From: Rich Kilmer <rich@...>
Date: 2002-11-14T10:49:41+09:00
Subject: Re: FXRuby issue

One other one in responder2.rb:

SELTYPE(sel) needs to be Fox::SELTYPE(sel)

-rich

> -----Original Message-----
> From: Rich Kilmer [mailto:rich@infoether.com] 
> Sent: Wednesday, November 13, 2002 8:43 PM
> To: ruby-talk ML; lyle@knology.net
> Cc: ruby-talk@ruby-lang.org
> Subject: Re: FXRuby issue
> 
> 
> Lyle,
> 
> I already made this change to FreeRIDE this afternoon.  Every 
> class that uses Fox constants now specifically includes Fox.  
> I agree that this is better/safer from a namespace 
> perspective.  Are you going to patch FXRuby with the changes 
> I/you list here (explicit reference to Fox:: in responder.rb 
> and responder2.rb)?
> 
> Thanks for always being so supportive!
> 
> Regards,
> 
> Rich
> 
> > -----Original Message-----
> > From: lyle@knology.net [mailto:lyle@knology.net]
> > Sent: Wednesday, November 13, 2002 7:35 PM
> > To: rich@infoether.com
> > Cc: ruby-talk@ruby-lang.org
> > Subject: Re: FXRuby issue
> > 
> > 
> > 
> > On Wed, 13 Nov 2002 15:07:41 -0500, "Rich Kilmer"
> > <rich@infoether.com> wrote :
> > 
> > > I had to patch responder2.rb and responder.rb to make Fox 
> work on my
> > > machine:
> > > 
> > > responder2.rb:
> > > 
> > > 36. @pseudoTarget = Fox::FXPseudoTarget.new
> > 
> > Yes, this one I'm OK with.
> > 
> > > responder.rb:
> > > 
> > > 7. include Fox
> > 
> > No; doing this would pollute the global namespace (more on
> > that in a moment). A better solution is to use the 
> > fully-qualified names for the various Fox module-level 
> > methods and constants that the Responder module methods need, 
> > e.g. this code:
> > 
> >   # Define range of function types
> >   def FXMAPTYPES(typelo, typehi, func)
> >     addMapEntry(MKUINT(MINKEY, typelo),
> >                 MKUINT(MAXKEY, typehi),
> >                 func)
> >   end
> > 
> > becomes:
> > 
> >   # Define range of function types
> >   def FXMAPTYPES(typelo, typehi, func)
> >     addMapEntry(Fox::MKUINT(Fox::MINKEY, typelo),
> >                 Fox::MKUINT(Fox::MAXKEY, typehi),
> >                 func)
> >   end
> > 
> > I did a little digging to try to understand why FXRuby worked
> > the way it did in version 1.0.13 and what changed in version 
> > 1.0.14, and I think it all boils down to some changes made to 
> > 'responder2.rb'. For some time (i.e. at least in version 
> > 1.0.13 and perhaps earlier) it has been true that when you 
> > require the 'fox' feature:
> > 
> >     require 'fox'
> > 
> > it was also requiring some of the FXRuby library modules, and
> > in particular it was doing:
> > 
> >     require 'fox/responder2'
> > 
> > Now for FXRuby-1.0.13, the beginning of 'responder2.rb'
> > looked like this:
> > 
> >     require 'fox'
> >     require 'fox/responder'
> >     
> >     include Fox
> > 
> > and that last line was the culprit: it was dumping
> > *everything* from the Fox module into the global namespace, 
> > which kinda defeats the purpose of having put those things 
> > into their own module in the first place. This code was 
> > changed in version 1.0.14, where 'responder2.rb' no longer 
> > "include"-s the Fox module into the global namespace.
> > 
> > So the behavior in FXRuby versions 1.0.14 (and later) is
> > "safer" in the sense that none of the stuff from the Fox 
> > module should ever escape into the global namespace unless 
> > you explicitly make it so (i.e. by doing an 'include Fox' at 
> > the top level). Since this is the behavior you were relying 
> > on, is it possible to just modify the affected FreeRIDE 
> > file(s) to do an 'include Fox' immediately after they require 
> > 'fox'? [I realize that I am asking this question without 
> > knowing how many source files we're talking about.] Making 
> > this change to the FreeRIDE sources would have the 
> > consequence that everything from the Fox module will get 
> > dumped into the global namespace -- but that's what was 
> > happening *anyways* when you were running with FXRuby 1.0.13 ;)
> > 
> > 
> 
> 
> 
> 
>