From: Andy Glew (glew_at_[hidden])
Date: 2000-01-07 10:43:55
Now that Aleksey et al have shown me how to create pre and post
methods using temporary values, I think that this shows a possible
resolution to (a) Boris Fomitchev's suggestion that shared_ptr<>
have a template parameter indicating synchronization,
and (b) Beman Dawes' reluctance to allow yet another template
parameter.
Boris should be made happy by mutexed< shared_ptr<T> >,
with implicit locking to make all accesses atomic.
If the default implementation of mutexed<T> is too slow
for shared_ptr<T>, then it can be specialized.
Beman's objection that the synchronized and unsynchronized
versions are different types is still valid --- shared_ptr<T>
and mutexed< shared_ptr<T> > *are* different types.
The usual issues about conversions apply.
--- Andy "Krazy" Glew, glew_at_[hidden], UW Madison and Intel. DISCLAIMER: private posting, not representative of university or Intel. Please respond by email in addition to replying to newsgroup.