Re: first sight

From: bill_kempf (williamkempf_at_[hidden])
Date: 2002-01-14 10:12:24

--- In boost_at_y..., Jens Maurer <Jens.Maurer_at_g...> wrote:
> Lee Brown wrote:
> > I do have one question:  why doesn't Boost.threads seem to support
> > thread cancelation?  It doesn't seem to be mentioned in the 
documentation.
> 
> Thread cancelation is difficult to do correctly in a C++ program,
> because you want to run all the destructors of stack objects of
> the thread being canceled.
> Bill Kempf might want to elaborate.

It's not so much that it's difficult to implement, as it's difficult 
to implement portably.  There's also the issue of carefully and 
correctly specifying cancellation behavior.  As Jens points out you 
need to insure that the stack is properly unwound, and this means an 
exception or an exception like concept.  The best choice seems to be 
to use an actual exception, but then several questions arise.  Should 
the user be able to catch the exception?  If they do, should they be 
able to "eat" the exception (i.e. catch but not re-throw)?  How can 
you (easily) deal with the possiblity of an exception thrown during 
stack unwinding of another exception?  Should asynchronous 
cancellation be allowed (this means that ALL functions can throw an 
exception, which is a huge burden on the user, the compiler and the 
runtime, and means some standards conforming code today would no 
longer be conforming if this were part of the standard)?

Cancellation is being evaluated, and will eventually be added to 
Boost.Threads.  Because of the issues above, however, I can't predict 
when this will be.

Bill Kempf