Boost logo

Boost Users :

Subject: Re: [Boost-users] [Serialization] Runtime error in load_pointer
From: Robert Ramey (ramey_at_[hidden])
Date: 2010-01-19 12:54:50


elizabeta petreska wrote:
> Again thank you all who helped me in this .
>
>>> This is likely a secondary effect from some other error.
> It was secondary error.
>
>>> A complete shot in the dark but it is possible it might be the
>>> problem.
>
> This was the problem, so it wasn't shot in the dark at all.
> The problem was that my classes have the following class members :
> shared_ptr<const T> , vector<const T>, vector<shared_ptr<const T>>
> and many others that have const inside them. I can not remove const
> from these, altghough I wish :)
> I was casting const away (http://tinyurl.com/yae5stv ) for all this
> members only at the point of loading, for example in many cases I was
> saving vector<const T> and loading vector<T>. Now I added const cast
> at the point of saving too, so saving and loading is symmetric. It
> seems it works Ok now, and I hope this is the right way to solve the
> problem.

I'm curious about this.

a) I can't understand why vector<const T> is even valid. My understanding
is that the value of type of a container must be Assignable and that
a const T is by definition not assignable. So I would expect this to
show up as an issue long before serialization is involved.

b) I can't see how one gets the asymetry in the first place. E.G.

class X {
    vector<const T> m_v;
    template<class Archive>
    void save(Archive & ar, const unsigned int version) const {
        ar << m_v;
    }
    template<class Archive>
    void load(Archive & ar, const unsigned int version){
        ar >> m_v;
    }
    ...
};

Hmmm I suppose that one might use a cast on a load but not on the save.
This is a case which in fact I've never considered. I'll think about this.

Meanwhile, any C++ STL gurus can feel free to comment on a) above.

Robert Ramey


Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net