Setting Session Options in Rails 2.3

I'm trying to figure out how I should set my session options for Rails
2.3. In Rails 2.2:

  session :cookie_only => false, :only => :swf_upload

In Rails 2.3, it looks like I need to use:

  request.session_options['cookie_only'] = false

But how do I tell it that I only want that option on a specific method
in the controller (the :only option)?


p.s. The old way works, but the deprecation warnings are driving me
nuts when I run my tests.

I am having exactly the same problem. However, for me, old:

class ImagesController < ApplicationController
  session :cookie_only => false

does *not* work anymore. TomRossi7, can you double check does it work
for you, please?

The only way I worked this around, is to get sessions directly from
SessionStore, which is extremely bad. Also I can't find tests
regarding cookie_only option in 2.3.

Have you tried using the following in your controller?

  session_options['cookie_only'] = false

I think that will work, I just don't know how to specify which methods
I want to apply the option.

Hi all,

Reviving this old thread with a I am developing a site that needs to be
accessible to mobile phones, and I am having issues with session

While the default "cookie_only" session tracking makes a lot of sense
(it prevents session fixation attacks), there are these cases when your
clients do not support cookies (say, many mobile phones, for example)
and you just need to do request-parameter-based session tracking.

One should be able to use the "cookie_only" session option and set it to
"false" where appropriate, but this is seriously broken in Rails 2.3.

1) Setting a default by adding "config.action_controller.session =
{:cookie_only => false}" in environment.rb DOES work.

2) Inside your controller (say, a before_filter method),
"request.session_options[:cookie_only] = false" to false will NOT work.

Unfortunately, 1) is not an acceptable solution, as it opens the whole
site to session fixation attacks.

FWIW, I think I have traced down the culprit to the load_session method
in active_store.rb around line 165:

          sid = request.cookies[@key]
          unless @cookie_only
            sid ||= request.params[@key]

This code becomes a problem in 2.3, because the session has been pushed
down to the Rack middleware layer, where your abstract_store gets
initialized once and for all, way before any of your controller code
gets executed. As a result, your abstract_store's @cookie_only and @key
are set once and for all, according to the default session options. Any
further changes to the session_options[:cookie_only] or
session_options[:key] will simply be ignored.

Here is a tentative monkey patch that restores the proper functionality
of session_options[:cookie_only] and session_options[:key].


hello Martin, Please forgive my terrible english.

I tried your mokye patch,
but it didn't work well.

I tried like next,

スクルト神守 wrote:

hello Martin, Please forgive my terrible english.

I tried your mokye patch,
but it didn't work well.

Yes, I just found out myself the patch is only a partial solution: it
worked on my simple prototype, but once integrated into my production
system, it stopped working. Like you, I found that "load session" was
happening before "set option".

The problem is that the session is automatically loaded as soon as
something tries to use it. So if some plugin uses the session before
your controller gets control, then the session will be loaded before
your before_filter can set the options correctly.

What you can do to find the culprit: use the ruby debugger and add a
"debugger" statement inside the load_session monkey patch function, then
run your server with "script/server --debugger". When you try to access
you site, your console will drop into the debugger, and you can issue a
"where" command, that will give you a stack trace, so you can find out
what triggered the call to load_session.

In my case, it was the render_component plugin.

Unfortunately, I do need the render_component plugin (for my admin pages
that use ActiveScaffold)...

So, the problem is still open... Can anyone help? Should I fill a bug

hello Martin, thank you for your answer!!

Unfortunately, the problem still be here, but,
thank you for telling me the debugging way!
it is very useful!! ^^)