linuxthreads, POSIX 1003.1j and CLOCK_MONOTONIC

[H. J. Lu]

On Wed, Jul 10, 2002 at 07:16:17PM -0700, Brian Kuschak wrote:
> 
> The issue is that the existing 1003.1c implementation
> has race conditions with respect to system clock
> discontinuities.  The POSIX working group knows this,
> which is why they addressed it in 1003.1j.  
> 

That is the real problem for us. I use

  while ((timeout->tv_sec > 0 || timeout->tv_nsec > 0)
	 && __libc_nanosleep(timeout, timeout) != 0);

I can live with the library overhead.


H.J.