Implicit conversion from float -5.739910313901348 to int loses precision