From: David Abrahams (dave_at_[hidden])
Date: 2003-10-23 08:52:00
"Maxim Egorushkin" <e-maxim_at_[hidden]> writes:
> Michael Grundberg wrote:
> MG> In other words, we do:
> MG> // Our resource int i;
> MG> // Give it to the library boost::shared_ptr<int> p1(&i,
> MG> null_deleter()); std::set<boost::shared_ptr<int> > s;
> MG> s.insert(p1);
> MG> // A bit later boost::shared_ptr<int> p2(&i, null_deleter());
> MG> s.erase(p2);
> MG> Which doesn't remove the pointer from s. This is causing us alot
> MG> of bugs, since the compilations succeed but the code is broken.
> Looks like you are misusing shared_ptr<>.
> The key point is that shared_ptr<> is an exclusive owner of a raw
> pointer. No more then one shared_ptr<> should be constructed using the
> same raw pointer (unless it is 0 pointer). Violating the rule leads to
> multiple releases on the raw pointer. Using null_deleter allows you to
> remain unpunished :).
I'm not sure that this should be classified as an abuse. It seems a
little unclear that this side effect of allowing custom deleters was
anticipated. There are other ways to induce the same problem which
don't involve null_deleter.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk