Implicit conversion from float -58.19999999999999 to int loses precision