RE: [boost] Re: smart_ptrs and auto_ptr-like move semantics (was'boost::something')

From: Yitzhak Sapir (yitzhaks_at_[hidden])
Date: 2002-06-13 11:02:01

> -----Original Message-----
> From: Peter Dimov [mailto:pdimov_at_[hidden]]
> From: "Alexander Terekhov" <alexander.terekhov_at_[hidden]>
> > > > explicit scoped_ptr(std::auto_ptr<T> &p): ptr(p.release()) {}
> > > >
> > > > and it works as expected, taking away ownership from 
> auto_ptr and
> giving
> > > > it to scoped_ptr.
> > > >
> > > > Is this something my compiler (MSVC6SP5) allows as an 
> extenstion?
> > >
> > > Yes, this is a Microsoft extension. VC 7 doesn't even warn. Odd.
> >
> > a) why not simply: explicit scoped_ptr(std::auto_ptr<T> p):
> ptr(p.release()) {} <?>
> 
> Yes, this would work. :-)

Is it possible to have such a thing?  I would want the ability to create something (such as a thread).  In the client I would have a scoped_ptr around the thread class so that it joins() and exits as necessary.  However in the creation function I'd want to return auto_ptr to maintain the pointer during the passing from the creation function to the scoped_ptr.  Using a scoped_ptr/auto_ptr is more specific of what I want to do than just using shared_ptr.