What mathematics probably hate most about floats is that rounding breaks commutativity.

Actually, the maths people in my circles tend to complain more about how the normal floating-point equality test doesn't define an equivalence relation, but really any of the violations of field properties can attract complaints.