Implicit conversion from float -7.490347490347489 to int loses precision