Implicit conversion from float -39.61956521739131 to int loses precision