From: Craig Files <Craig_Files@...>
Date: 2001-08-02T06:18:53+09:00
Subject: [ruby-talk:18994] Re: Possible SWIG Ruby bug

Craig Files wrote:

  After get no response, I thought that I would try to determine if this was a SWIG or a Ruby bug.  It is a SWIG bug, in that
SWIG does not differentiate between the pointer returned from a constructor or a method.  If anybody is interested, I can
email the changes to the SWIG ruby.cxx file that I have made.... I have also posted this bug on SourceForge, but my solution
was not accepted by the bug tracker.

> Hi,
>
>   It seems that when using SWIG with Ruby, there is a bug that causes a destructor for an object to be called more than once.  If an object is newed
>
> and deleted within C++, but a pointer to that object is returned to Ruby during execution of a Ruby program, the object is deleted twice (once in
>
> C++ and once in Ruby).
>
> The following code recreates the bug.  The resulting core file shows the point of failure in the destructor for class Bar.
>
> Has anyone else encountered this bug when using SWIG?  Is there a way around this problem?  That is, we would like the ability
>
> to delete a object Bar if we create it in Ruby (i.e. bar = Bar.new()).
>
> ----------------------------------------- Foo.h
>
> class Bar {
>   public:
>     int * data;
>     Bar() : data (new int) {}
>     ~Bar() { delete data; }
> };
>
> class Foo {
> private:
>   Bar *V;
> public:
>   Foo() : V(new Bar) {}
>   ~Foo() { delete V; }
>   const Bar *GetBar() { return  V; }
> };
>
> ----------------------------------------- Foo.i
>
> %module Foo
> %{
> #include "Foo.h"
> %}
>
> class Bar {
> public :
>   Bar();
>   ~Bar();
> };
>
> class Foo {
> public:
>   Foo();
>   ~Foo();
>   const Bar *GetBar();
> }
>
> ------------------------------------------- foo.rb
>
> require 'Foo'
>
> f = Foo::Foo.new()
> b = f.GetBar()
> puts "exiting"
>
> -------------------------------------------- output
>
> exiting
> test.rb:5: [BUG] Segmentation fault
> ruby 1.6.4 (2001-06-04) [i686-linux]
> Aborted
>
>    Thanks,
>
>       Craig
>
> --
> ___________________________________________________________________________
>  Craig Files                                |                  /
>   http://fccf0183l.ftc.agilent.com/~cfiles/ |       __o       /
>   craig_files@agilent.com                   |       \<,      /
> ____________________________________________|_____()/ ()____/

--
___________________________________________________________________________
 Craig Files                                |                  /
  http://fccf0183l.ftc.agilent.com/~cfiles/ |       __o       /
  craig_files@agilent.com                   |       \<,      /
____________________________________________|_____()/ ()____/ (970)288-0183