Implicit conversion from float -30.89235127478753 to int loses precision