* Fixed plugin generator so that generated unit tests would subclass ActiveSupport::TestCase, also introduced a helper script to reduce the needed require statements #1137 [Mathias Meyer]
* Rails::Initializer#initialize_time_zone raises an error if value assigned to config.time_zone is not recognized. Rake time zone tasks only require ActiveSupport instead of entire environment [Geoff Buesing]
* Added config.cache_store to environment options to control the default cache store (default is FileStore if tmp/cache is present, otherwise MemoryStore is used) [DHH]
* Changed the default database from mysql to sqlite3, so now running "rails myapp" will have a config/database.yml that's setup for SQLite3 (which in OS X Leopard is installed by default, so is the gem, so everything Just Works with no database configuration at all). To get a Rails application preconfigured for MySQL, just run "rails -d mysql myapp" [DHH]
* Turned on ActionView::Base.cache_template_loading by default in config/environments/production.rb to prevent file system stat calls for every template loading to see if it changed (this means that you have to restart the application to see template changes in production mode) [DHH]
* Introduce SecretKeyGenerator for more secure session secrets than CGI::Session's pseudo-random id generator. Consider extracting to Active Support later. #10286 [Hongli Lai]
* Changed the default logger from Ruby's own Logger with the clean_logger extensions to ActiveSupport::BufferedLogger for performance reasons [DHH]. (You can change it back with config.logger = Logger.new("/path/to/log", level).)
* Added a default 422.html page to be rendered when ActiveRecord::RecordInvalid, ActiveRecord::RecordNotSaved, or ActionController::InvalidAuthenticityToken is raised [DHH]
* db:create creates the database for the current environment if it's on localhost. db:create:all creates local databases for all environments. #8783 [matt]
* Generators: look for generators in all gems, not just those suffixed with _generator, in the gem's generators or rails_generators directory. Allow use of the rails_generators directory instead of the standard generators directory in plugins also. #8730 [Dr Nic, topfunky]
* Fixed that dispatcher preparation callbacks only run once in production mode. Mock Routes.reload so that dispatcher preparation callback tests run. [Rick]
* Removed breakpointer and Binding.of_caller in favor of relying on ruby-debug by Kent Sibilev since the breakpointer has been broken since Ruby 1.8.4 and will not be coming back [DHH]
To use the new debugger, start your server with script/server --debugger and insert a call to 'debugger'
(instead of 'breakpoint') where you want to jump into the debugger.
BACKWARDS INCOMPATIBILITY NOTE: You must remove the default line 12 from config/environments/development.rb:
config.breakpoint_server = true
This configuration option is no longer available. Rails will fail to start in development mode as long as
* Canonicalize RAILS_ROOT by using File.expand_path on Windows, which doesn't have to worry about symlinks, and Pathname#realpath elsewhere, which respects symlinks in relative paths but is incompatible with Windows. #6755 [Jeremy Kemper, trevor]
* Deprecation: remove components from controller paths. [Jeremy Kemper]
* Split out the basic plugin locator functionality into an abstract super class. Add a FileSystemLocator to do the job of checking the plugin_paths for plugins. Add plugin_locators configuration option which will iterate over the set of plugin locators and load each of the plugin loaders they return. Rename locater everywhere to locator. [Marcel Molina Jr.]
* Split plugin location and loading out of the initializer and into a new Plugin namespace, which includes Plugin::Locater and Plugin::Loader. The loader class that is used can be customized using the config.plugin_loader option. Those monkey patching the plugin loading subsystem take note, the internals changing here will likely break your modifications. The good news is that it should be substantially easier to hook into the plugin locating and loading process now. [Marcel Molina Jr.]
* Added source-annotations extractor tasks to rake [Jamis Buck]. This allows you to add FIXME, OPTIMIZE, and TODO comments to your source code that can then be extracted in concert with rake notes (shows all), rake notes:fixme, rake notes:optimize and rake notes:todo.
