RB position/linear velocity/angular velocity is NaN

[korokuchum]

Library and Version

5.6.1

Operating System

Windows 11

Steps to Trigger Behavior

No certain steps

Expected Behavior

No Nans present after the simulation

Actual Behavior

Rigid body position, linear ang angular velocity is NaN after simulation step.

I can't figure out where those NaNs came from. I used debug PhysX build and added a bunch of asserts to vec3/vec4/quat/matrix etc. and can't catch the moment when those NaNs appear.
Thus I concluded that it is caused either by invalid memcopy or sse operation.
Could you please tell me where should I look to figure this out?

[korokuchum]

I've tracked down nans to friction patches.
body0/1Anchors have nans for some reason.
If you could let me know where it is written I probably can fix this.

[vreutskyy]

Hi @korokuchum. Some code snippet or a repro sample could help us investigate this issue. Otherwise it's hard to tell what's going on there.

[korokuchum]

That's the thing. I can't isolate it in my project.
It happens spontaneously and it seems there is no common variable leading to this behavior.

[korokuchum]

Looks like update of PhysX library from 5.3.1 to 5.6.1 caused the issue but I have no idea why.
I haven't found any clues in the changelogs.
If you have any ideas for me I would love to hear them.