`btKinematicCharacterController::getLinearVelocity()` returns displacement, not velocity

[Sweetzonzi]

Description

getLinearVelocity() of PhysicsCharacter directly returns m_walkDirection, but m_walkDirection is used as per-step displacement in stepForwardAndStrafe (m_targetPosition = m_currentPosition + walkMove), without being multiplied by dt.

// getLinearVelocity() — returns m_walkDirection as-is
btVector3 btKinematicCharacterController::getLinearVelocity() const {
    return m_walkDirection + (m_verticalVelocity * m_up);
}

// stepForwardAndStrafe — m_walkDirection used as displacement
m_targetPosition = m_currentPosition + walkMove;

// playerStep — passed directly, no dt scaling
stepForwardAndStrafe(collisionWorld, m_walkDirection);

By contrast, the setVelocityForTimeInterval path (m_useWalkDirection = false) correctly multiplies by dt:

btVector3 move = m_walkDirection * dtMoving;
stepForwardAndStrafe(collisionWorld, move);

Result: every other CollisionObject's getLinearVelocity() returns m/s, but KCC returns m/step — off by a factor of dt.

Suggestion

If changing the native behavior is not feasible, the Javadoc should at least note that callers must divide by dt to recover actual velocity. The current Javadoc makes no mention of this discrepancy, making it a subtle trap for users.

[stephengold]

Thank you for bringing this issue to my attention.

Changing behavior at this point would probably cause more trouble than it solves, but I'm definitely open to correcting the documentation, especially if the name is so misleading. I'll want to study the code myself, of course.