Implicit conversion from float -6.84317718940936 to int loses precision