Stuff… And Things…

I do… stuff and build… things.

Cross-Domain AJAX Request in Development

| Comments

Sometimes an AJAX request on a page you’re developing needs to hit a server on a different domain. Web browsers’ Same-Origin Policy means (among other things) that other domains called from AJAX need to be whitelisted using the Access-Control-Allow-Origin header.

While this limits the damage malicious Javascript injected into a page can do, it’s annoying in development. You could whitelist your dev box, 0.0.0.0, or 127.0.0.1 (depending on how you work), but that’s ugly.

Fortunately, there’s an easier way, temporarily disable same-origin policy.

Luminate Online Server API Gem

| Comments

As part of the work Indra did for Stand Up To Cancer’s telethon, I need to make use of Convio/blackbaud’s Luminate Online Server APIs. Luminate Online is a widely used fundraising platform and relationship manager for non-profits.

I’ve turned that development work into open source in the form of gem providing Ruby bindings to the APIs.

I have only implemented a handful of endpoints, Luminate Online doesn’t have any kind of test mode or sandbox, any testing has to be done with live data. However, I’ve used a meta programming approach making it easy to quickly add additional endpoints.

All of the endpoints I have implemented have been used in anger.

The details are in the repo: https://github.com/spikex/luminate

It’s in the Logs, Stupid

| Comments

Yesterday, I was involved in a fire drill around the launch of a new Rails site on a very tight time frame. The site worked fine in development/staging, but the index was taking upwards of 10 seconds to render in production.

Because it worked outside of production we leapt to the conclusion it was related to the hosting infrastructure. We checked Apache, Passenger, server load, network configuration, and so on. Nothing.

Finally, I thought to check log/production.log, and there is was:

1
2
3
Processing by BlahController#index as */*
  Rendered blah/_carousel.html.erb (5795.2ms)
  Rendered blah/index.html.erb within layouts/site (5801.5ms)

We quickly tracked it down some image processing that wasn’t being cached. It didn’t show outside of production because the image assets were different. I won’t bore you with the details.

However, I will bore enlighten you as to my point. When debugging a problem, start with the simple things. The Rails logs aren’t very detailed, but they provide more than enough information to quickly drill down to problems in your code.

Building Ruby 1.8.7 on FreeBSD 10

| Comments

If you try to build Ruby 1.8.7 with RVM you will see:

1
2
3
4
5
# rvm install 1.8.7
[...]
./lib/fileutils.rb:1430: [BUG] unexpected local variable
ruby 1.8.7 (2012-02-08 patchlevel 375) [amd-freebsd10]
*** [./.rbconfig.time] Signal 6

This is because FreeBSD 10 ships with the Clang compiler instead of GCC and Ruby versions priory to 1.9.3-p194 do not support Clang.

The work-around, is to install GCC, either with pkg:

1
# pkg install gcc

or from Ports:

1
2
# cd /usr/ports/lang/gcc
# make install clean

Once you have gcc you can install 1.8.7 with:

1
# CC=/usr/local/bin/gcc47 rvm install 1.8.7

This should work with older versions of Ruby 1.9 as well.

What on earth would you want to install 1.8.7? Two words: Legacy Apps…

Dynamic Keys for Strongbox

| Comments

Previously, Strongbox, my gem for using Public Key Encryption with ActiveRecord, allowed only one key pair for encrypting all of the records for a given ActiveRecord model. I’ve had a number of requests to make it possible to dynamically choose the keys on a per record basic and version 0.6.0 adds this feature.

The values of :public_key, :private_key, and :key_pair can be in one of the following formats:

A string containing path to a file. This is the default interpretation of a string.

1
2
encrypt_with_public_key :secret,
   :key_pair => File.join(RAILS_ROOT,'config','keypair.pem')

A string contanting a key in PEM format, needs to match this the regex /^-+BEGIN .* KEY-+$/

1
2
3
encrypt_with_public_key :secret,
  :key_pair =>
    "-----BEGIN RSA PRIVATE KEY-----\nProc-Type: 4,ENCRYPTED\nDEK-Info: DES-EDE3-CBC,BFE700E4DDA4C434\n\nDfZD7FKM4zLJdb[...]"

A symbol naming a method to call. Can return any of the other valid key formats.

1
2
3
encrypt_with_public_key :secret,
  :public_key => :public_key_method,
  :private_key => :private_key_attribute

An instance of OpenSSL::PKey::RSA. Must be unlocked to be used as the private key.

1
2
3
4
KEY_PAIR = OpenSSL::PKey::RSA.new(key,password)
# [...]
encrypt_with_public_key :secret,
   :key_pair => KEY_PAIR

Using this, you can automatically create per record public keys:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
require 'openssl'

# Assumes the migration contains:
#  t.string :description
#  t.binary :secret
# that you are collecting a password to encrypt the private key,
# and that the secret is small

class User < ActiveRecord::Base
  attr_accessor :password
  encrypt_with_public_key :secret,
    :key_pair => :key_pair

  def after_initialize
     rsa_key = OpenSSL::PKey::RSA.new(2048)
     cipher =  OpenSSL::Cipher::Cipher.new('des3')
     key_pair = rsa_key.to_pem(cipher,self.password) + rsa_key.public_key.to_pem
  end
end

Important Caveat - Currently, Strongbox encrypts the attribute as soon as it’s assigned (this will change in version 1.0). The means that the public key must be available before the attribute is assigned, hence the use of after_initialize to generate the key pair. Even so, this will fail if you do something like:

1
user = User.new(params[:user])

because the attributes are set before after_initialize is called.

Instead, use something like:

1
2
3
user = User.new
user.password = params[:password]
user.attributes = params[:user]

Version 1.0 will allow you to control when the encryption occurs, making this less of an issue.

