Slicing Boxes - found out why it was buggy

[Carli]

Hi,

I found out, why/when exactly the boxes where slicing over the ground:
I created a ground box with size 1|1|1 and scaled the matrix to to 1000|20|1000 - the boxes are slicing over the ground box and were keeping rotated
when I scaled the ground to 1000|1|1000, it works fine.

i suspect that if i would let the boxes hit the side of the ground box, the bug would be 1000 times stronger
i also suspect that there is no normalization when using scaled matrices

so my question:
should i use scale matrices or is there an other way to get the collision to an other size... (or maybe you implement a normalization of the normals - which costs computing time)

[JernejL]

As far as i know, the only proper way to scale a body in newton is a convex matrix modifier using a convex hull object, are you using this or setting custom matrices?

[Carli]

i abused the position matrix to set the size.

[JernejL]

i abused the position matrix to set the size.

As i said that's not supported and the results of it are "undefined".. you need to use convex hull modifier, it's the only supported way to change size of objects.

[Carli]

But a convex hull modifier would not affect the graphics...

[JernejL]

But a convex hull modifier would not affect the graphics...

You'll just need to apply additional step to add scaling back to the matrix, what you are doing currently is just not supported..

[Carli]

i didn't know if i said it before but i'm working at a script->physic->graphic interface for a game engine and the idea behind is that the graphics are independend from the rest - that means you could also stream the object matrices per network (just like masterservers do)

the object oriented idea is that you just have interfaces from where you call "methods" but no one knows the code behind it - that means you can replace each code for a new platform for example. i think newton is in this way a bit too specialized and i have to write too many wrappers.

You'll just need to apply additional step to add scaling back to the matrix, what you are doing currently is just not supported..

that all would cause network traffic

[JernejL]

i didn't know if i said it before but i'm working at a script->physic->graphic interface for a game engine and the idea behind is that the graphics are independend from the rest - that means you could also stream the object matrices per network (just like masterservers do)

the object oriented idea is that you just have interfaces from where you call "methods" but no one knows the code behind it - that means you can replace each code for a new platform for example. i think newton is in this way a bit too specialized and i have to write too many wrappers.

You'll just need to apply additional step to add scaling back to the matrix, what you are doing currently is just not supported..

that all would cause network traffic

What's the difference if you call NewtonConvexHullModifierGetMatrix instead of NewtonBodyGetMatrix? that matrix would still be a ordinary 4x4 matrix, so your argument about extra network traffic makes no sense.

It was mentioned on forum for a dozen of times that scaling bodies by setting matrix is unsupported, the only proper way to scale them is using NewtonConvexHullModifierSetMatrix, i'll fix the documentation on wiki to make this clearer.