Round bug with series

[Former user]

Originally reported on Google Code with ID 244 ``` What steps will reproduce the problem? 1. iterate sum : for i = 0 to infinite ans=ans+(1/2^i) 2. Iterate until 67 3. Then iterate one more time : you've got the answer "2" which is wrong...

What is the expected output? What do you see instead? Excepted output is a number infinitively close to 2 but never equal

What version of the product are you using? On what operating system? 0.10.1 on kubuntu (with KDE 4.2)

Please provide any additional information below. No I'm not evil, but I have to say sometimes I'm doing strange thinks, that's right :) ```

Reported by `keltroth` on 2009-01-31 18:17:53

[Former user]

``` There's nothing wrong with the result. The sum is mathematically equal to 2-2^(-68), and that's exactly what you get. To 50 decimal places: 1.99999999999999999999661186821098279864372670999728 Subtracting the summed up result from the theoretical summation value yields 0, in perfect accordance with theory and anything you can expect from SpeedCrunch. Note: SpeedCrunch uses internally a 79 digits representation of numbers which will be rounded for output to the required number of digits. So, assumed the deviation from 2 is less than 10^-50, the displayed result is 2 in all display modes. The full internal 79-digit value continues to be used in further calculations. ```

Reported by `wolf.lammen` on 2009-01-31 21:01:22 - Status changed: `Invalid`

[Former user]

``` It's just kind a point of vue... Having 2 as an answer is mathematicaly wrong... Even if speedcrunch knows it 1.9999999999999999999999999999999999999999999999 and something... ```

Reported by `keltroth` on 2009-02-01 01:01:49