FE_UNDERFLOW gives a fair amount of hits with LMMS but is of lower
importance than the other tests and slows down debugging considerably.
Commenting out for the time being.
TimePos::quantize works for negative values, but ends
up snapping the TCO to the opposite direction. This is because the
snapping happens in the direction of the origin, which is left for
positive values and right for negative values.
That wasn't accounted for in the pasteSelection method
and we ended up with wrong positions when pasting before the
TCO(s) we copied.
This PR fixes the issue by ensuring that we snap in the same direction when halfway through an interval, regardless of negative or positive offset.
* Fixes a calculation on TimePos::quantize
Since we are working with integers, using "offset /
(interval/2)" would be problematic if interval was odd. We instead
multiply both sides by two and use "(2 * offset) / interval" to obtain
the result for snapUp.