Subj : Re: Boost.Threads on its way to C++0x
To   : comp.programming.threads
From : gottlobfrege
Date : Tue Apr 26 2005 08:00 am


Peter Dimov wrote:
> gottlobfrege@gmail.com wrote:
> > - a static_local template:
> >
> > int func()
> > {
> >   // this thread-safely inits foo only once
> >   static static_local<Foo> foo(foo_param1, foo_param2,...);
>
> How do you solve the race condition WRT the hidden compiler-generated

> boolean flag?

There are lots of ways, but they all boil down to the same thing: ie
you ignore the compiler's flag and use your own.  And use call_once to
set your flag.

More likely, you use call_once to init a CriticalSection, then use the
CriticalSection to init (in-place new?) your internal Foo object.

Or use a 'StaticLocalCriticalSection' object (which does the call_once
stuff internally, or does the call_once-like magic itself, and then
write call_once using the StaticLocalCriticalSection...).

Or use a Once object and a Once::Sentry object, which is created using
a StaticLocalCriticalSection, or vice versa, etc.

Even though all the ways are 'equivalent', some take longer than
others, and, if you are careful, you can avoid creating any
mutexes/semaphores/criticalsections in the non-contention case.  It
depends what you know about the context - ie is the call_once init_func
going to throw, or just call InitCriticalSection() and you're
comfortable assuming that works, etc...

.