tr1::shared_ptr and v3's atomic ops

[Jonathan Wakely <cow at compsoc dot man dot ac dot uk>]

I have a patch to fix a MT race condition in tr1::shared_ptr.
I haven't tested it since Benjamin re-organised the headers, I'll do so
whehn I get home tonight, and will post it here later.

In the meantime...

Since I incorporated shared_ptr into v3 Peter Dimov has modified the
Boost version so that (on Windows) it doesn't use any locks. I had
already done something similar for the version I added to v3, except
that I couldn't make shared_count::add_ref_lock() lock-free, since
v3's atomic compare_and_swap() was removed some years ago:
http://gcc.gnu.org/ml/gcc-patches/2001-10/msg00215.html

I have two questions about this

1) could compare_and_swap be resurrected, or is the porting headache too
much bother?

2) Do the v3 atomic ops give any guarantees about memory synchronisation
on systems with relaxed memory visibility rules ?  Peter Dimov has asked
on the Boost list for help regarding v3's atomic ops, as he wants to
have a lock-free shared_ptr on unix as well as on Windows:
http://lists.boost.org/MailArchives/boost/msg81481.php

jon

-- 
"The following story is true.  By which I mean false.  It's all lies.
 But they're entertaining lies.  And in the end, isn't that the real truth.
 The answer, is no."
	- Leonard Nimoy, in "The Simpsons"