does the header file of polymorphic_derived1's include the following?
 

BOOST_CLASS_TYPE_INFO(

polymorphic_derived1,

extended_type_info_no_rtti<polymorphic_derived1>

)

Also, the *.cpp file of polymorphic_derived1 should contain

BOOST_CLASS_EXPORT(polymorphic_derived1)

Robert Ramey

"Daniel Roberts" <impulsionaudio@gmail.com> wrote in message news:14125aaf0901151528p8f354ffp9409248decf7958c@mail.gmail.com...
Thanks for the response. Yes, I am using the no rtti option. I downloaded test_no_rtti.cpp and got it to work as-is. I then made the following changes:

1) Move polymorphic_base's definition into its own header file.
2) Move polymorphic_derived1's definition into its own header file.
3) Move the definition of polymorphic_derived1::get_key() into an implementation file.

Making these changes causes the assertion failure. I am using boost v1.36.0 and  Visual Studio 2008 9.0.30729.1 SP

Thanks,
Daniel Roberts


I presume you're using the "no rtti" option? get_key used only for those systems which
use which depend on extended_type_info_no_rtti to handle extended_type_info functionality.

with which compiler?, which version of boost?
 "Daniel Roberts" <impulsionaudio@gmail.com> wrote in message news:
14125aaf0901142057s6ba59d08tc98a6990a5f78bca@mail.gmail.com...
 I have a derived class that exports its type and overrides get_key for serialization through a base class pointer. If I move the definition of the overridden get_key function out of the header into a cpp file, an assertion fails on line 49 of extended_type_info.cpp.

 Take a look at the example "test_no_rtti".  the class polymorphic_derived has a get_key
 declaration - but the get_key definition is not inline.  This would seem identical your case.

 You might want to make a small variation of test_no_rtti such that the definition of
 polymorphic_derived1::get_key defintion is moved to a separate source and verify that
 it builds and runs correctly.

 If I define any other functions in the cpp file, I get a linker error saying that the get_key symbol is already defined. It looks to me that any serializable derived class must be completely defined in the header file. Is this correct?

 If so, why does such a limitation exist?


 I don't think this limitation exists

 Thank you

 Robert Ramey


_______________________________________________
Boost-users mailing list
Boost-users@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users