A series of commits relating to preloading and scopes caused a regression. Cloning the relation calls initialize_copy which resets a number of instance variables to nil. Without this the scope thinks that it is already loaded when it is called again. Reverts the following commits: 13f1401 8491740 dffbb52 Fixes #6575, #6576 & #6577
…m 'therubyracer'. Dependency.rb expects the symbol to be named :platforms as opposed to platform. RubyMine's inspections indicate that the symbol should be named :platforms. Updating tests.
This reverts commit ebfa58a. This change caused a regression in my application's tests: * Failing build http://travis-ci.org/#!/resolve/refinerycms/builds/1461242 Broken tests pass after this change is reverted.
Thanks to Ben Murphy for reporting this CVE-2012-2661
Conflicts: activesupport/lib/active_support/json/encoding.rb activesupport/test/json/encoding_test.rb
This is a back-port of rails/rails#6525. See the commit notes there for details.
railites_order method, introduced in 40b19e0, had a bug that was causing loading application instance twice in initializers if railties_order already included application instance. So for example railties_order = [Foo::Engine, :main_app, Bar::Engine] would result in such railties array: [MyApp::Application, Foo::Engine, MyAppApplication, Bar::Engine] In order to fix it, we need to check for existence of application in both railties_order and railties arrays.
…issues [#6464] Fixes #6464 Synchronize the contents of the release method in ConnectionPool due to errors when running in high concurrency environments. Detected invalid hash contents due to unsynchronized modifications with concurrent users org/jruby/RubyHash.java:1356:in `keys' /usr/local/rvm/gems/jruby-1.6.7@new_import/gems/activerecord-3.2.3/lib/a ctive_record/connection_adapters/abstract/connection_pool.rb:294:in `release' /usr/local/rvm/gems/jruby-1.6.7@new_import/gems/activerecord-3.2.3/lib/a ctive_record/connection_adapters/abstract/connection_pool.rb:282:in `checkin'