Implicit conversion from float -105.54317548746516 to int loses precision