[boost] Re: Boost.Thread

From: Alexander Terekhov (terekhov_at_[hidden])
Date: 2002-09-16 12:14:47

"Victor A. Wagner, Jr." wrote:
> 
> At Monday 2002/09/16 05:29, you wrote:
> 
> >"Hillel Y. Sims" wrote:
> > >
> > > "Eric Woodruff" <Eric.Woodruff_at_[hidden]> wrote in message
> > > news:014f01c25a8f$605ebd80$1800000a_at_soy...
> > > > When return values enter the picture, join needs to become the return
> > > value
> > > > accessor, allowed to be called more than once
> > >
> > > If this is the case, how would the thread library know when to finally
> > clean
> > > up the finished thread's resources?
> >
> >When the last reference goes away (i.e. managing thread objects via
> >smart pointers; I'd even choose *AUTO*_ptr like move semantics by
> >default with some "shared ownership" adapter for {current} standard
> >containers and algorithms). http://www.terekhov.de/mythread.c
> >
> >---
> >int my_thread_join( my_thread_t mt,void** value_ptr );
> >
> >int my_thread_tryjoin( my_thread_t mt,void** value_ptr );
> >
> >int my_thread_timedjoin( my_thread_t mt,
> >                          struct timespec* timeout,
> >                          void** value_ptr );
> >---
> 
> So, we're going to intentionally put a "timing window" problem into user
> code.... multiple join() will work just fine _provided_ that all the 
> relevant join() requests actually occur _before_ the thread terminates.  
> Otherwise, all bets are off.

What makes you think so?

> I don't like it.

``Me too.'' ;-)

regards,
alexander.