奇怪的编译器行为

Question

I just hit a strange behavior when compiling a small C++ program with g++ (4.6.3). Compare the two floats celsius and test:

#include <iostream>
using namespace std;

int main () {
    float fahrenheit = 0;
    float celsius = 0;
    float test = 0;

    cout << "Temperature in °F: " << endl;
    cin >> fahrenheit;

    celsius = 5 / 9 * (fahrenheit - 32);
    test =  5 * (fahrenheit - 32) / 9;

    cout << "\nWrong:\t" << fahrenheit << " °F = " << celsius << " °C" << endl;
    cout << "Correct:" << fahrenheit << " °F = " << test << " °C" << endl;
}

I compiled this with just "g++ test.cpp -o test".

This is the output of the program:

$ ./test
Temperature in °F: 
1000

Wrong:  1000 °F = 0 °C
Correct:1000 °F = 537.778 °C

Can someone explain to me why g++ does (reproducibly) compute celsius to 0, whereas test contains the correct solution? Does it have anything to do with some optimization, although I didn't set any args for that? Or is this really a bug in some way?

Answer 1

5 / 9 = 0

0 * anything = 0

When you divide to integers, you get an integer back. You can just say 5.0 / 9 and it should be a lot better.

When you're writing code and don't get what you expect, your first inclination should be "what dod I do wrong?" not "what did the compiler do wrong?"

It'll help you look in the right place most of the time.