JRuby 9000 is the new major version of JRuby, representing years of effort and large-scale reboots of several JRuby subsystems.

Major features of JRuby 9000:

• Ruby 2.x compatibility
• A new optimizing runtime based on a traditional compiler design
• New POSIX-friendly IO and Process
• Fully ported encoding/transcoding logic from MRI

• Ruby 2.3 Compatible (language and standard library) (i)
• Fixed performance and behavior regressions from 9.0.5.0
• Added several JIT optimizations missed in initial 9k release
• Fixed most of the outstanding parser bugs
• RubyGems updated to 2.6.4
• Many improvements to our Java Integration
• 114 issues fixed for 9.1.0.0

(i) We have a large majority of Ruby 2.3 features but we are not completely done. You can follow our remaining work by tracking this issue.

• removing default -Xmx (500MB) setting from jruby.bash. Most users that get JRuby will end up using the bash script. https://github.com/jruby/jruby/issues/3739

• Thread.new and others spawning threads (Fiber, Enumerator) should be a bit faster to boot, most importantly they no longer possibly slow-down due low entropy sources on *nix systems due a Java SecureRandom instantiation on JRuby's ThreadContext. https://github.com/jruby/jruby/pull/3723

• Java stack-trace filtering was tuned to not exclude everything under the org.jruby package prefix, we still avoid the additional noise from the stack-trace but no longer filter potentially unknown packages (e.g. org.jruby.rack) or extension stacks (org.jruby.ext). This change also affects backtrace information on the Ruby side which might now include more .java parts.

• the internal JavaPackageModuleTemplate has been refactored into a standalone module-like class, all package stubs org.kares.jruby now return a JavaPackage instance (which is a Module)

• names throw and catch are no longer allowed as package names using the method format (org.jruby.catch will now dispatch to Kernel.catch) - those are not valid Java package names anyways

• singleton_class (as well as hook methods such as singleton_method_added, singleton_method_removed) is no longer allowed as a package-name using the method ('.') format (org.jruby.singleton_class will return the Ruby meta-class)

• pkg.to_s now returns package-name opposed to "module name" e.g. java.lang.to_s == 'java.lang' (note that inspect still behaves the same as before: java.lang.inspect == 'Java::JavaLang')

• improved java.util.Map equality (==, eql?) to other map/hash instances - previously wasn't 100% correct

• java.util.Map proxies now support all of Hash methods (including new ones such as dig, fetch_values and comparison operators e.g. <=)

• Java arrays now handle value equality == with Ruby arrays (use eql? if types need to match as well). this also affects === with native arrays.

• Java's exception hierarchy java.lang.Throwable no longer provides a to_str method, we feel like this unintentionally slipped probably due misunderstanding the difference between Ruby's to_s and to_str

• java.util.Collection instances are no longer to_ary (Array-like) convertible, please note that to_a still exists and (for compatibility) java.util.List provides to_ary.

• unified java.lang.Comparable#compareTo with incompatible types now raising TypeError

• java.util.Iterable#each (also java.util.Collection#each) now return an Enumerator when no block given

• java.util.Iterable#each_with_index returns Enumerator when no block given

• java.util.List#rindex now using listIterator to loop backwards instead of get(index)

• java.util.List#sort dup-s - returns the same type as target (previously always returned an ArrayList)

• java.util.List#[] extension works as with Array (no Java index out of bounds errors)

• java.util.List#[]= extension behaves like with an Array (growing the list as needed) also element removal should be more efficient across different kinds of list implementations

• Ruby's dup and clone methods are expected to be working for most Collection types (except internal impl classes such as ones returned from java.util.Collections factories)

• (native) Java arrays now correctly dup

• added java.util.regex.Pattern#casefold?

• java.util.regex.Matcher is expected to quack very much like a Ruby MatchData

• java.util.regex.Matcher supports named groups with begin/end (available on Java 8+)

• JRuby now provides ruby_xxx names for potentially conflicting methods e.g. java.util.List#sort on Java 8 java.util.ArrayList.class_eval { alias sort ruby_sort } will make all array lists sort the Ruby way while the instance is being used in Ruby scripts.

• Java interfaces no longer provide the (useless) new method (e.g. java.lang.Runnable.new)

• method arities for generated Java methods (on Ruby class' become_java!) have been improved so that when a Ruby method has fixed arity (e.g. foo(arg)) it will map to a correct Java method signature (foo(IRubyObject arg)) instead of always generating a generic var-args (IRubyObject[] args) method.

• Warbler has had issues with pre-compiled .rb files due broken IR de-serialization logic, we expect all issues to be fixed and added specs to cover previously failing issues.

• jrubyc --jdk5 and -5 switches were removed (Java 5 has not been supported for a while)

• jrubyc --dir option now handles absolute paths correctly with the --target option