I ran into this issue with Ruby 1.9 (not verified w/ 1.8.7) and the SDL C libraries on Mac OSX.  The specific issue here is related to the Objective-C wrapping of SDL (I guess?).  This issue is not necessarily seen on other platforms.

When running Autospec, noise abounds!

/Users/peterfitzgibbons/.rvm/rubies/ruby-1.9.1-p378/bin/ruby /Users/peterfitzgibbons/ruby-projects/wizardry/vendor/gems/rspec-1.3.0/bin/spec --autospec /Users/peterfitzgibbons/ruby-projects/wizardry/spec/gosu_star_catcher_spec.rb -O spec/spec.opts
Profiling enabled.
/Users/peterfitzgibbons/ruby-projects/wizardry
./Users/peterfitzgibbons/ruby-projects/wizardry
./Users/peterfitzgibbons/ruby-projects/wizardry
.

Top 10 slowest examples:
0.2427560 GosuStarCatcher when initialized should have set dimensions
0.1411460 GosuStarCatcher when initialized should have a background image
0.1077080 GosuStarCatcher when drawn should have background

Finished in 0.526593 seconds

3 examples, 0 failures
2010-03-28 21:47:13.642 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x1005200f0 of class NSCFNumber autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.645 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x100567690 of class NSConcreteValue autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.646 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x11236f080 of class NSCFNumber autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.647 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x1005300c0 of class NSConcreteValue autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.649 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x100538b40 of class NSCFDictionary autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.667 ruby[72302:903] *** attempt to pop an unknown autorelease pool (0x100db0a00)
2010-03-28 21:47:13.668 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x101976320 of class NSCFNumber autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.668 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x1019f1d30 of class NSConcreteValue autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.688 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x11236f080 of class NSCFNumber autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.689 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x100589980 of class NSConcreteValue autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.693 ruby[72302:903] *** __NSAutoreleaseNoPool(): Object 0x1019c1710 of class NSCFDictionary autoreleased with no pool in place - just leaking
2010-03-28 21:47:13.706 ruby[72302:903] *** attempt to pop an unknown autorelease pool (0x100db1a00)

Solution?  Run autospec in terminal window-x with a file redirect of stdout :

autospec 2>>autospec.log

If you have issues, tail the log.  Most of the time, your useful info will be written to &1 (stdout) because autospec is trappping all exceptions and wrapping.

Yeay, nice output!

/Users/peterfitzgibbons/.rvm/rubies/ruby-1.9.1-p378/bin/ruby /Users/peterfitzgibbons/ruby-projects/wizardry/vendor/gems/rspec-1.3.0/bin/spec --autospec /Users/peterfitzgibbons/ruby-projects/wizardry/spec/gosu_star_catcher_spec.rb -O spec/spec.opts
Profiling enabled.
/Users/peterfitzgibbons/ruby-projects/wizardry
./Users/peterfitzgibbons/ruby-projects/wizardry
./Users/peterfitzgibbons/ruby-projects/wizardry
.

Top 10 slowest examples:
0.2427560 GosuStarCatcher when initialized should have set dimensions
0.1411460 GosuStarCatcher when initialized should have a background image
0.1077080 GosuStarCatcher when drawn should have background

Finished in 0.526593 seconds

3 examples, 0 failures

Following Tutorial steps 1 and 2.  Establishing a screen and background.  Not much exciting on the game progress.  As seems usual, Iteration 1 makes minimal app progress and continues to clean up infrastructure issues of Iteration 0

Github Tag for Iteration-01

Observations :

• Most Gosu objects are opaque due to the Ruby wrapper of C++ Gosu library
• Heavy use of mocking and expectations will be required to properly Spec the game

Additions to environment:

• mocha gem : additional mocking tools above Rspec’s internal mocker

I am naming the codebase of my Gosu Ruby Tutorial : Gosu Star Catcher

It’s on GitHub http://github.com/pjfitzgibbons/Gosu-Star-Catcher .

I’m using a basic Ruby library layout.. hand created.

