From: Chris Shea <cmshea@...>
Date: 2008-04-09T07:21:08+09:00
Subject: Re: Marshal and singleton.rb - bug?

On Tue, Apr 8, 2008 at 3:19 PM, Eric Hodel <drbrain@segment7.net> wrote:
> On Apr 8, 2008, at 10:31 AM, Chris Shea wrote:
>
> > I was looking over the documentation for singleton.rb (because I want
> > to help it), and while I was double-checking assumptions I had, I
> > noticed that Marshal.load of a serialized singleton instance didn't
> > return the singleton instance, but a new one. I assume that the
> > expected behavior is that Marshal.load would be to return the sole
> > instance.
> >
> > I wrote a simple test case here: http://pastie.textmate.org/177309
> >
>
>  There's assert_same which does the #equal? comparison.

Good to know. Thanks.

> > My installation of Ruby (ruby 1.8.6 (2007-03-13 patchlevel 0)
> > [i686-darwin9.1.0]) fails the last two, as does trunk (r15929).
> >
> > Are my assumptions correct?
> >
>
>  In order for this to work as you expect, Marshal would need to replace the
> existing instance, which is out of it's scope.  It might be possible to
> "fix" by overwriting allocate, but that seems dangerous.

Marshal shouldn't need to replace the existing instance, just return it.

In Singleton::SingletonClassMethods

    def _load(str)
      instance
    end

Perhaps by mistake, JRuby 1.1 (at least) passes those tests.

Chris