Subj : Re: pthread function behavior in C++ program
To   : comp.programming.threads
From : Peter Dimov
Date : Sun Jun 12 2005 03:25 pm

Torsten Robitzki wrote:
>> It's better because it's only one step from the "right thing". They
>> only need to re-cancel (not rethrow!) instead of terminating. Not
>> running destructors and exception handlers will "only" leak and
>> deadlock, but such an implementation would be much more difficult to
>> fix.
>
> What do you have in mind when you distinct between "re-cancel" and
> "rethrow"?

If you have

void f() throw()
{
    try { cancellation_point(); } catch(...) {}
}

an automatic rethrow at the end of the catch clause will lead to process 
termination anyway because of the throw() specification.

> If I have to choose between an implementation that
> terminates my application every once and then or let my application
> leak like a sieve I would choose neither of them ;-) but try to find
> a solution without cancellation.

Yes, practically speaking, you are right; but from the perspective of 
someone who is interested to see usable cancellation in (standard) C++ at 
some point in the future, one of these is still better than the other. :-)

.