Building Emacs.App From Source on OS X Lion

| Comments

As always, I like building Emacs for my Mac from source. It lets me live on the cutting edge and have tigher control of the version I’m running. If building software from source isn’t your thing then skip the rest of this article and consider installing Emacs using Homebrew, MacPorts, or Fink

I’ve written about building Emacs in the past, but OS X Lion brings a few complexities to the process.

Octopress Blog Re-Launch

| Comments

After years of running on Wordpress, today I’m relaunching my blog using Octopress. Octopress is a blogging framework build on top of Jekyll which in turn is system designed for publishing static sites from source files, be they Markdown, HAML, SASS, etc. This style fits nicely in to my everyday workflow, so my hope is it will get me writing on a regular basis.

Wordpress is great if you want a WYSIWYG blog and I still recommend to my clients. But, if you spend your days in Emacs or Vim, running rake tasks and git commits, you might want to give Octopress a try.

Legacy Development With Pow

| Comments

Pow is a zero-config Rack server that makes developing Rack apps (include Rails apps) a snap on Mac OS X.

You install Pow:

1
curl get.pow.cx | sh

You create a symlink to your app:

1
2
cd ~/.pow
ln -s ~/dev/myapp

You point your browser to http://myapp.dev/ and there’s you app!   Awesome!

However, there is a downside:  Pow doesn’t play nicely with Apache (or any server listening on port 80). Life isn’t all greenfield, if in the course of the day you need to work on PHP or CGI legacy apps Pow is not so simple.   Pow creates a firewall rule that redirects port 80 to its port; to access Apache you need to either toggle the firewall rule on and off or move Apache to a different port all together. And now you’re running two web servers.  There has to be a better way.

The Win And there is, make your legacy app a Rack App.  Thanks to the rack-legacy gem, this is actually quite simple.

First, install the rack-legacy and rack-rewrite gems:

1
gem install rack-legacy rack-rewrite

(To sudo or not to sudo, that’s up to you).

Next,  in the top level of your legacy app create this config.ru file:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
require 'rack'
require 'rack-legacy'
require 'rack-rewrite'

INDEXES = ['index.html','index.php', 'index.cgi']

use Rack::Rewrite do
  rewrite %r{(.*/$)}, lambda {|match, rack_env|
    INDEXES.each do |index|
      if File.exists?(File.join(Dir.getwd, rack_env['PATH_INFO'], index))
        return rack_env['PATH_INFO'] + index
      end
    end
    rack_env['PATH_INFO']
  }
end

use Rack::Legacy::Php, Dir.getwd
use Rack::Legacy::Cgi, Dir.getwd
run Rack::File.new Dir.getwd

The Details Rack::Legacy::Php runs any requested file with the extension .php. Rack::Legacy::Cgi runs any requested file that is set executable (which means you’ll need to make sure your .html files are not). Files that don’t end in .php and aren’t executable are served as static content by Rack::File.

The INDEXES array contains a list of files to check for if a directory is requested (just like Apache’s DirectoryIndex directive). You can change the order or use different names (default.htm anyone?).

The Catch rack-legacy uses the php-cgi command line program to run scripts and while PHP ships with current versions of OS X, php-cgi is not included. You’ll need to install PHP using MacPort/Homebrew/Fink/etc. That’s beyond the scope of this post but, if you’re doing this kind of development, it’s probably not beyond you.

The Caveat This is probably not a fast as running PHP using the Apache module and it’s certainly not as fast as something like FastCGI. If you are primarily developing legacy apps you probably should stick to Apache. However, if you mostly work with Rack apps and just occasionally need legacy support, this is a great way to go.

Deploying “Modular” Style Sinatra Apps With Phusion Passenger

| Comments

There’s plenty of documentation on how to deploy “Classic” style Sintra applications with Phusion Passenger, but it’s not immediately obviously how to deploy the new “Modular” style app (created with Sinatra::Base). Fortunately, it’s simple, the resulting class can be passed to “run” inside you “config.ru” file, something like:

1
2
3
4
require 'rubygems'
require File.join(File.dirname(__FILE__), 'lib/my_app')

run MyApp

That’s it.

Instructions you find for “Classic” Sintra apps have you setting “:env” to the ENV[‘RACK_ENV’] before running the app. As of Sintra 1.0 it’s “:environment” and it’s automatically set to ENV[‘RACK_ENV’] (or “:development” if RACK_ENV is not set). You’ll also see instructions for setting “:run” to false, this is not nessecary for “Modular” apps.

Handling 403 Forbidden Login Pages With (Ruby) Mechanize

| Comments

Instead of having a dedicated login page, some sites return a 403 Forbidden HTTP status code and include the login form in an HTML body of a custom 403 page. For example, Drupal admin pages work this way. While this may seem a little odd, it works; all modern browser will display the HTML and few, if any, will note the Forbidden status.

Mechanize on the other hand raise an exception when it receives a 403 status. Fortunately, it returns the page it received as part of that exception. Here’s how to handle it:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
mechanize = Mechanize.new

begin
  login_page = mechanize.get("http://localhost/admin")
rescue Mechanize::ResponseCodeError => exception
  if exception.response_code == '403'
    login_page = exception.page
  else
    raise # Some other error, re-raise
  end
end
login = login_page.form_with(:action => '/login') do |f|
  f.field_with(:name => 'user').value = user
  f.field_with(:name => 'password').value = password
end.submit
raise 'Login Failed' if login.body !~ /Logged in!/

This code also works in the case where you don’t get a forbidden status, so it can be used generically.

For bonus points you can use the same code in a Cucumber step by changing:

1
login_page = mechanize.get("http://localhost/admin")

to:

1
login_page = webrat.adapter.mechanize.get(path_to('the admin page'))

(assuming you’ve set up “the admin page” in paths.rb).