NewtonCustomJoints on linux64/32

[Carli]

Here you have:


In the menu you press "ctrl+s" and then "s".
Ingame, you press "f" the error will happen.
the code that causes the error is in "src/newtonjoints.pas".

[Julio Jerez]

when you are calling

Code: Select all

NewtonUserJoint *CreateCustomPlayerController (const dFloat* localFrame, const NewtonBody* player, dFloat maxStairStepFactor, dFloat cushion)

with maxStairStepFactor = 1.0

maxStairStepFactor is a fraction of the palyer height, which is determine from the size of the collision shape in the body alone teh up axis of localFrame.
is you pass 1.0 it means the steps the player can take are as high or higher that he head, it is like a person having his knees above his head
the critiria I use to set that paramenter is in general for most people the knee height is the same % of his own height.

Code: Select all

joint =  CreateCustomPlayerController (localFrame, playerBody, 0.2, 0.05)

it should be fine

I tryed by hacking the value and it does not crash, but you have some other problems, I think, when I walk around if I hit a catus I bounce off very fast.
I do not know if that is part of the game or something you have not ajusted because the game kept crashing.
So I can walk around as long as I do not hit anything, then I lose control.

[Carli]

Yes and that's what i want to fix by using the built in character controller.

The landscape is represented as large blocks of compound collision. Each compound collision contains "needles", that means a convex hull with a high y size of 8 points. In an other post you said something that the collision detection of some large thin objects is not really exact. Maybe that's the thing?

Edit: Thanks, it's working with the new values

[Carli]

An other question:
Which values should I pass for »heading« in SetVelocity of the Player?
Sometimes, the Player makes a 180 turn and i'm walking backwards.
Is it a bug or a feature? Which values do i have to pass that this does not happen?

Edit: sorry, fault on my side.

[Carli]

Very strange.

On 64bit, the front and side vector have the opposite direction than in the 32bit version.

Any idea why?

Edit: ok, it's again my fault (i used mem garbage on the stack for calculating stuff)