Re: Thread-safe singleton w/ Boost.Threads

From: williamkempf_at_[hidden]
Date: 2001-09-25 15:42:36

--- In boost_at_y..., "Peter Dimov" <pdimov_at_m...> wrote:
> From: <scleary_at_j...>
> > >
> > > static pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER;
> >
> > Boost.Threads is designed to be portable.  There is no way to do 
this
> under
> > Win32 to my knowledge :(
> 
> Looking at the pthreads-win32 source for inspiration, I can imagine
> something like:
> 
> struct static_mutex
> {
>     long once;
>     CRITICAL_SECTION * pcs;
> };
> 
> #define BOOST_STATIC_MUTEX_INITIALIZER { -1, 0 }
> 
> void lock(static_mutex & m)
> {
>     if(m.pcs == 0)
>     {
>         if(::InterlockedIncrement(&m.once) == 0) // first post!
>         {
>             CRITICAL_SECTION * tmp = new CRITICAL_SECTION;
>             ::InitializeCriticalSection(tmp);
>             m.pcs = tmp;
>         }
>         else while(m.pcs == 0)
>         {
>             ::Sleep(0);
>         }
>     }
> 
>     ::EnterCriticalSection(m.cs);
> }
> 
> Anyway, my question was about implementing a singleton with 
Boost.Threads.
> ;-)
> 
> IMO, anything that is easy to do under pthreads should be at least 
as easy,
> and less error-prone, under Boost.Threads, otherwise why bother? The
> singleton is an example that fails to meet that (perhaps 
idealistic) goal.

It's not an idealistic goal at all.  And my reading of discussions on 
this topic and on DCL indicate that there should not be an issue with 
using boost::call_once to do this, though I am disturbed that the 
POSIX standard appears to be ambiguous in this regard.

Bill Kempf