* Added config/initializers where all ruby files within it are automatically loaded after the Rails configuration is done, so you don't have to litter the environment.rb file with a ton of mixed stuff [DHH]
* Make config.plugins affect the load path and the dependencies system. Allows you to control plugin loading order, and keep disabled plugins off the load path. [James Adam]
* Added one-letter aliases for the three default environments to script/console, so script/console p will load the production environment (t for test, d for development) [DHH]
* Added ActiveRecord::Base.clear_active_connections! in development mode so the database connection is not carried over from request to request. Some databases won't reread the schema if that doesn't happen (I'm looking at you SQLite), so you have to restart the server after each migration (= no fun) [DHH]
* Make sure that exceptions which are thrown outside of the user code try their best to be handeled in ApplicationController#rescue_action [Tobias Luetke]
* Only wrap request processing with our USR1 signal handler so FastCGI can trap it and raise an exception while waiting for connections. Idle processes exit immediately rather than waiting for another request; active processes gracefully exit when the request is finished. [Jeremy Kemper]
* Alter prior change to use require_dependency instead of require_or_load. Causes ApplicationController to be reloaded again. Closes #6587. [Nicholas Seckar]
* Update to latest Prototype, which doesn't serialize disabled form elements, adds clone() to arrays, empty/non-string Element.update() and adds a fixes excessive error reporting in WebKit beta versions [Thomas Fuchs]
* Added config.plugins to control which plugins are loaded #6269 [skaes]. By default, everything in vendor/plugins will be loaded, but if you specify config.plugins, only those will be loaded. Example:
* Fixed migration creation to work with namespaced models, so script/generate model Gallery::Image will use create_table :gallery_images #6327 [BobSilva]
* Fixed rename_table on SQLite tables with indexes defined #5942 [brandon@opensoul.org]
* Remove the uncanny default of adding all app/models/*/ directories to the load path. This change will break application which expect the current behavior. As
* Fix Dispatcher.reset_application! so that AR subclasses are removed and Observers re-initialized *after* Reloadable classes are removed. Closes #5743. [Rick Olson]
* Added Mongrel-spawning capabilities to script/process/spawner. Mongrel will be the default choice if installed, otherwise FCGI is tried [DHH]. Examples:
spawner # starts instances on 8000, 8001, and 8002 using Mongrel if available
spawner fcgi # starts instances on 8000, 8001, and 8002 using FCGI
spawner mongrel -i 5 # starts instances on 8000, 8001, 8002, 8003, and 8004 using Mongrel
spawner -p 9100 -i 10 # starts 10 instances counting from 9100 to 9109 using Mongrel if available
spawner -p 9100 -r 5 # starts 3 instances counting from 9100 to 9102 and attempts start them every 5 seconds
Also note that script/process/reaper is Mongrel capable. So the combination of spawner and reaper is a built-in alternative to something like mongrel_cluster.
* Added uninstall.rb hook to plugin handling, such that plugins have a way of removing assets and other artifacts on removal #5003 [takiuchi@drecom.co.jp]
* The app generator sets a session key in application.rb so apps running on the same host may distinguish their cookies. #2967 [rcoder, rails-bug@owl.me.uk]
* Added pid file usage to script/process/spawner and script/process/reaper along with a directive in default config/lighttpd.conf file to record the pid. They will all save their pid file in tmp/pids [DHH]
* Windows: include MinGW in RUBY_PLATFORM check. #2982 [okkez000@gmail.com, Kaspar Schiess]
* Stop swallowing errors during rake test [Koz]
*1.2.2* (February 5th, 2007)
* Fix gem deprecation warnings, which also means depending on RubyGems 0.9.0+ [Chad Fowler]
* Require the dispatcher for Rails::Configuration#to_prepare. [Rick]
*1.2.1* (January 16th, 2007)
* Updated to Active Record 1.15.1, Action Pack 1.13.1, Action Mailer 1.3.1, Action Web Service 1.2.1
*1.2.0* (January 16th, 2007)
* Update to Prototype 1.5.0. [Sam Stephenson]
* Generator: use destination path for diff tempfiles. #7015 [alfeld]
* Fixed that webrick would strip leading newlines and hang connection #4156 [psross]
* Ensure plugins are in the Dependencies.load_once_paths collection by default. [Rick]
If you really want your plugins to reload, add this to the very top of init.rb:
Dependencies.load_once_paths.delete(lib_path)
* Fix scaffold_resource generator so it respects the --pretend argument when creating the routes file. Closes #6852 [fearoffish]
* Fix Webrick Daemon dispatching bug regarding a bad current working directory. Closes #4899 [Rick Olson]
* Make config.plugins affect the load path and the dependencies system. Allows you to control plugin loading order, and keep disabled plugins off the load path. [James Adam]
