[boost] Re: Moving Boost.Thread forward

From: Michael Glassford (glassfordm_at_[hidden])
Date: 2003-12-16 11:32:26

Alexander Terekhov wrote:

> Peter Dimov wrote:
> 
>>Michael Glassford wrote:
>>
>>>One other potential problem that needs to be addressed in the same
>>>section of code: if the "called once" function throws an exception,
>>>should the flag indicating that it has been called be set or not? If the
>>>flag is set, the function will not be called again even though it
>>>failed; on the other hand, if the flag is is not set, the function may
>>>continue to be called and continue to throw an exception. What does
>>>the pthread implementation do in this case?
>>
>>Pthreads don't need to deal with exceptions, 
>

Actually, I meant the pthread implementation of Boost.Thread, which does 
need to deal with exceptions, and looks like this:

    void call_once(void (*func)(), once_flag& flag)
    {
       pthread_once(&once, &key_init);
       pthread_setspecific(key, &func);
       pthread_once(&flag, do_once);
    }

> 
> Except that 
> 
> <quote>
> 
> The pthread_once() function is not a cancellation point. However, 
> if init_routine is a cancellation point and is canceled, the effect 
> on once_control shall be as if pthread_once() was never called.
> 
> </quote>

I did run across the above quote after reading Peter's reply above. In 
C++, does "if init_routine is a cancellation point and is canceled" 
roughly translate into "if init_routine throws an exception"? I confess 
to almost complete ignorance of pthreads (what little I know I've picked 
up by following conversations here and occasionally in 
comp.programming.threads), so in addition to a reply to my question, I'd 
appreciate pointers to a good tutorial and/or reference.

Thanks,

Mike