From: Talbot, George (Gtalbot_at_[hidden])
Date: 2006-11-10 10:10:09
> Talbot, George wrote:
>
> > I thought that the current implementation of shared_ptr is lock-free
> > and multithreaded in the sense that I can pass shared_ptr between
> > threads, and the reference count is updated correctly, etc. (I'm
> > using BOOST 1.33.) Am I incorrect in thinking this?
Peter Dimov wrote:
> This is not the same as atomic. An atomic type allows you to modify a
> variable from thread A while, at the same time, thread B is also
reading
> or
> writing it. shared_ptr doesn't offer this level of thread-safety. As
an
> example, consider:
>
> shared_ptr<X> px;
>
> // thread A
>
> shared_ptr<X> px2( px );
>
> // thread B
>
> px = px3;
>
> (you can also have thread B doing compare and swap, it doesn't
matter.)
>
> Imagine that thread A reads the two pointers from px and is preempted
> right
> after that. Thread B overwrites the contents of px, and in doing so,
> causes
> its reference count to drop to zero; the object and the control block
are
> destroyed. Thread A wakes up and attempts to increment the now non-
> existent
> reference count. Bad things happen.
>
> But maybe I'm misunderstanding the intent of your compare and swap
> primitive.
I apologize if this is rapidly becoming a refresher course for me in
multithreaded programming. ;^)
Forgetting about compare-and-swap for a second, what you're saying is
that the _initialization_ of px2 from px in thread A above is not
atomic? Is this sequence atomic?
// Before the treads start
shared_ptr<X> px(blah);
// thread A
shared_ptr<X> px2;
px2 = px;
// thread B
px = px3
I thought that constructors, destructors and the assignments were atomic
from the documentation, but it doesn't mean I'm reading it right. I'm
reading http://www.boost.org/libs/smart_ptr/compatibility.htm. Under
"Features That Improve Robustness" it says "The manipulation of use
counts is now thread safe on Windows, Linux and platforms that support
pthreads. See <boost/detail/atomic_count.hpp> file for details."
I really don't understand, based upon that statement and the statements
made in the comments of atomic_count.hpp of exactly what operations are
thread safe and what aren't when using shared_ptr.
I'm sorry if I'm being too dense here. I think a section in the
documentation saying exactly what is guaranteed and what isn't would
help a lot.
-- George T. Talbot gtalbot_at_[hidden]