.

If I keep just delta, the collision will be matching the prediction. If I keep delta*1000, the body will hit the ground way earlier than expected (physics is altered).
This is indeed seen in the playground: same forces on same bodies end up in two different behaviors.

What do you think?
Thank you,
Joez
 

[santarcade]

I've been playing even with setting different timeSteps: if useDeltaForWorldStep is true, after a while it seems the world simulation losses sync (I guess the accumulator jumps in, feeling the need to recover steps).

So far, I have been able to achieve a stable and consistent results changing the line with

this.world.step(this._fixedTimeStep, this._useDeltaForWorldStep ? delta*(this._fixedTimeStep/(1/60)) : 0, 3);

Does it make sense to you?
Best,
Joez
 

[santarcade]

Hi guys,
I tested even OimoJS and my conclusion is that there is definitively something wrong with physics in version 2.4.

Check this demo up: http://www.babylonjs-playground.com/#2B84TV#5

The timeStep in Oimo is inconsistent: so fast to the point that after a few bounces on the plane, the sphere will pass through it.

I tried to check the code, but could not find a reason for this: any idea?

The only consistent version seems to be the first commented line:

var physEngine = new BABYLON.CannonJSPlugin(true);

Joez