From: ts <decoux@...>
Date: 2002-04-20T00:39:07+09:00
Subject: Re: GC problem in C++ extension

>>>>> "P" == Paul Brannan <pbrannan@atdesk.com> writes:

P> Hmm... if I change line 327 of gc.c from:
P>   VALUE obj;
P> to:
P>   static VALUE obj;
P> then my program works.

 OK here your example :


pigeon% cat uu.cc
#include <string>
#include <iostream>
#include <ruby.h>
#include <intern.h>
  
#if defined(RUBY_METHOD_FUNC)
#undef RUBY_METHOD_FUNC
#endif
typedef VALUE (*RUBY_METHOD_FUNC)();
  
class Foo
{
public:
  Foo(char const * s) : s_(s) { }
  ~Foo() { std::cout << s_ << std::endl; }
  static void free(Foo * f) { delete f; }
private:
  std::string s_;
};
  
VALUE test_foo(VALUE)
{
  Foo * f = new Foo("testing...");
  VALUE v = Data_Wrap_Struct(rb_cObject, 0, (RUBY_DATA_FUNC)Foo::free, f);
  return Qnil; // line 25
}
  
int main(int argc, char * argv[])
{
  ruby_init();
  ruby_init_loadpath();
  ruby_options(argc, argv);
  rb_define_global_function("foo", RUBY_METHOD_FUNC(test_foo), 0);
  ruby_run();
}
pigeon% 

pigeon% g++ -g -O2  -I/usr/local/lib/ruby/1.6/i686-linux -I. -I/usr/local/include  -o uu uu.cc -L/usr/local/lib -L/usr/local/lib/ruby/1.6/i686-linux -L/usr/local/lib -lruby -lm -lcrypt -ldl
pigeon%
 
pigeon% uu -e 'foo; GC.start; puts "exiting..."'
testing...
exiting...
pigeon% 


 This is 1.6.7 and I've the same result with 1.6.5

P> then my program works.  It seems that the problem is rb_newobj() is

 No, there is no problem in rb_newobj().

 What you see is that it's difficult to know which variables are on the
 stack and worst it's perhaps compiler dependent.

 Change your compiler and you can have a different result.


Guy Decoux