How to make minutes and seconds timer in C

Question

I'm struggling to make a timer in c that counts minutes and seconds. I'm trying to test it by printing the time to the console but it doesn't seem to show anything. Does anything look wrong in my code?

#include <stdio.h>
#include <stdlib.h>
#include <windows.h>

#define TRUE 1

int main( void )
{

    int min = 0;
    int sec = 0;

    while (TRUE)
    {
        sec++;
        Sleep(1000);
        printf("%2d:%2d", min, sec);
        if (sec == 59)
        {
            min++;
            sec = 0;
        }
    }
    return 0;
}

Answer 1

For performance reason, printf is buffered. That is, it won't display until the buffer is full. First thing to try is to add a new-line character to the end:

    printf("%2d:%2d\n", min, sec);

If that doesn't work, you can force the output buffer to flush by calling fflush(stdout);

Answer 2

I would just check the system time rather than keeping track of seconds/minutes. This is because, your Sleep may not be exactly 1000ms, so over time your counter will not be accurate.

Since you're using Windows, here's a slightly modified version of your code:

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <windows.h>

int main()
{
    for (;;)
    {
        time_t now = time(NULL);
        struct tm* ptm = localtime(&now);
        printf("%02d:%02d\n", ptm->tm_min, ptm->tm_sec);
        Sleep(1000);
    }

    return 0;
}

I hope that helps.

Answer 3

As you ask, there are several things wrong in your code, I'll tell you as I read it. See below.

#include <stdio.h>
#include <stdlib.h>
#include <windows.h>

#define TRUE 1

int main( void )
{

    int min = 0;
    int sec = 0;

    while (TRUE)
    {
        sec++;
        Sleep(1000);

1. There's a problem with Sleep(). The thing is that you ask the kernel to pause your program for 1000 milliseconds, but that means that the kernel will awake your program 1 second after your call it to pause. This doesn't take into account the fact that the kernel will put in the schedule queue your process and it will, in general never take the cpu immediately, but after some delay. Then, even if you get the cpu immediately, your code will need some time to execute, making the total loop longer than 1000 ms. And your clock will be slow. It is better to get the system time and show it on the screen, or to take a timestamp when you start... and then show the difference in time from the start time to the timestamp you get at each display. The system time is maintained by an interrupt, that happens at regular intervals (by means of a precise clock oscillator) so you'll get a good clock time that way, instead of your slow clock (how slow it is will depend on things like how many other processes you are running on the system)

            printf("%2d:%2d", min, sec);

2. This has already been stated in other answers, but let me explain how it works so you can understand how buffering works. A buffer is a large block of memory (normally it is 512 bytes) that is filled by printf() so it only calls write() when it has filled a complete buffer of data. This allows stdio to save system calls to do the actual writing and so, be more efficient when transferring large amounts of data.

   On interactive applications, this is not applicable, as no output would be done if you don't force the buffers to fflush() before any input is done, so when the output is a tty device (something that stdio can know from the file descriptor associated to standard output) then it switches to line mode buffering, and that means that printf() will flush out the buffer when: 1) it is filled up, or 2) when a newline \n character is found in the output. So, one way to solve your problem is to put a \n at the end of the string, or to call fflush(stdout); after calling printf().

            if (sec == 59)

3. as you increment your seconds before comparing, you have to check against 60 and not 59, as you compare with 60 to convert it to 0 after you have already incremented the seconds.

            {
                min++;

4. You should have to do the same with the minutes, when minutes get to 60. As you have not included this code, I assume you are not considering an hours chrono.

                sec = 0;
            }
        }
        return 0;
    }

---

A complete solution for what I mean can be:

#include <stdlib.h>
#include <stdio.h>
#include <time.h>
#include <windows.h>

int main()
{
    long start_timestamp;
    long display_timestamp;

    start_timestamp = time(NULL);

    for(;;) { /* this is the same as while(TRUE) */
            display_timestamp = time(NULL);
            int elapsed = display_timestamp - start_timestamp;
            int min = elapsed / 60;
            int sec = elapsed % 60;
            /* the \r in next printf will make to print a single carry
             * return in the same line, so the time updates on top of
             * itself */
            printf("\r%02d:%02d", min, sec); fflush(stdout);
            Sleep(100); /* as you print the time elapsed, you don't
                         * mind if you shorten the time between 
                         * displays. */
    }
}

Answer 4

Does anything look wrong in my code?

Several things look wrong. The first is stdout line buffering - see eduffy's answer for that.

The second problem is that you're doing sec++; sleep(1000);, which means that sec will be incremented once every 1000 seconds or more.

The third problem is that if(sec == 59; is wrong and needs to be if(sec == 60). Otherwise you'll have 59 seconds per minute.

The fourth problem is that sleep(1) will sleep for at least 1 second, but may sleep for 2 seconds, or 10 seconds, or 1234 seconds. To guard against this you want something more like this:

expiry = now() + delay;
while(true) {
    sleep(expiry - now() ):
    expiry += delay;
 }

The basic idea being that if one sleep takes too long then the next sleep will sleep less; and it'll end up being "correct on average".

The last problem is that sleep() doesn't really have enough precision. For 1 second delays you want to be able to sleep for fractions of a second (e.g. like maybe 9/10ths of a second). Depending on which compiler for which OS, there's probably something better you can use (e.g. maybe nanosleep()). Sadly there may be nothing that's actually good (e.g. some sort of "nanosleep_until(expiry_time)" that prevents jitter caused by IRQs and/or task switches that occur after you determine now but before you call something like "nanosleep()").