notes:

* Thanks Tomek for pointing out the iteration problem and possible efficient solutions.

* Thanks Tomek for the feedback.

* We’ve addressed negative numbers, numbers in (0, 1), overflow problems, absolute and relative error problems.
* Instead of |(a - b) / b| < eps we can use |(a - b) / max(a, b)| < eps. Since a, b >=0 in our problem, we won’t get the nasty cases I was talking about previously, except for the c = 0 case.
* Some tests for your own code -1, 0, 1, 2, 8, 0.001, 1e30, 1e60