Implicit conversion from float -32.11538461538461 to int loses precision