* Don't generate a components directory in new Rails apps. [Jeremy Kemper]
* Fixed script/process/spawner to work properly with Mongrel including in -r (daemonize mode) [DHH]
* Deprecated the name route "root" as it'll be used as a shortcut for map.connect '' in Rails 2.0 [DHH]
* Fixed that script/server running against Mongrel should tail the proper log regardless of the environment [DHH]
* Update initializer to load Rails::VERSION as soon as possible. Closes #6698. [Nicholas Seckar]
* Added ActiveRecord::Base.clear_active_connections! in development mode so the database connection is not carried over from request to request. Some databases won't reread the schema if that doesn't happen (I'm looking at you SQLite), so you have to restart the server after each migration (= no fun) [DHH]
* Made RAILS_GEM_VERSION work for beta gems too, so specifying 1.1.6 will give you 1.1.6.4520 if available [DHH]
* Update to Prototype and script.aculo.us [5579]. [Sam Stephenson, Thomas Fuchs]
* Made script/server work with -e and -d when using Mongrel [DHH]
* Make sure that exceptions which are thrown outside of the user code try their best to be handeled in ApplicationController#rescue_action [Tobias Luetke]
* Rails::VERSION::STRING should always be available without having to require 'rails/version'. #6244 [fearoffish]
* Add grep-based fallback to reaper, to work in pidless setups [Jamis Buck]
* Only wrap request processing with our USR1 signal handler so FastCGI can trap it and raise an exception while waiting for connections. Idle processes exit immediately rather than waiting for another request; active processes gracefully exit when the request is finished. [Jeremy Kemper]
* Alter prior change to use require_dependency instead of require_or_load. Causes ApplicationController to be reloaded again. Closes #6587. [Nicholas Seckar]
* Rake: use absolute paths to load lib and vendor tasks so they may be run outside of RAILS_ROOT. #6584 [jchris]
* scaffold_resource generator uses _path named routes and head instead of render :nothing => true. #6545 [Josh Susser]
* Generator can show diff on file collision to help you decide whether to skip or overwrite. #6364 [jeffw, Jeremy Kemper]
* Generated directories are recursively svn added, like mkdir -p. #6416 [NeilW]
* resource and scaffold_resource generators add a restful route to config/routes.rb [Jeremy Kemper]
* Revert environment changes for autoload_paths. [Koz]
* Clean up the output of rake stats, de-emphasise components and apis, and remove the indents for tests [Koz]
* Added option to script/process/spawner of specifying the binding address #5133 [zsombor]
* Update environment.rb comments to include config.autoload_paths. Closes #6478 [caio]
* Update scaffold to use new form_tag block functionality. Closes #6480. [BobSilva]
* Plugin generator: check for class collisions. #4833 [vinbarnes@gmail.com]
* Mailer generator: handle mailers in modules, set mime_version in unit test. [Jeremy Kemper]
* Set $KCODE to 'u' by default to enable the multibyte safe String#chars proxy. [Koz]
* Added config.plugins to control which plugins are loaded #6269 [skaes]. By default, everything in vendor/plugins will be loaded, but if you specify config.plugins, only those will be loaded. Example:
* Clean up html on included error pages. [Tim Lucas]
* Fixed default 404.html and 500.htmls to remove extreme ugliness and include human language [DHH]
* Update to latest Prototype and script.aculo.us trunk versions [Thomas Fuchs]
* PostgreSQL: db:test:purge closes open database connections first. #6236 [alex]
* Fixed test:uncommitted on Windows (backslash issue) #4999 [paul@paulbutcher.com]
* Fixed migration creation to work with namespaced models, so script/generate model Gallery::Image will use create_table :gallery_images #6327 [BobSilva]
* Fixed rename_table on SQLite tables with indexes defined #5942 [brandon@opensoul.org]
* Added default timeout setting of 5 seconds to SQLite3 database.yml configurations [DHH]
* Added generated attribute options to script/generate model, like the one found in scaffold_resource and resource [DHH]. Examples:
./script/generate model post title:string created_on:date body:text published:boolean
* Added script/generate resource which works just like scaffold_resource, but creates empty placeholders instead of predefined [DHH]
* script/runner can run files, pass on arguments, and be used as a shebang. #6286 [Tuxie, dlpond]
#!/usr/bin/env /path/to/my/app/script/runner
# Example: just start using your models as if you are in script/console
* Remove the uncanny default of adding all app/models/*/ directories to the load path. This change will break application which expect the current behavior. As
* Assign Routing.controller_paths; fix script/about and rails info controller. [Nicholas Seckar]
* Don't warn dispatcher of Reloadable deprecations. [Nicholas Seckar]
* Rearrange application resetting and preparation, fix bug with leaking subclasses hash in ActiveRecord::Base [Rick Olson]
ActiveRecord::Base.reset_subclasses is called before Dependencies are cleared and classes removed.