Here are the components involved at the current moment :

rvm 0.1.23 by Wayne E. Seguin (wayneeseguin@gmail.com) [http://rvm.beginrescueend.com/]

ruby 1.9.1p378

*** LOCAL GEMS ***

archive-tar-minitar (0.5.2)
autotest (4.2.9)
autotest-fsevent (0.2.2)
bundler (0.9.11)
columnize (0.3.1)
ffi (0.6.3)
gosu (0.7.18)
linecache19 (0.5.11)
nice-ffi (0.3)
rake (0.8.7)
rsdl (0.1.4)
rspec (1.3.0)
ruby-debug-base19 (0.11.23)
ruby-debug19 (0.11.6)
ruby-sdl-ffi (0.2)
ruby_core_source (0.1.4)
rubygame (2.6.2)
rubygems-update (1.3.6)
sys-uname (0.8.4)

Editor is Textmate

I’ve been into a game development project using Ruby.  This is inspired by my friend Jared working on re-implementing Wizardry 1 (of Apple ][e fame) on (of all things) Excel and VBA.

So, I’ve reviewed several libraries.  Currently I’m trying Gosu for Ruby.

I will blog of my experience in following the Gosu Ruby Tutorial

The code was not much to look at.  We mostly rspec’d a few methods and learned that just because you ‘def to_s’ doesn’t mean the implicit string conversion uses it!
I think the real issue we uncovered was that the implicit string conversion produces escaped string and calling #to_s produces unescaped strings.

Ruby Gems we used to setup the environment.

• rspec
• ZenTest
• redgreen (for windows users)

heredocs — we talked about how to produce heredocs   You can either use multi-line double-quoted, or <<-eot

text = <<-eot
This is the heredoc text
It can even include evaluations #{Time.now} !!
eot

More about heredocs : http://blog.jayfields.com/2006/12/ruby-multiline-strings-here-doc-or.html

A windows-implementation of textmate : http://www.e-texteditor.com/

The Ruby Way — Hal Fulton

http://www.google.com/url?sa=t&source=web&ct=res&cd=1&ved=0CAwQFjAA&url=http%3A%2F%2Fwww.amazon.com%2FRuby-Way-Second-Techniques-Programming%2Fdp%2F0672328844&ei=OuLwSsj_PIa1tgf59N26Cw&usg=AFQjCNHhpmT3C35Xiianfordqb21_DsWIA&sig2=Pb631QnepV_sAD8wmRuPTA

The Rails Way — Obie Fernandez

http://www.google.com/url?sa=t&source=web&ct=res&cd=3&ved=0CBYQFjAC&url=http%3A%2F%2Fwww.amazon.com%2FRails-Way-Obie-Fernandez%2Fdp%2F0321445619&ei=V-LwSunCEYeVtgfWib26Cw&usg=AFQjCNGpPcvEJDxIOGEwWZZIp9HXnbPC0A&sig2=0ofdwEJfrMOm7U3bR7c5OQ

The Rails Way Blog therailsway.com

Enumerable Module

In this chapter, I’ll discuss the adapter initialization; how Rails 2.x reads from database.yml and launches the database initialization.

Rails Boot

Here is a condensation of the Rails boot process from Rails Guts.  As Mike Gunderloy suggests, it will be really helpful to have a blank rails app open with

 rake gems:freeze 

invoked :

1. script/server (vendor/rails/railties/lib/commands/server.rb) requires environment.rb from the local path (./)
2. in environment.rb

    Rails::Initializer.run do |config| 

3. Initializer.run has a default parameter

    configuration = Configuration.new 

   (that’s Rails::Configuration in railties/lib/initializer.rb)

4. Configuration#initialize is long, including

    self.database_configuration_file = default_database_configuration_file 

5. Configuration#default_database_configuration_file is :

   def default_database_configuration_file
    File.join(root_path, 'config', 'database.yml')
    end
   

6. Notice that the default command on Rails::Initializer#run was :process, so Initializer#process is now called, which includes #initialize_database …

   def initialize_database
    if configuration.frameworks.include?(:active_record)
    ActiveRecord::Base.configurations = configuration.database_configuration
    ActiveRecord::Base.establish_connection
    end
    end
   

