Implicit conversion from float -5.257234726688111 to int loses precision