ActiveRecord::Base.instantiate_observers is called during a Dispatcher preparation callback.
* Add missing mock directories from the autoload_paths configuration. [Rick Olson]
* Nested controller scaffolding also nests the generated layout. [iain d broadfoot]
* Add "require 'dispatcher'" to webrick server in the continuing quest to squash webrick weirdness. [Nicholas Seckar]
* Add autoload_paths support to Initializer. [Nicholas Seckar]
* Fix Dispatcher.reset_application! so that AR subclasses are removed and Observers re-initialized *after* Reloadable classes are removed. Closes #5743. [Rick Olson]
* Clarify usage of script/plugin source. Closes #5344. [james.adam@gmail.com]
* Add Dispatcher.to_prepare and config.to_prepare to provide a pre-request hook. [Nicholas Seckar]
* Tweak the Rails load order so observers are loaded after plugins, and reloaded in development mode. Closed #5279. [Rick Olson]
* Added that you can change the web server port in config/lighttpd.conf from script/server --port/-p #5465 [mats@imediatec.co.uk]
* script/performance/profiler compatibility with the new ruby-prof, including an option to choose the results printer. #5679 [shugo@ruby-lang.org]
* Fixed the failsafe response so it uses either the current recognized controller or ApplicationController. [Rick Olson]
* Make sure script/reaper only reaps dispatcher pids by default, and not the spawner's pid. [Jamis Buck]
* Fix script/plugin about so it uses about.yml and not meta.yml. [James Adam]
* Dispatcher processes rescued actions with the same controller that processed the request. #4625 [sd@notso.net]
* rails -d frontbase to create a new project with a frontbase database.yml. #4945 [mlaster@metavillage.com]
* Ensure the logger is initialized. #5629 [mike@clarkware.com]
* Added Mongrel-spawning capabilities to script/process/spawner. Mongrel will be the default choice if installed, otherwise FCGI is tried [DHH]. Examples:
spawner # starts instances on 8000, 8001, and 8002 using Mongrel if available
spawner fcgi # starts instances on 8000, 8001, and 8002 using FCGI
spawner mongrel -i 5 # starts instances on 8000, 8001, 8002, 8003, and 8004 using Mongrel
spawner -p 9100 -i 10 # starts 10 instances counting from 9100 to 9109 using Mongrel if available
spawner -p 9100 -r 5 # starts 3 instances counting from 9100 to 9102 and attempts start them every 5 seconds
Also note that script/process/reaper is Mongrel capable. So the combination of spawner and reaper is a built-in alternative to something like mongrel_cluster.
* Update scaffolding functional tests to use :id => people(:first) instead of :id => 1. #5612 [evan@protest.net]
* db:test:clone should remove existing tables before reloading the schema. #5607 [sveit@tradeharbor.com]
* Fixed migration generation for class names like ACLController #5197 [brad@madriska.com]
* Added show_source_list and show_call_stack to breakpoints to make it easier to get context #5476 [takiuchi@drecom.co.jp]. Examples:
* Mongrel support for script/server. #5475 [jeremydurham@gmail.com]
* Fix script/plugin so it doesn't barf on invalid URLs [Rick]
* Fix plugin install bug at dir with space. (closes #5359) [Yoshimasa NIWA]
* Fix bug with 'script/plugin install' so it reports unknown plugin names correctly. [Rick]
* Added uninstall.rb hook to plugin handling, such that plugins have a way of removing assets and other artifacts on removal #5003 [takiuchi@drecom.co.jp]
* Create temporary dirs relative to RAILS_ROOT when running script/server #5014 [elliot@townx.org]
* Minor tweak to dispatcher to use recognize instead of recognize!, as per the new routes. [Jamis Buck]
* Make "script/plugin install" work with svn+ssh URLs. [Sam Stephenson]
* Added lib/ to the directories that will get application docs generated [DHH]
* The app generator detects the XAMPP package's MySQL socket location. #3832 [elliot@townx.org]
* The app generator sets a session key in application.rb so apps running on the same host may distinguish their cookies. #2967 [rcoder, rails-bug@owl.me.uk]
* Distinguish the spawners for different processes [DHH]
* Added -n/--process to script/process/spawner name the process pid (default is dispatch) [DHH]
* Namespaced OrderedHash so the Rails implementation does not clash with any others. (fixes #4911) [Julian Tarkhanov]
* Replace Ruby's deprecated append_features in favor of included. [Marcel Molina Jr.]
