Serialization with versioning not working with (pure virtual) base classes

[anonymous]

The versioning feature of the serialization component works fine when using with non-polymorphic classes or derived classes. However, for (pure virtual) base classes the version number set with BOOST_CLASS_VERSION(MyBaseClass, 1) seems to be ignored and during serialization the default version (=0) is always used.

[Robert Ramey]

I can't imagine why this would be a problem.

In order to help with this, I'd really have to see a small example demonstrating the problem.

[anonymous]

Replying to ramey:

I can't imagine why this would be a problem.

In order to help with this, I'd really have to see a small example demonstrating the problem.

I stepped through the code and found that what I'm observing (version number ignored) is due to a different reason, though serializing an inheritance hierarchy may be heavily affected by this issue.

Here is what I found: Consider the following class structure that are contained in two different modules (compilation units).

=========== ModuleA.dll ===========

class ClassA {

template<class Archive> void serialize(Archive &ar, const unsigned int uiVersion);

};

BOOST_CLASS_VERSION(ClassA, 1)

=========== ModuleA.dll ===========

=========== ModuleB.dll ===========

class ClassB : public ClassA {

template<class Archive> void serialize(Archive &ar, const unsigned int uiVersion) {

ar & BOOST_SERIALIZATION_BASE_OBJECT_NVP(ClassA);

}

}

BOOST_CLASS_VERSION(ClassB, 1)

=========== ModuleB.dll ===========

Now consider the following call stack:

[...]
ModuleB.dll!ClassB::serialize(...) <- this triggers BOOST_SERIALIZATION_BASE_OBJECT_NVP(ClassA);
[...]
ModuleB.dll!oserializer<Archive, ClassA>::save_object_data(...) <- this call uses ::boost::serialization::version<ClassA>::value

The problem now is that ::boost::serialization::version<ClassA>::value evaluates to 0 as the specialization (version #1) is only known inside ModuleA.

I think this should be a new ticket if this is indeed the issue...

[Robert Ramey]

is the resolved? As I haven't heard anything lately, I'm going to close this ticket