Implicit conversion from float -67.03624733475482 to int loses precision