Re: [Boost-users] [thread] bug: thread::sleep can't 'Idle'

From: Ian McCulloch (ianmcc_at_[hidden])
Date: 2006-10-12 09:21:10

Roland Schwarz wrote:

> Johan Nilsson wrote:
>> Out of curiosity, who actually prefers using clock times for
>> thread::sleep()? IIRC the implementation still converts the UTC time into
>> a relative time before entering sleep, so it's only misleading to have an
>> absolute time as parameter to sleep.
> 
> You would need to ask the pthreads folks. ;-)
> This behaviour is inherited from there.
> 
> Roland

That is kinda misleading.  pthreads doesn't have a native sleep function. 
There were proposals for one, and some systems have a non-standard
extension, but the usual way is to just call nanosleep().

In boost::threads, almost all of the code paths do in fact convert the
absolute time into a relative duration.  The only path that doesn't is when
there is no nanosleep() and no pthread_delay_np() present; then the sleep
is implemented by a pthread_cond_timedwait() which uses an absolute time,
which has a resonable rationale for doing so.

The rationale for using absolute times in a condition variable wait don't
apply to sleep(), IMHO.  The pthreads sleep extension uses relative time. 
There are alternatives to pthread_cond_timedwait() that don't use absolute
time, for example select().  So a relative sleep is certainly
implementable.

Cheers,
Ian McCulloch