Re: m4/clock_time.m4 from r117596

[Paul Eggert]

Dmitry Antipov wrote:
> Are you sure about that?  On my (Fedora 20) system, 'man timer_settime' and
> 'man timerfd_settime' says nothing about resolution and rounding.

It's a POSIX requirement for timer_settime, and I expect timerfd_settime 
to be similar.  I'd be surprised if the Linux kernel did it any differently.

A bit of reflection should make it clear why this is so (and why 
clock_settime is different): an application must always allow for the 
act of computation to advance the clock a little bit, so if the system 
rounds timers up and clocks down it won't hurt apps any more than 
they're hurt already simply by running.

Here's the POSIX requirement:

"Time values that are between two consecutive non-negative integer 
multiples of the resolution of the specified timer shall be rounded up 
to the larger multiple of the resolution."

http://pubs.opengroup.org/onlinepubs/9699919799/functions/timer_settime.html