In what situation the output could go wrong like this?

Question

I am trying to do the problem 200B on codeforces. I tested my code, and the output was all right. But when I uploaded it to the online judge system, I failed on the very first test case. It said my output was -0.000000000000, instead of 66.666666666667.
But I have compiled and run on Visual Studio C++ 2010, MacOS clang 13.0.0, and Linux GCC 6.3.0, the outputs were all the same as mine, 66.666666666667. I am very curious and want to figure out in what situation the output could be -0.000000000000.
On my computer,
Input:
3
50 50 100
Output:
66.666666666667
On the online judge system,
Input:
3
50 50 100
Participant's output
-0.000000000000
Jury's answer
66.666666666667
Checker comment
wrong answer 1st numbers differ - expected: '66.66667', found: '-0.00000', error = '1.00000'

#include <stdio.h>

int main(void)
{
    int n;
    double sumOrange = 0;
    double sumDrink = 0;
    scanf ("%d", &n);
    
    while (n-- > 0) {
        int m;
        scanf("%d", &m);

        sumOrange += m / 100.0;
        sumDrink++;
    }

    printf("%.12lf\n", (sumOrange / sumDrink) * 100.0);
    return 0;
}

I just don't understand why my output could be -0.000000000000. Please help, thanks.

Update: Tested on different versions of GCC (4.9, 5.1, 6.3), the wrong output does not appear. Guess the cause might lie in the specific implementation of printf.

By the way use this format string %.12f instead of this %.12lf — Vlad from Moscow, Dec 23 '21 at 06:59
Leon, Aside: could remove the `/ 100.0` and `* 100.0`. They only inject minute errors in the calculations. — chux - Reinstate Monica, Dec 23 '21 at 08:33
@chux-ReinstateMonica, yeah, you are right. I was just confused about the output. So I tried every possible way to find the cause, but failed. — Wu Xiliang, Dec 23 '21 at 09:12
@VladfromMoscow, Thanks for you comment. I have just updated my code and passed the tests, but still I want to know when, in what situation, or by what compiler, the compiled executable could produce this error. I just have tried Windows compiler VS 2010, Linux GCC and MacOS XCode (clang), the output is all the same, no problem. — Wu Xiliang, Dec 23 '21 at 09:16
For `sumDrink++'` See: [Is floating point math broken?](http://stackoverflow.com/questions/588004/is-floating-point-math-broken) and [Why Are Floating Point Numbers Inaccurate?](https://stackoverflow.com/questions/21895756/why-are-floating-point-numbers-inaccurate) and [Floating point comparison `a != 0.7`](https://stackoverflow.com/questions/6883306/floating-point-comparison-a-0-7). As an example, with `float f = 1.1;` is actually `1.100000023842`. After `f++;`, you have `2.099999904633` -- do be aware of the floating-point math aspect of post-increment. — David C. Rankin, Dec 24 '21 at 03:59

score 1 · Answer 1 · answered Dec 23 '21 at 07:28

1

The problem is because printf function in GNU gcc C11 does not support %.12lf format. It should be changed to %.12f For more information, you can read the article below:

Correct format specifier for double in printf

answered Dec 23 '21 at 07:28

konata39

196
6

Thanks. I feel so happy to know the cause of the problem, but what I really want is to get the wrong output if possible. – Wu Xiliang Dec 23 '21 at 09:18
Maybe it is depends on GCC version. The GCC version in codeforces is 5.1, and your VS C++ version is 13.0(clang), so it can only represent wrong answer in lower GCC. Maybe the newer version of clang is fix `%lf` problem, and it will auto convert %`lf` to `%f`. Therefore, you cannot represent the error output in VS C++. – konata39 Dec 23 '21 at 09:29
1

[`%.12lf` works fine in GCC 5.1](https://godbolt.org/z/9WjT1EG5x). It is actually the GNU C Library which implements `printf`, not GCC. But the problem is in the online judge’s C implementation, the version of which is not stated. – Eric Postpischil Dec 23 '21 at 11:39
@EricPostpischil I tried GCC 4.9 on Debian jessie, the output is also okay. I agree with you. The cause should lie in the specific implementation of `printf`. Really happy that I have learned something I didn't understand. – Wu Xiliang Dec 24 '21 at 00:02

In what situation the output could go wrong like this?

1 Answers1