## 1.1.3 * The `rails runner` command no longer passes environment switches to files which it runs. Issue #272. * Various issues solved to do with termination / processes hanging around longer than they should. Issue #290. ## 1.1.2 * Detect old binstubs generated with Spring 1.0 and exit with an error. This prevents a situation where you can get stuck in an infinite loop of spring invocations. * Avoid `warning: already initialized constant APP_PATH` when running rails commands that do not use spring (e.g. `bin/rails server` would emit this when you ^C to exit) * Fix `reload!` in rails console * Don't connect/disconnect the database if there are no connections configured. Issue #256. ## 1.1.1 * Fix `$0` so that it is no longer prefixed with "spring ", as doing this cause issues with rspec when running just `rspec` with no arguments. * Ensure we're always connected to a tty when preloading the application in the background, in order to avoid loading issues with readline + libedit which affected pry-rails. ## 1.1.0 * A `bin/spring` binstub is now generated. This allows us to load spring correctly if you have it installed locally with a `BUNDLE_PATH`, so it's no longer necessary to install spring system-wide. We also activate the correct version from your Gemfile.lock. Note that you still can't have spring in your Gemfile as a git repository or local path; it must be a proper gem. * Various changes to how springified binstubs are implemented. Existing binstubs will continue to work, but it's recommended to run `spring binstub` again to upgrade them to the new format. * `spring binstub --remove` option added for removing spring from binstubs. This won't work unless you have upgraded your binstubs to the new format. * `config/database.yml` is watched * Better application restarts - if you introduce an error, for example by editing `config/application.rb`, spring will now continue to watch your files and will immediately try to restart the application when you edit `config/application.rb` again (hopefully to correct the error). This means that by the time you come to run a command the application may well already be running. * Gemfile changes are now gracefully handled. Previously they would cause spring to simply quit, meaning that you'd incur the full startup penalty on the next run. Now spring doesn't quit, and will try to load up your new bundle in the background. * Fix support for using spring with Rails engines/plugins ## 1.0.0 * Enterprise ready secret sauce added ## 0.9.2 * Bugfix: environment variables set by bundler (`BUNDLE_GEMFILE`, `RUBYOPT`, etc...) were being removed from the environment. * Ensure we only run the code reloader when files have actually changed. This issue became more prominent with Rails 4, since Rails 4 will now reload routes whenever the code is reloaded (see https://github.com/rails/rails/commit/b9b06daa915fdc4d11e8cfe11a7175e5cd8f104f). * Allow spring to be used in a descendant directory of the application root * Use the system tmpdir for our temporary files. Previously we used `APP_ROOT/tmp/spring`, which caused problems on filesystems which did not support sockets, and also caused problems if `APP_ROOT` was sufficiently deep in the filesystem to exhaust the operating system's socket name limit. Hence we had a `SPRING_TMP_PATH` environment variable for configuration. We now use `/tmp/spring/[md5(APP_ROOT)]` for the socket and `/tmp/spring/[md5(APP_ROOT)].pid` for the pid file. Thanks @Kriechi for the suggestion. Setting `SPRING_TMP_PATH` no longer has any effect. ## 0.9.1 * Environment variables which were created during application startup are no longer overwritten. * Support for generating multiple binstubs at once. Use --all to generate all, otherwise you can pass multiple command names to the binstub command. * The `testunit` command has been extracted to the `spring-commands-testunit` gem, because it's not necessary in Rails 4, where you can just run `rake test path/to/test`. * The `~/.spring.rb` config file is loaded before bundler, so it's a good place to require extra commands which you want to use in all projects, without having to add those commands to the Gemfile of each individual project. * Any gems in the bundle with names which start with "spring-commands-" are now autoloaded. This makes it less faffy to add additional commands. ## 0.9.0 * Display spring version in the help message * Remove workaround for Rubygems performance issue. This issue is solved with Rubygems 2.1, so we no longer need to generate a "spring" binstub file. We warn users if they are not taking advantage of the Rubygems perf fix (e.g. if they are not on 2.1, or haven't run `gem pristine --all`). To upgrade, delete your `bin/spring` and re-run `spring binstub` for each of your binstubs. * Binstubs now fall back to non-spring execution of a command if the spring gem is not present. This might be useful for production environments. * The ENV will be replaced on each run to match the ENV which exists when the spring command is actually run (rather than the ENV which exists when spring first starts). * Specifying the rails env after the rake command (e.g. `rake RAILS_ENV=test db:migrate`) now works as expected. * Provide an explicit way to set the environment to use when running `rake` on its own. * The `rspec` and `cucumber` commands are no longer shipped by default. They've been moved to the `spring-commands-rspec` and `spring-commands-cucumber` gems. ## 0.0.11 * Added the `rails destroy` command. * Global config file in `~/.spring.rb` * Added logging for debugging. Specify a log file with the `SPRING_LOG` environment variable. * Fix hang on "Run `bundle install` to install missing gems" * Added hack to make backtraces generated when running a command quieter (by stripping out all of the lines relating to spring) * Rails 4 is officially supported ## 0.0.10 * Added `Spring.watch_method=` configuration option to switch between polling and the `listen` gem. Previously, we used the `listen` gem if it was available, but this makes the option explicit. Set `Spring.watch_method = :listen` to use the listen gem. * Fallback when Process.fork is not available. In such cases, the user will not receive the speedup that Spring provides, but won't receive an error either. * Don't preload `test_helper` or `spec_helper` by default. This was causing people subtle problems (for example see #113) and is perhaps surprising behaviour. It may be desirable but it depends on the application, therefore we suggest it to people in the README but no longer do it by default. * Don't stay connected to database in the application processes. There's no need to keep a connection open. * Avoid using the database in the application processes. Previously, reloading the autoloaded constants would inadvertantly cause a connection to the database, which would then prevent tasks like db:create from running (because at that point the database doesn't exist) * Removed ability to specify list of files for a command to preload. We weren't using this any more internally, and this is easy to do by placing requires in suitable locations in the Rails boot process (which is not explained in the README). * Seed the random number generator on each run. ## 0.0.9 * Added `Spring::Commands::Rake.environment_matchers` for matching rake tasks to specific environments. * Kill the spring server when the `Gemfile` or `Gemfile.lock` is changed. This forces a new server to boot up on the next run, which ensures that you get the correct gems (or the correct error message from bundler if you have forgotten to `bundle install`.) * Fixed error when `Spring.watch` is used in `config/spring.rb` ## 0.0.8 * Renamed `spring test` to `spring testunit`. * Implemented `spring rails` to replace `spring [console|runner|generate]`. * `config/spring.rb` is only loaded in the server process, so you can require stuff from other gems there without performance implications. * File watcher no longer pays attention to files outside of your application root directory. * You can use the `listen` gem for less CPU intensive file watching. See README.