Implicit conversion from float -1.2328767123287747 to int loses precision