Re: [boost] no semaphores in boost::thread

From: Stefan Seefeld (seefeld_at_[hidden])
Date: 2003-06-04 10:00:32

Robin Hu wrote:

 >     Semaphore can always be stimulated by using mutex + cond_var. and
 >     the only place that semaphore is a must is in signle-handler, but
 >     this is not within the area of threading programming. ;-)

right, semaphores can be emulated by other primitives. But does
that mean they *have to* be implemented that way ? I thought the point
of not including semaphores in boost::threads was that they were
error-prone or 'broken' in some other way, so I wanted to understand
why.

Actually, I'v already seen threading libraries where the windows version
implemented condition variables in terms of mutexes and semaphores, not
the other way around...

Regards,
                Stefan