* Added script/process/inspector to do simple process status information on Rails dispatchers keeping pid files in tmp/pids [DHH]
* Added pid file usage to script/process/spawner and script/process/reaper along with a directive in default config/lighttpd.conf file to record the pid. They will all save their pid file in tmp/pids [DHH]
*1.1.6* (August 10th, 2006)
* Additional security patch
*1.1.5* (August 8th, 2006)
* Mention in docs that config.frameworks doesn't work when getting Rails via Gems. #4857 [Alisdair McDiarmid]
* Change the scaffolding layout to use yield rather than @content_for_layout. [Marcel Molina Jr.]
* Includes critical security patch
*1.1.4* (June 29th, 2006)
* Remove use of opts.on { |options[:name] } style hash assignment. References #4440. [headius@headius.com]
* Updated to Action Pack 1.12.3, ActionWebService 1.1.4, ActionMailer 1.2.3
*1.1.3* (June 27th, 2006)
* Updated to Active Record 1.14.3, Action Pack 1.12.2, ActionWebService 1.1.3, ActionMailer 1.2.2
* Make all ActionView helpers available in the console from the helper method for debugging purposes. n.b.: Only an 80% solution. Some stuff won't work, most will. [Marcel Molina Jr.]
* Added 'port open?' check to the spawner when running in repeat mode so we don't needlessly boot the dispatcher if the port is already in use anyway #4089 [guy.naor@famundo.com]
* Added that rake rails:freeze:edge will now just export all the contents of the frameworks instead of just lib, so stuff like rails:update:scripts, rails:update:javascripts, and script/server on lighttpd still just works #4047 [DHH]
* Added namespaces to all tasks, so for example load_fixtures is now db:fixtures:load. All the old task names are still valid, they just point to the new namespaced names. "rake -T" will only show the namespaced ones, though [DHH]
* CHANGED DEFAULT: ActiveRecord::Base.schema_format is now :ruby by default instead of :sql. This means that we'll assume you want to live in the world of db/schema.rb where the grass is green and the girls are pretty. If your schema contains un-dumpable elements, such as constraints or database-specific column types, you just got an invitation to either 1) patch the dumper to include foreign key support, 2) stop being db specific, or 3) just change the default in config/environment.rb to config.active_record.schema_format = :sql -- we even include an example for that on new Rails skeletons now. Brought to you by the federation of opinionated framework builders! [DHH]
* Added -r/--repeat option to script/process/spawner that offers the same loop protection as the spinner did. This deprecates the script/process/spinner, so it's no longer included in the default Rails skeleton, but still available for backwards compatibility #3461 [ror@andreas-s.net]
* Changed the default configuration for lighttpd to use tmp/sockets instead of log/ for the FastCGI sockets [DHH]
* Added a default configuration of the FileStore for fragment caching if tmp/cache is available, which makes action/fragment caching ready to use out of the box with no additional configuration [DHH]
* Changed the default session configuration to place sessions in tmp/sessions, if that directory is available, instead of /tmp (this essentially means a goodbye to 9/10 White Screen of Death errors and should have web hosting firms around the world cheering) [DHH]
* Added tmp/sessions, tmp/cache, and tmp/sockets as default directories in the Rails skeleton [DHH]
* Added that script/generate model will now automatically create a migration file for the model created. This can be turned off by calling the generator with --skip-migration [DHH]
* Added -d/--database option to the rails command, so you can do "rails --database=sqlite2 myapp" to start a new application preconfigured to use SQLite2 as the database. Removed the configuration examples from SQLite and PostgreSQL from the default MySQL configuration [DHH]
* Added SIGTRAP signal handler to RailsFCGIHandler that'll force the process into a breakpoint after the next request. This breakpoint can then be caught with script/breakpointer and give you access to the Ruby image inside that process. Useful for debugging memory leaks among other things [DHH]