   1. Configuration#database_configuration is where the database.yml is loaded and returned to ActiveRecord::Base.configurations :

      def database_configuration
       require 'erb'
       YAML::load(ERB.new(IO.read(database_configuration_file)).result)
       end
      

7. Now were getting somewhere!!  AR::Base#establish_connection(nil) :
   1. because the call is with spec = nil, #establish_connection is re-executed with RAILS_ENV
   2. this time through, spec = RAILS_ENV (ie, :development), so #establish_connection is called again, this time with the database.yml hash-key of RAILS_ENV (from Configuration#database_configuration)
   3. 3rd time’s a charm!  Now, with spec set to config hash, require the adapter gem,

       gem "activerecord-#{spec[:adapter]}-adapter" 

      and execute the adapter-method (which is expected to be def’d from the adapter’s gem) :

       adapter_method = "#{spec[:adapter]}_connection" 

The adapter-method is the entry-point of our new adapter, so in my next post I’ll pick-apart the #establish_connection process and prepare for our multi-connection pooling adapter’s version.

Well, I don’t know how many parts this will have… so lets just say 5 until I re-edit the introduction.

Inspiration and Perspiration

Thomas Edison told us how to invent.  So let’s get to it.  The Adapter we’ll build will assist in handling the problem of database-failover in Rails 2.3.2 applications

Inspired – Seamless Database Pool

Seamless Database Pool gem provides us with a template to follow.  The SDP is focused purely on connection logic.  The true db adapter is used within the pool connections.  Our Adapter, like SDP, will simply serve as a proxy to handle the event when an ActiveRecord method (#select, #create, #update, #delete) encounters a DatabaseConnectionError.

Perspired – Activerecord Failover Adapter

Ok, we’re going to have to work for it this time.  Event though we have SDP as a template, there is A LOT of logic in there that is really not useful to us.  SDP is a load-balancing pool based upon a weighted one-write/many-read database cluster / replication group.

Requirements

Ahh, the real meat of our cookout.

1. Configure Master + Slaves(multiple) in database.yml just like SDP
2. Use Master Always.  Well, until failover.
3. Use periodic timeouts to attempt re-connect to master
4. Retry db queries on alternate connection
5. Rollback transactions upon connection failure… then retry them?
   1. Re-try the transaction in the alternate connection
6. Allow #establish_connection to target master/slave (especially needed for targeted migrations)
7. Allow db:migrate to target master/slave through use of RAILS_ENV

In my next post, we’ll start with #1 Configure Master + Slavees in database.yml

I recently completed configuration of continuous-integration between two servers.

Here’s the steps and notes that I have for the move.

System Layout

The system of servers/services comes together like this :

Git / Redmine

Git and Redmine run on server ‘redmine-server’, which is a CentOS 5 Intel box.  OpenSSH is installed, Redmine runs on Ruby 1.8.7, yum package install.
User on this server is ‘git’

Cerberus continuous integration

Cerberus runs on an OSX box named ‘ci-server’, Leopard.
Ruby on this box is installed using ruby-switcher, with 1.8.7p173 and 1.9.1p129r23412.
My customization of that ruby-switcher is here ruby_switcher.sh.
User on this server is ‘cerberus’

SSH setup Git -> Cerberus

The Git repository lives on a different server, so SSH is employed to protect passwords-cleartext over the wire.

There is an additional constraint in that Git hooks are “non-interactive”, so ssh keys must be configured to not challenge.  There are [security implications] to this.  The option I have chosen is passwordless “backup keys” for specific purposes.

git@redmine-server $ ssh-keygen -t rsa -C git_ci_server@redmine_server
Generating public/private rsa key pair.
Enter file in which to save the key (/home/git/.ssh/id_rsa): \
  .ssh/id_rsa_git_ci_server
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in .ssh/id_rsa_git_ci_server.
Your public key has been saved in .ssh/id_rsa_git_ci_server.pub.
The key fingerprint is:
42:62:16:4c:00:74:7c:a9:96:90:5a:75:5d:08:c5:f9 git_ci_server@redmine_server

