[boost] Re: Re: Proposals: p2m offsets, prebuild<>, squad_ptr<>(ofcourse)

From: Philippe A. Bouchard (philippeb_at_[hidden])
Date: 2002-09-03 13:41:25

Philippe A. Bouchard wrote:

[...]

Correction.  virtual ~T() doesn't mean the destructor is not trivial:

inline placed_ptr<T, simple_type>::destroy(T * p, __true_type const &)
{
        ...

        for (list<void *>::iterator i = g.begin(); i != g.end(); ++ i)
        {
                free(p - (sizeof(ptr_header) + ptr_header::s_padding));
        }
}

inline placed_ptr<T, complex_type>::destroy(T * p, __true_type const &)
{
        ...
                free(dynamic_cast<void *>(p) - ...);
}

inline placed_ptr<T, simple_type>::destroy(T * p, __false_type const &)
{
        p->~T();
        free(p - (sizeof(ptr_header) + ptr_header::s_padding));
}

inline placed_ptr<T, complex_type>::destroy(T * p, __false_type const &)
{
        ...
        free(dynamic_cast<void *>(p) - ...);
}

inline placed_ptr<T, simple_tree>::~placed_ptr()
{
        if (m_ptr && ! -- get_count())
        {
                destroy(m_ptr, __type_traits<T>::has_trivial_destructor());
        }
}

inline placed_ptr<T, complex_tree>::~placed_ptr()
{
        if (m_ptr && ! -- get_count())
        {
                destroy(m_ptr, _type_traits<T>::has_trivial_destructor());
        }
}

... and placement operator new (size_t, gc_type const &) could clean the 
garbage in the heap if malloc() returns 0 also.

It seems you had the same idea Larry...  I more reason not to not implement 
it.

-- 
Philippe A. Bouchard