Implicit conversion from float -1.7541229385307417 to int loses precision