Move the key to the target server ci-server

[git@redmine-server]$ scp .ssh/id_rsa_git_ci_server.pub \
  cerberus@ci-server:id_rsa_git_ci_server.redmine-server

Password:

id_rsa_git_ci_server.pub                     100%  410     0.4KB/s   00:00

[git@redmine-server]$ ssh deployer@web-ci
Password:

Last login: Mon Jun 15 09:20:00 2009 from redmine-server

[cerberus@ci-server]$ cat id_rsa_git_ci_server.redmine-server
ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAtOVHRlhLmUI/9XglQs6u9F1reUwAq2U0clJI2l
j5kAX5XkVW7FoiiGo4BfAqP3j69ynMBnLx/eUtcYcxHUKZfikwQbXnUdAUd8mIXoNhOfVZ/gqr
i0yS0XHpkGU2lJve3soqEIyYu2gxf5QPP18ibJeyFZ5QRTPUATfi1c6Pb12tsr4O9H+b5T08cU
ejvQP5ISyYH3ysPZwLQ3d1Ods9ZItm7h9/eoZQxgXpVLxUkQ3NO+JoktKlruGajTRgm5xP7NBd
gUZsC2vs84/mW64RWSzbl8fA5oKUhKQIDt7QSOGx2UVGYAoarFGTsvps5g9WnXTJ37XRT+ORHI
cpmO90mQ== git_ci_server@redmine_server
[cerberus@ci-server]$ cat id_rsa_git_ci_server.redmine-server >> .ssh/authorized_keys
[cerberus@ci-server]$ exit

SSH Backup-Key

From ‘man ssh’ :

 -i identity_file
    Selects a file from which the identity (private key) for RSA or DSA authentication is read.  The default is
    ~/.ssh/identity for protocol version 1, and ~/.ssh/id_rsa and ~/.ssh/id_dsa for protocol version 2.  Identity files
    may also be specified on a per-host basis in the configuration file.  It is possible to have multiple -i options
    (and multiple identities specified in configuration files).

This Git server does not have ssh-agent running, so using -i will cause the indicated file to be the only identity used in the ssh session.

Now test the login using the backup-key switch of ’ssh’

[git@redmine-server]$ ssh -i .ssh/id_rsa_git_ci_server cerberus@ci-server
Last login: Mon Jun 15 10:53:38 2009 from redmine-server
[cerberus@ci-server]$

The backup-key switch will be used in the Git post-receive hook for executing Cerberus build upon Git Push

Git Post-Receive

Simple.  Use ssh to call a build script on ci-server, which we’ll build after the break…

#!/bin/sh
#
# An example hook script for the post-receive event
#
# This script is run after receive-pack has accepted a pack and the
# repository has been updated.  It is passed arguments in through stdin
# in the form
#  <oldrev> <newrev> <refname>
# For example:
#  aa453216d1b3e49e7f6f98441fa56946ddcd6a20 68f7abf4e6f922807889f52bc043ecd31b79f814 refs/heads/master
#
# see contrib/hooks/ for an sample, or uncomment the next line (on debian)
#

ssh -i ~/.ssh/id_rsa_git_ci_server cerberus@ci-server \
 bin/cerberus-build racc-webapp

NOTE: the use of the -i switch

This hook is applied to each repository that will be monitored by Cerberus Continuous Integration.  It must be made executable so Git will execute it!

[git@redmine-server]$ cp post-receieve repos/my_app/.git/hooks/post-receive
[git@redmine-server]$ chmod +x repos/my_app/.git/hooks/post-receive

Cerberus configuration

Using the ruby-switcher, install the ruby of your choice (’install_ruby_187p173′).  The use it (’use_ruby_187′).

Install cerberus gem (note we are using local gem storage to accomodate ruby-switcher):

[cerberus@ci-server]$ gem install cerberus

Configure cerberus globals, especially email server config (for SMTP send).

