From: Paul Brannan <paul@...>
Date: 2001-12-11T01:31:51+09:00
Subject: [ruby-talk:28098] Re: Boost.Ruby ?

On Sun, Dec 09, 2001 at 10:00:34PM +0900, Issac wrote:
> I'm starting to wonder about whether it's a good idea to extend 
> Ruby with C++.  When I tried to extend Ruby with a C module,

Be sure to read [ruby-talk:17943] (where Chris Uzdavinis makes some good
points about exception safety in C++ extensions).

> everything went smoothly.  As soon as I tried to do it with C++
> everything started falling to pieces.  First, gnu c++ complained
> about some mismatched parameters even though the code 
> compiled just fine with gcc:
> 
> Test.cpp: In function `void Init_Test()':
> Test.cpp:46: ANSI C++ prohibits conversion from `(long unsigned int)' to `(...)'
> 
> Test.cpp:47: ANSI C++ prohibits conversion from `(long unsigned int, long unsign
> ed int)' to `(...)'
> make: *** [Test.o] Error 1

You need to make a cast:

  typedef VALUE (*RUBY_METHOD_FUNC)();
  rb_define_method(
    rb_cMyClass,
    "foo",
    RUBY_METHOD_FUNC(foo),
    -1);

> So then I added -fpermissive to the CXXFLAGS in the Makefile.
> This got the Test.cpp file to compile, but the linker wasn't pleased:
> 
> Test.exp(.edata+0x30):fake: undefined reference to `Init_Test'
> collect2: ld returned 1 exit status
> dllwrap: gcc exited with status 1
> make: *** [Test.so] Error 1
> 
> I think this is because of C++'s name-mangling.  In Test.o I found 
> Init_Test__Fv and _Init_Test__Fv but no Init_Test.  Still, I don't
> know how to fix this.

Try this:

  extern "C" Init_Test () {
    ...
  }

Paul