Implicit conversion from float -4.176706827309232 to int loses precision