I also configured the builder/rake to handle new migrations (Rails 2.3) and Rspec (rake spec)

[cerberus@ci-server]$ cat .cerberus/config.yml
publisher:
#  active: mail jabber rss campfire
  mail:
    address: smtp.test.com
    port: 25
    domain: test.com
    authentication: :none
    sender: Cerberus CI <cerberus@test.com>
#    address: smtp.gmail.com
#    port: 587
#    domain: gmail.com
#    authentication: plain
#    user_name: someuser
#    password: somepassword
#    on_event: all
#  jabber:
#    jid: somemailbox@gmail.com/cerberus
#    password: somepassword
#  twitter:
#    login: twitter_username
#    password: twitter_password
#  irc:
#    nick: cerb
#    server: irc.freenode.net
#    channel: cerberus
#  campfire:
#    url: http://someemail:password@cerberustool.campfirenow.com/room/51660
#  rss:
#    file: /usr/www/rss.xml
builder:
  rake:
    task: db:migrate spec
#changeset_url: POINT_TO_YOUR_TRAC/changeset/
#hook:
#  rcov:
#    on_event: successful, setup #by default - run hook for any state
#    action: 'export CERBERUS_HOME=/home/anatol && sudo chown www-data -R /home/anatol/cerberus && rcov' #Add here any hook you want

Configure the cerberus project. Note how the config points back to the repository

[cerberus@ci-server]$ cerberus add git@redmine-server:my_app/.git \
  APPLICATION_NAME=my-app SCM=git

Edit the configuration of the project to add project-level mailing addresses.  The BUILD WILL FAIL WITHOUT THIS!!!

[cerberus@ci-server]$ cat .cerberus/config/my-app.yml
---
publisher:
  mail:
    recipients: someuser@test.com
scm:
  url: git@redmine-server:repos/my_app/.git
  type: git

An initial build ‘cerberus build my-app’ is required to set the sources.  This one will likely fail, as it is necessary to configure non-SCM files like ‘database.yml’.

You should receive email from “Cerberus CI” declaring your failure.

You have completed config.  Now FIX YOUR BROKEN CHECKIN FTW!!

Episode 2: Using the story from the ground up.

I could go into great detail… but I’m beat.  Creating the dev environment from the ground up was a lot of work.  I also needed to get some project details under way.

Here is what I needed to get Features and Specs to load up the library properly

Cucumber

Create a cucumber.yml.  This is necessary to set the default profile, which I wanted to get a little more pretty printing during the cucumber command.

In a step definition (grid_steps.rb), I needed the $LOAD_PATH to point to the lib, so I added the following to the top of support/env.rb

$:.unshift File.expand_path(File.dirname(__FILE__) + "/../../lib")

Also support RSpec expectations/matchers within the steps, so

require 'spec/expectations'

RSpec

Update the $LOAD_PATH here from a different location in the tree than features, from ./spec

$:.unshift File.expand_path(File.dirname(__FILE__) + "/../lib")

Autotest integration

In order to get cucumber integrated with autotest, add this to cucumber.yml

autotest: --format pretty features
autotest-all: --format pretty features

Also autospec requires an environment switch, which I have added to ./bashrc for convenience

export AUTOFEATURE=true

Amber to Red to Green to Amber to Red to Green…

With all the prettyness set up with cucumber and RSpec, we now have a new color in the Red-Green cycle.

Cucumber Features’ Given/When/Then lines are displayed in Amber until each is properly trapped by a defined Given/When/Then in a steps definition file (./features/step_definitions/*.rb).

RSpec specs are displayed in amber if the spec is not defined (name only) like this

it "this spec is not defined yet"

Well, that’s about all I can handle for now.

I will make an attempt to trap the Amber-Red-Green cycle in git commits during my work tomorrow.

Here’s the entirement development environment for TicTacToe.  This was actually rebuilt for this project as I deleted my last dev environment (A VirtualBox instance) during a system-repair process (wipe-drive-reinstall of Vista

• Ubuntu 8.10 Desktop as VirtualBox Guest on Vista Home Premium on Compaq C771.  NAT, 1gb ram, 16gb vhd, shared folders
• Ruby 1.8 + dependencies from Apt
• Rubygems 1.3.1 installed from .tgz (Ubuntu 8.10 has trouble with Apt’s rubygems install)
• Rails, RSpec, RSpec-Rails, Cucumber gems installed
• NetBeans 6.5.1 / All Modules.  NBGit installed for Git integration.