On Friday 12 March 2004 12:14 am, Jonathan Pryor wrote: > 2. C++ compilers use a technique called "name mangling", so that you > can have function overloads. For example, your "print (const char*)" > function is *actually* the linker symbol:
> The right way is to instead tell the C++ compiler to disable name > mangling. This is done by using an ``extern "C"'' modifier in your C++ > code. The downside to this is that you can not have two C++ function > overloads that are both marked as ``extern "C"'', as this would cause > confusion. I would also add that the ``extern "C"'' does not help in situations where you need to communicate with C++ classes. In this case, with GCC an intermediate C library is almost certainly necessary. If something similar VS.NET's Managed Extensions for C++ existed for GCC, this might be avoided. There at least 3 reasons why an intermediate layer is needed. First, there is no way to create a new instance of a C++ object from outside C++ unless the C++ library provides an appropriate function. There is no obvious way to determine how much memory to allocate for a given C++ object, so even if it were possible to call a C++ constructor directly to initialize the object, allocating the object in the first place seems impossible. Second, C++ compilers use name-mangling to avoid duplication of symbols when several classes define the same method and when overloaded methods are used. This mangling is platform dependent. It is possible to use system-specific utilities to produce a list of mangled and unmanagled names and then match them up to create a conversion table. So this is more of an annoyance than an an outright dead-end. Third, when optimization is specified, C++ compilers frequently inline any methods those bodies are defined within the class scope (and possibly other methods as well). Moreover, if every invocation of a method is inlined, some compilers do not by default emit an out-of-line copy for the method. _______________________________________________ Mono-list maillist - [EMAIL PROTECTED] http://lists.ximian.com/mailman/listinfo/mono-list