When applying velocity on a part that’s perpendicular to a rotated wall, so the part moves straight onto the wall’s long surface, it will behave differently compared to the exact same case with a wall that’s not rotated. This works as expected as long as the wall’s orientation on the XZ plane equals to 0°, 90°, 180° or 270°.
Backstory
I was working on a custom character controller when this problem came up which I spent countless hours debugging until I came up with the very simplified reproduction file provided below.
Setup
The moving parts are duplicates and consist of the same properties and configurations.
Modified properties that’re worth looking into are
CustomPhysicalProperties
Wall parts’ Orientation
AlignOrientation’s CFrame, MaxTorque and Responsiveness
LinearVelocity’s VectorVelocity and MaxForce
The moving parts are essentially made up of an AlignOrientation object and LinearVelocity object as shown below.
It’s worth noting that both the moving parts’ Friction values are set to 0 and FrictionWeight values are set to 100 for a desired nearly friction-less behavior.
I believe something internally is causing AlignOrientation’s CFrame property to stay == CFrame.new(), although printing seems to indicate otherwise. Rollback indicates this issue was not occurring before the most recent update
EDIT: AlignOrientation doesn’t seem to work properly specifically in OneAttachment mode
If that’s true, just as a note to the engineers, this has to be related to BodyGyro as well as this was also the case with BodyVelocity & BodyGyro instead of LinearVelocity & AlignOrientation. I switched to the latter prior to this bug report to see if it’d make a difference, which it did not.
This issue has been fixed by recent improvements to the way forces are limited for the LinearVelocity constraint. Unfortunately, the issue will remain for BodyVelocity due to the way forces are specified for that class.
Please report back if there still seem to be any issues with the LinearVelocity constraint. Thanks.
