Re: [boost] Ending/canceling/suspending a thread (proposal)

From: Peter Dimov (pdimov_at_[hidden])
Date: 2001-08-23 10:13:57

From: "Scott McCaskill" <scott_at_[hidden]>
> I just remembered another problem that someone else pointed out when this
> was discussed a while back--using exceptions for thread cancellation would
> mean that you would have to make sure that you never called any
cancellation
> points (directly or indirectly) from a destructor.  That's a much bigger
> concern--we couldn't even have scoped_lock then, not and also have robust
> cancellation.

True; but robust cancelation requires stack unwinding, since in C++
destructors release resources. What are the defining characteristics of a
robust cancelation scheme that does stack unwinding, executing destructors
and catch() handlers, but is not equivalent to throwing an exception? [BTW
cancelation points would probably not throw when std::uncaught_exception()
returns false.]

--
Peter Dimov
Multi Media Ltd.