RE: [FR-devel] FreeRIDE Error -- make_current is an undefined method.

[Rich Kilmer]

I see the jist of the problem.  At line 158 in EditPanes.rb
renderer().make_current is called and renderer (the method) returns nil
because the Fox renderer is NOT installed as the manager yet.  A couple
of things to note:

1) The renderer method contains this code:

def renderer
 # get a renderer for this editpane
 @renderer = @base_slot["../../render/address@hidden"].manager if   
   @renderer == nil
 return @renderer
end

Just something to keep in mind.  I would highly recommend in using the
Databus and manager objects that you don't keep a handle to the active
manager, as a rule of thumb, but to slots.  Something like this:

@render = @base_slot["../../render/address@hidden"]

def make_current
  @render.manager.make_current
  @log_info << "Making EditPane #{base_slot.path} current"
end

The reason for this is in the ability to replace the manager of a slot
at run time.  I know this may appear rare, but it may not.  Always going
through the slot enable the abstraction to be complete.  This does not
fix our problem, its just a pointer.

2)  The problem of the renderer not existing during the make_current
call:

This is our problem...it appears that the EditPaneRenderFox renderer is
not RUNNING at the time that this code is being called and is causing
our exception.  There will need to be some refactoring to fix the
problem.  I think that the way the abstraction between the EditPane and
the Renderer needs to be looked at.  

A temporary fix is (in EditPane.rb):

require 'singleton'
class NilRenderer
  include Singleton
  def method_missing(meth, *args, &block); end
end

def renderer
 # get a renderer for this editpane
 @renderer = @base_slot["../../render/address@hidden"].manager if   
   @renderer == nil
 return NilRenderer.instance unless @renderer
 return @renderer
end

This "NilRenderer" will catch all method calls and just do nothing.  The
renderer method will keep getting called until the Renderer is set by
the EditPaneRenderFox plugin.  It's a hack, but it works.

-rich


> -----Original Message-----
> From: address@hidden [mailto:freeride-
> address@hidden On Behalf Of Rich Kilmer
> Sent: Sunday, July 14, 2002 10:29 PM
> To: address@hidden
> Subject: RE: [FR-devel] FreeRIDE Error -- make_current is an undefined
> method.
> 
> We must have a race condition where some dependency is not where it
> should be and it depending on the startup it works correctly (or
> sometimes not).  I just checked out FreeRIDE from CVS and on a 1.7.2
> Ruby install on Fox 1.0.11 on Mandrake 8.2 I get the error.
> 
> So...we need to analyze the dependencies of that slot that is being
set
> to make sure they are in order.
> 
> Log file attached
> 
> -rich
> 
> > -----Original Message-----
> > From: address@hidden
[mailto:freeride-
> > address@hidden On Behalf Of Curt Hibbs
> > Sent: Sunday, July 14, 2002 10:08 PM
> > To: address@hidden
> > Subject: RE: [FR-devel] FreeRIDE Error -- make_current is an
undefined
> > method.
> >
> > Laurent Julliard wrote:
> > >
> > >   I have committed all the changes ok in the CVS repo plus I
> > > double-checked that there was no error when running FR from my
> working
> > > copy.
> > >
> > >   The patch suggested by Mizuho San is not necessary. The code as
it
> is,
> > > is ok. The problem comes from the fact that the renderer() method
> call
> > > in EditPanes line 158 return nil which makes me think that the
> > > EditPanesRenderFox plugin is not initialized.
> > >
> > > If this is so it means that the order in which plugins are
> initialized
> > > on your machine and mine is not the same. I guess this is a
> dependency
> > > problem that doesn't show up on Linux but shows up on Windows and
it
> is
> > > probably related to the order in which the Core engine look at the
> > > plugin directories and load them. So It's probably a dependency
> problem
> > > that is not correctly expressed and doesn't show on Linux.
> > >
> > > I have attached my freeride.log (simply start and choose Exit in
the
> > > file menu) so you can see how it compares with yours. I'm
interested
> in
> > > seeing your log file as well.
> > >
> >
> > I'm not trying to be lazy here, I'm just trying to get a Ruby system
> (with
> > FOX) built with full debug symbols so I can work on the threading
> problem,
> > and I wanted to check with you for the obvious, easy things befor
> looking
> > more deeply.
> >
> > Anyway, I've attached my log file (although I haven't even looked at
> it
> > yet
> > myself -- maybe later tonight). Interestingly, it does work for me
> under
> > 1.7.2, but not with 1.6.7 so you suspicion that the order of plugin
> > initialization may be fact be the problem.
> >
> > Curt
> >
> > >
> > > >
> > > > R:\freeride>ruby -I. -Iso -Iredist -e "require
> 'freeride/freeride';
> > > > FreeBASE::Core.startup('./config
> > > > /properties.xml','./config/default.xml')"
> > > > ./plugins/System_GuiCore/EditPanes.rb:158:in `make_current':
> undefined
> > > > method `make_current' for nil
> > > >  (NameError)
> > > >         from ./plugins/System_GuiCore/EditPanes.rb:43:in `start'
> > > >         from ./plugins/System_GuiCore/EditPanes.rb:40:in `call'
> > > >         from ./freebase/databus.rb:407:in `notify'
> > > >         from ./freebase/databus.rb:404:in `each_value'
> > > >         from ./freebase/databus.rb:404:in `notify'
> > > >         from ./freebase/databus.rb:226:in `data='
> > > >         from ./plugins/System_GuiCore/EditPanes.rb:129:in
> `initialize'
> > > >         from ./plugins/System_GuiCore/GuiSetup.rb:110:in `new'
> > > >          ... 57 levels...
> > > >         from ./freebase/core.rb:67:in `initialize'
> > > >         from ./freebase/core.rb:35:in `new'
> > > >         from ./freebase/core.rb:35:in `startup'
> > > >         from -e:1
> > > >
> > > > R:\freeride>
> > > >
> > > >
> > > > _______________________________________________
> > > > Freeride-devel mailing list
> > > > address@hidden
> > > > http://mail.freesoftware.fsf.org/mailman/listinfo/freeride-devel
> > > >
> > >
> > >
> > >