Rails - Apr 2009 - non-repeatable "undefined error 'path'" error

I''m seeing an error the first time I run a create after restarting my
webserver (thin).  Same create succeeds on all following attempts.

Details:

ruby 1.9.2dev
rails 2.3.2
rack 0.9.1
thin 1.0.0
rmagick 2.9.1

I''m trying to get attachment_fu working in my current setup and have
run into a rather strange situation.  The first image I try to load
produces the following diag in the logfile:

--------------------------
Processing ApplicationController#create (for 127.0.0.1 at 2009-04-25
09:50:27) [POST]
  Parameters: {"authenticity_token"=>"EjZpcSwOswNpjZCHHhypx/
LdfR8m0AmW1Z2caF4/BVI=",
"mugshot"=>{"uploaded_data"=>#<File:/var/
folders/JP/JP7pZWhAHLy-es7AYViHF++++TI/-Tmp-/RackMultipart3767-2>},
"commit"=>"Create"}

NameError (undefined method `path'' for class
`ActionController::UploadedStringIO''):
  <internal:prelude>:8:in `synchronize''
  thin (1.0.0) lib/thin/connection.rb:63:in `pre_process''
  thin (1.0.0) lib/thin/connection.rb:54:in `process''
  thin (1.0.0) lib/thin/connection.rb:39:in `receive_data''
  eventmachine (0.12.6) lib/eventmachine.rb:240:in `run_machine''
  eventmachine (0.12.6) lib/eventmachine.rb:240:in `run''
  thin (1.0.0) lib/thin/backends/base.rb:57:in `start''
  thin (1.0.0) lib/thin/server.rb:150:in `start''
--------------------------

If I then simply reissue the request from the browser the create
action succeeds.  From that point on, all attempts to follow the
new-
>create sequence succeed first try.
Out of curiosity, I tried the same with a new webrick in place of
thin.  In that case the first attempt produces:

--------------------------
Processing ApplicationController#create (for ::1 at 2009-04-25
10:20:30) [POST]
  Parameters: {"authenticity_token"=>"EjZpcSwOswNpjZCHHhypx/
LdfR8m0AmW1Z2caF4/BVI=",
"mugshot"=>{"uploaded_data"=>#<File:/var/
folders/JP/JP7pZWhAHLy-es7AYViHF++++TI/-Tmp-/RackMultipart4180-2>},
"commit"=>"Create"}

NameError (undefined method `path'' for class
`ActionController::UploadedStringIO''):
  <internal:prelude>:8:in `synchronize''
  /opt/RoR/lib/ruby/1.9.1/webrick/httpserver.rb:111:in `service''
  /opt/RoR/lib/ruby/1.9.1/webrick/httpserver.rb:70:in `run''
  /opt/RoR/lib/ruby/1.9.1/webrick/server.rb:183:in `block in
start_thread''
--------------------------

The browser retry yields a "Webrick - stopped (abort trap)" so no
further attemps are possible.  Ignoring for a momemt the different
ways the two servers handle the error allows me to focus on the
common:

--------------------------
NameError
(undef...''path''...''Act...::UploadedStringIO'')
  <internal:prelude>:8:in `synchronize''
--------------------------

Unfortunately, that''s as far as I can get.  I''m clueless on
how to
resolve the ''synchronize'' reference.  There''s
obviously a point in the
sequence (?rack?) that is not correctly dealing with the
initialization.

The fact that thin will allow a do-over is fortunate for me because it
lets me believe the problem isn''t in my code (for a change;-).

Actually it''s not my code but the trivial "mugshots" example
that can
be found through attachment_fu/README.  But, in the interest of full
disclosure:

from mugshots_controller.rb
--------------------------
  def new
    @mugshot = Mugshot.new
  end

  def create
    @mugshot = Mugshot.new(params[:mugshot])
    logger.info "Mugshot attributes hash: #
{@mugshot.attributes.inspect}"
    logger.info "Mugshots_url:  #{mugshots_url.inspect}"
    if @mugshot.save
      flash[:notice] = ''Mugshot was successfully created.''
      redirect_to mugshots_url
    else
      render :action => :new
    end
  end
--------------------------
from mugshots/new.html.erb
--------------------------
<%= error_messages_for :mugshot %>

<% form_for(:mugshot, :url => mugshots_path,
                      :html => { :multipart => true }) do |f| -%>

 <p>
    <label for="mugshot">Upload A Mugshot:</label>
    <%= f.file_field :uploaded_data %>
  </p>
 <p>
    <%= submit_tag ''Create'' %>
  </p>
<% end -%>
--------------------------

thanks in advance,
rick

I think it''s a bit too early to run all that stuff on 1.9.2.

Dmitry

certainly a valid reply, not exactly what I was expecting though...

On Apr 25, 1:21 pm, Dmitry Sokurenko
<Dmitry.Sokure...-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
wrote:
> I think it''s a bit too early to run all that stuff on 1.9.2.
>
> Dmitry