[boost] [serialization] Big gotcha with BOOST_CLASS_EXPORT

From: Robert Ramey (ramey_at_[hidden])
Date: 2004-01-21 12:42:40

Vladimir Prus wrote:

>I've just spend about an hour with a colleague looking for a very subtle 
>problem with serialization and BOOST_CLASS_EXPORT.

>What we were trying to do is to have some code which reads pointers to base 
>class and does something with it. On the saving side, we've got several 
>functions which store various derived classes. 

>Now the example:

>    class B { ... } ;
>    class D { ... } ;
>    BOOST_CLASS_EXPORT(D);

>    D* d = new D;
>    oa << d;

>Looks innocent, but it does not store the name of derived class that is 
>stored. So, after loading with

>    B* b;
>    is >> b;

>you get object of type B, not D. I attach the program which reproduces this
>problem.

>The gotcha is that if you write

>    B* d = new D;
>    oa << d;

>everything works. But it's nonintuitive, and the mistake is very easy to make 
>and very hard to notice! In my case, one function was intended for saving 
>specific derived type, so it naturally declared a pointer of that type, and 
>it took an hour to debug the problem.

Only an hour? 

>Is it possible to always save proper identification when saving a pointer to 
>polymorphic objects?

This looks like an error to me.  The intention is that it work as you expected.
I'll look into it.

Robert Ramey