Implicit conversion from float -1.7127071823204432 to int loses precision