sleep() function does not work and print is executed instantaneously

Question

static void timeDelay(int no_of_seconds)
{
#ifdef _WIN32
    Sleep(1000 * no_of_seconds);
#else
    sleep(no_of_seconds);
#endif
}

void somefunction(){
printf("\t\t Load ... \n\t\t"); 
fflush(stdout); 
for (int i = 1; i <= 60; i++)
        {
            fflush(stdout);
            timeDelay(1);
            if (i == 31)
                printf("\n\t\t");
            printf("*****");
        }
        
  }

I have included the header files too:

#ifdef _WIN32
    #include <windows.h>
#else
    #include <unistd.h>

The stars are printed instantaneously. I added fflush(stdout) after seeing the answers here. I also commented out the if (i==31) {} portion to check if that's causing the problem but it isn't. So what is wrong with my code?

I mean that I have tried your program and it works as expected. There is a delay and there is no instantaneous output. — n. m. could be an AI, Jun 28 '20 at 14:18
That's what you said, but which *system* are you running. The Linux version of the code, or the Windows version of the code? — Weather Vane, Jun 28 '20 at 14:19
@Archer Is it possible that you are using SSH or any other kind of remote access to the box you are compiling and running the program in, and the terminal application or the server itself is performing some kind of buffering. Can you provide more insights about which OS, which environment, etc you are running your code in? — mcleod_ideafix, Jun 28 '20 at 14:24
@mcleod_ideafix Windows 10 -> Linux shell provided by Ubuntu WSL. — Archer, Jun 28 '20 at 14:25
I am trying to ask whether the code that prints stars immediately is calling `Sleep()` or whether it is calling `sleep()`. — Weather Vane, Jun 28 '20 at 14:25
It seems to be a problem with nanosleep. man page for sleep() says that it is based on nanosleep. I've tried using it directly and it doesn't execute the delay, but returns EINVAL. — mcleod_ideafix, Jun 28 '20 at 14:42
The code presented in the question works for me on WSL just as seems to be intended: there is a delay before printing each group of stars. I am inclined to believe, therefore, that either the program you are actually running was not built from corresponding source, or that the problem arises from something in the parts that have not been presented. — John Bollinger, Jun 28 '20 at 15:25

mcleod_ideafix · Answer 1 · 2020-06-28T15:25:28.737

According to the sleep() manpage...

NOTES On Linux, sleep() is implemented via nanosleep(2). See the nanosleep(2) man page for a discussion of the clock used.

So I have rewritten your program to use nanosleep. As you are working with WSL, I've dropped any reference to Win32, only Linux. Well, the thing is that this program exits nanosleep() prematurely, with an "Invalid argument" error. I cannot see why is that.

#include <stdio.h>
#include <unistd.h>
#include <time.h>
#include <errno.h>

void timeDelay (time_t no_of_seconds)
{
  struct timespec req, rem;
  int res;

  req.tv_sec = no_of_seconds;
  req.tv_nsec = 0;

  do
  {
    rem.tv_sec = 0;
    rem.tv_nsec = 0;
    res = nanosleep (&req, &rem);
    req = rem;
  }
  while (res == EINTR);
  if (res)
    perror("nanosleep");
}

void somefunction()
{
  printf("\t\t Load ... \n\t\t");
  fflush(stdout);
  for (int i = 1; i <= 60; i++)
  {
    fflush(stdout);
    timeDelay(1);
    putchar('*');
  }
}

int main()
{
  somefunction();
  return 0;
}

I've also tried with NULL instead of rem, and reissuing nanosleep() with the original time instead of the remaining time, and putting the test and the perror() within the loop to print all possible errors from nanosleep. No matter what I do, I always receive an EINVAL from the first call to nanosleep()

So, there seems to be a real problema with nanosleep() on WSL. See https://github.com/microsoft/WSL/issues/4898 . It mentions a problem for WSL being unable to read the realtime clock from a certain version of glibc. I tried this in my WSL terminal:

$ sleep 1
sleep: cannot read realtime clock: Invalid argument

There is a mention of a workaround here: https://github.com/microsoft/WSL/issues/4898#issuecomment-612622828

I've tried with another approach: using clock_nanosleep() so I can choose another source for the clock: in the above program, just change the call to nanosleep() with this other to clock_nanosleep()

clock_nanosleep (CLOCK_MONOTONIC, 0, &req, &rem);

There is no problem with CLOCK_MONOTONIC, and now the program works!

I agree, this does not solve the issue. Nor does it really even address the question. How `sleep()` is implemented is irrelevant. That it behaves as it is documented to do (or that it didn't, if that were the case) is the point. And it *does* behave as documented for me on WSL. — John Bollinger, Jun 28 '20 at 15:21
It IS relevant, as the underlying problem is that sleep() uses nanosleep(), and nanosleep() uses CLOCK_REALTIME which cannot be read in WSL. I've updated my answer anyway. — mcleod_ideafix, Jun 28 '20 at 15:26
`sleep()` works as documented on WSL (for me). Perhaps it is buggy on some WSL providers, but that's a rather extraordinary claim. One of the key points of WSL is that it provides a fully functional Linux environment. Furthermore, if it does happen to be buggy on some WSL providers, that does not necessarily mean that the problem is with `nanosleep()`, though that is plausible. — John Bollinger, Jun 28 '20 at 15:32
For some users it works fine. The issue is documented here also: https://discourse.ubuntu.com/t/ubuntu-20-04-and-wsl-1/15291 — mcleod_ideafix, Jun 28 '20 at 15:38
Great. *That* would be the basis for an actual answer. "`sleep()` does not work correctly on some WSL providers, such as [...]". *Why* it doesn't work is a detail of little significance to the OP's question. — John Bollinger, Jun 28 '20 at 16:06

sleep() function does not work and print is executed instantaneously

1 Answers1