Implicit conversion from float -9.9384236453202 to int loses precision