* Changed default lighttpd.conf to use CWD from lighttpd 1.4.10 that allows the same configuration to be used for both detach and not. Also ensured that auto-repeaping of FCGIs only happens when lighttpd is not detached. [DHH]
* Added Configuration#after_initialize for registering a block which gets called after the framework is fully initialized. Useful for things like per-environment configuration of plugins. [Michael Koziarski]
* Added check for RAILS_FRAMEWORK_ROOT constant that allows the Rails framework to be found in a different place than vendor/rails. Should be set in boot.rb. [DHH]
* Add plugin library directories to the load path after the lib directory so that libraries in the lib directory get precedence. #2910. [james.adam@gmail.com]
* PostgreSQL: the purge_test_database Rake task shouldn't explicitly specify the template0 template when creating a fresh test database. #2964 [dreamer3@gmail.com]
* Eliminate Subversion dependencies in scripts/plugin. Correct install options. Introduce --force option to reinstall a plugin. Remove useless --long option for list. Use --quiet to quiet the download output and --revision to update to a specific svn revision. #2842 [Chad Fowler, Rick Olson]
* Changed all script/* to use #!/usr/bin/env ruby instead of hard-coded Ruby path. public/dispatcher.* still uses the hard-coded path for compatibility with web servers that don't have Ruby in path [DHH]
* Added an omnipresent RailsInfoController with a properties action that delivers an HTML rendering of Rails::Info (but only when local_request? is true). Added a new default index.html which fetches this with Ajax. [Sam Stephenson]
* Added default lighttpd config in config/lighttpd.conf and added a default runner for lighttpd in script/server (works like script/server, but using lighttpd and FastCGI). It will use lighttpd if available, otherwise WEBrick. You can force either or using 'script/server lighttpd' or 'script/server webrick' [DHH]
* New configuration option config.plugin_paths which may be a single path like the default 'vendor/plugins' or an array of paths: ['vendor/plugins', 'lib/plugins']. [Jeremy Kemper]
* Plugins are discovered in nested paths, so you can organize your plugins directory as you like. [Jeremy Kemper]
* Refactor load_plugin from load_plugins. #2757 [alex.r.moon@gmail.com]
* The freeze_gems Rake task accepts the VERSION environment variable to decide which version of Rails to pull into vendor/rails. [Chad Fowler, Jeremy Kemper]
* Removed script.aculo.us.js, builder.js and slider.js (preperation for move of scriptaculous extensions to plugins, core scriptaculous will remain in Railties) [Thomas Fuchs]
* The freeze_edge Rake task does smarter svn detection and can export a specific revision by passing the REVISION environment variable. For example: rake freeze_edge REVISION=1234. #2663 [Rick Olson]
* Added that plugins can carry generators and that generator stub files can be created along with new plugins using script/generate plugin <name> --with-generator [DHH]
* Made the rails command use the application name for database names in the tailored database.yml file. Example: "rails ~/projects/blog" will use "blog_development" instead of "rails_development". [Florian Weber]
* Added Rails framework freezing tasks: freeze_gems (freeze to current gems), freeze_edge (freeze to Rails SVN trunk), unfreeze_rails (float with newest gems on system)
* Made the instantiated/transactional fixtures settings be controlled through Rails::Initializer. Transactional and non-instantiated fixtures are default from now on. [Florian Weber]
* Add db_schema_dump and db_schema_import rake tasks to work with the new ActiveRecord::SchemaDumper (for dumping a schema to and reading a schema from a ruby file).
* Added create_sessions_table, drop_sessions_table, and purge_sessions_table as rake tasks for databases that supports migrations (MySQL, PostgreSQL, SQLite) to get a table for use with CGI::Session::ActiveRecordStore
* Added -c/--svn option to the generator that'll add new files and remove destroyed files using svn add/revert/remove as appropriate #2064 [kevin.clark@gmail.com]
* Moved all the shared tasks from Rakefile into Rails, so that the Rakefile is empty and doesn't require updating.
