Implicit conversion from float -49.02208201892745 to int loses precision