[1446] PhysBones behavior change? Bones collide even when radius is zero
tracked
bd_
There appears to have been a behavior change with PhysBones in a recent update; now, bones which have a zero radius still collide if they are inside of a fixed PB collider component. This behavior can be observed with this avatar: https://vrchat.com/home/avatar/avtr_a991cdfc-edb9-4254-9e96-07e633787bd6
As a repro, sit down in 6-point FBT and move your head up and down. Your tail will spaz out. Adjusting the Spine collider to not be initially in collision with the tail fixes this. This was not how VRChat behaved prior to the latest live update (1446).
Example of this behavior: https://www.youtube.com/watch?v=FXeKwoD7Pnk
Log In
Kiro
I've looked into this issue but have been unable to produce a situation where colliders interact with zero radius bones, including on the avatar ID listed above.
Following your instructions I was able to reproduce the movement you were noticing in the tail when moving my head. However this doesn't appear to be a result of collision with zero radius bones. Instead, when sitting and moving your head the spine collider is pushed into the second bone in the chain which does have a radius. I tested this with our latest client 2024.2.1 and previous client 2024.1.2 with the same results.
I will mention that head movement affecting the spine bone depends on your IK settings, if you are locked via Head/Hip/Both. Locking to either Head or Both seems to introduce significant motion to the spine. Locking to Hip appears to have much less spine rotation and the tail stays fairly stable.
For now I'm marking this as a non-issue, however if you are able to find a reproduceable test case of colliders affecting a zero radius bone feel free to respond and I'll look into it.
bd_
Thanks for looking into it! It's hard with these kinds of things to tell if it's a new regression after the release lands, since I can't roll back to an older client to test... :/
StormRel
tracked