non-beta: When the controller is exactly upside down, rotation matrices have scaling
You don't have a section for bugs in non-Beta software, so I'm posting it here.
When the controller is held exactly upside down, pointing in any direction, it returns rotation matrices that are very invalid. The matrices include scaling in the Y and Z axes. It also returns slightly incorrect positions.
Roughly upside-down is fine, but exactly upside-down (roll=180 degrees) causes major problems.
You can verify this bug easily by running the SDK sample application and turning the controller upside-down and adjusting it slightly until the graphics jump out at you or grow tall.
If this is a hardware limitation that makes measuring at that orientation impossible, then you should just detect invalid data and return old data or failure instead. If it's not a hardware limitation then you should fix your code if you can find the bug, or at least detect the bug and return old data or failure.
Any progress on this bug? Any explanation?
Same problem here, going mad!!! how to fix it?