* Added Rails::Initializer and Rails::Configuration to abstract all of the common setup out of config/environment.rb (uses config/boot.rb to bootstrap the initializer and paths)
* Allow ERb in the database.yml file (just like with fixtures), so you can pull out the database configuration in environment variables #1822 [Duane Johnson]
* Added convenience controls for FCGI processes (especially when managed remotely): spinner, spawner, and reaper. They reside in script/process. More details can be had by calling them with -h/--help.
* Dropped the 'immediate close-down' of FCGI processes since it didn't work consistently and produced bad responses when it didn't. So now a TERM ensures exit after the next request (just as if the process is handling a request when it receives the signal). This means that you'll have to 'nudge' all FCGI processes with a request in order to ensure that they have all reloaded. This can be done by something like ./script/process/repear --nudge 'http://www.myapp.com' --instances 10, which will load the myapp site 10 times (and thus hit all of the 10 FCGI processes once, enough to shut down).
* Look for app-specific generators in RAILS_ROOT/generators rather than the clunky old RAILS_ROOT/script/generators. Nobody really uses this feature except for the unit tests, so it's a negligible-impact change. If you want to work with third-party generators, drop them in ~/.rails/generators or simply install gems.
* Added support for SQL Server in the database rake tasks #1652 [ken.barker@gmail.com] Note: osql and scptxfr may need to be installed on your development environment. This involves getting the .exes and a .rll (scptxfr) from a production SQL Server (not developer level SQL Server). Add their location to your Environment PATH and you are all set.
* Added a VERSION parameter to the migrate task that allows you to do "rake migrate VERSION=34" to migrate to the 34th version traveling up or down depending on the current version
* Extend Ruby version check to include RUBY_RELEASE_DATE >= '2005-12-25', the final Ruby 1.8.2 release #1674 [court3nay@gmail.com]
* Improved documentation for environment config files #1625 [court3nay@gmail.com]
* Add the option to manually manage garbage collection in the FastCGI dispatcher. Set the number of requests between GC runs in your public/dispatch.fcgi [skaes@web.de]
* Allow dynamic application reloading for dispatch.fcgi processes by sending a SIGHUP. If the process is currently handling a request, the request will be allowed to complete first. This allows production fcgi's to be reloaded without having to restart them.
* Added a 'whiny nil' that's aim to ensure that when users pass nil to methods where that isn't appropriate, instead of NoMethodError? and the name of some method used by the framework users will see a message explaining what type of object was expected. Only active in test and development environments by default #1209 [Michael Koziarski]
* Fixed the test_helper.rb to be safe for requiring controllers from multiple spots, like app/controllers/article_controller.rb and app/controllers/admin/article_controller.rb, without reloading the environment twice #1390 [Nicholas Seckar]
* Allow graceful exits for dispatch.fcgi processes by sending a SIGUSR1. If the process is currently handling a request, the request will be allowed to complete and then will terminate itself. If a request is not being handled, the process is terminated immediately (via #exit). This basically works like restart graceful on Apache. [Jamis Buck]
* Fixed that purge_test_database would use database settings from the development environment when recreating the test database #1122 [rails@cogentdude.com]
* Removed SCRIPT_NAME from the WEBrick environment to prevent conflicts with PATH_INFO #896 [Nicholas Seckar]
* Removed ?$1 from the dispatch.f/cgi redirect line to get rid of 'complete/path/from/request.html' => nil being in the @params now that the ENV["REQUEST_URI"] is used to determine the path #895 [dblack/Nicholas Seckar]
* Combined the script/environment.rb used for gems and regular files version. If vendor/rails/* has all the frameworks, then files version is used, otherwise gems #878 [Nicholas Seckar]
* Changed .htaccess to allow dispatch.* to be called from a sub-directory as part of the push with Action Pack to make Rails work on non-vhost setups #826 [Nicholas Seckar/Tobias Luetke]
* Added new generator framework that informs about its doings on generation and enables updating and destruction of generated artifacts. See the new script/destroy and script/update for more details #487 [Jeremy Kemper]
* Fixed dependency reloading by switching to a remove_const approach where all Active Records, Active Record Observers, and Action Controllers are reloading by undefining their classes. This enables you to remove methods in all three types and see the change reflected immediately and it fixes #539. This also means that only those three types of classes will benefit from the const_missing and reloading approach. If you want other classes (like some in lib/) to reload, you must use require_dependency to do it.
* Added that ApplicationController will catch a ControllerNotFound exception if someone attempts to access a url pointing to an unexisting controller [Tobias Luetke]
