Implicit conversion from float -17.00787401574803 to int loses precision