Overview
The Rubinius team is proud to announce the availability of 1.0.1.
1.0.1 is primarily a bugfix release, meant to address issues found in the 1.0.
We decided to do a 1.0.1 because there were some installations issues found in
1.0 that prevented some new users from using Rubinius. Additionally, a few
crash bugs were uncovered which we felt warranted a quick release.
As a bonus, this release includes fixes for running a number of popular gems.
The C-API support is extended to cover functionality required by popular gems
such as gherkin, curb, rjb/buildr, and many more.
Stats
- Commits: 199
- Days since last release: 23
Authors
- Brian Ford
- Dirkjan Bussink
- Evan Phoenix
- Ivan Samsonov
- Matthew Draper
- Mikko Perttunen
- Wilson Bilkovich
Summary
Critical
- Update StackVariable::last_match_ in the GC. Fixes #287.
- Set the stack start and size properly in a thread. @bugfix
- Revise GIL unlocking to keep Thread::saved_call_frame_ sane
- Handle very large values from getrlimit properly
- Add flag to disallow direct allocation. Fixes #335.
Fixes
- Fixed defined?(a && b). Closes #302.
- Fixed Fixnum and Bignum bit shifts.
- Imported lib/matrix.rb from 1.9 with many bug fixes. @bugfix
- Revert Matrix#** behavior for exception to 1.8 version. @bugfix
- Changed libgdtoa g_dfmt() to emit MRI format for minimal, unique String representation of Float.
- Added to_s_minimal primitive for Float#to_s.
- Use send to reach private extend_object methods. Closes #324.
- Don’t use a increment operation and the original value in the same expression
- Explicitly cast the return value of stack_pop() to void*
- Initial work to enable compiling with clang
- Use error state for read / write and print out the message
- Generalize jit generation for fixnum comparison operations
- Honor read-only-ness of certain globals. Fixes #283.
- Escape slashes in Regexp#to_s output. Fixes #308.
- Disambiguate a vcall send to the send_method opcode. Fixes #293.
- Evaluate next’s argument even though it’s ignored. Fixes #307.
- Report path to script bodies relatively. Fixes #308.
- Remove Numeric#==
- Use Kernel.exit! after a forked block returns. Fixes #289.
- Fix fork to allow SystemExit through properly
- Be persistant with write(2)
- Handle object_id’s for immediates. Fixes #315.
- Alter protocol for how Proc with a bound method is invoked. Fixes #277.
- Teach Signal.trap the rest of it’s tricks. Fixes #314.
- Make sure the arguments are Strings. @bugfix
- Cleanup and speed up Struct
- Add missing functionality to Signal.trap
- Introduce -Xprofile to be used instead of -P
- Cleanup and fix irb’s save history feature
- Search Object’s includes also. Fixes #328
- Make sure thet the default vis in a script body is private. Fixes #327
- Make STDOUT.reopen(path) effect the system fd. Fixes #322.
- Use the encoding of the Regexp in String#split on characters. Fixes #323
- Be sure to cleanup any temp modifications during #delete_if
- Let –enable-llvm do something. Fixes #336.
- Add flag support to new Dir.glob, always use it
- Handle write() setting EAGAIN properly
- Handle ECONNRESET like eof. Fixes #338
- Add UNIXSocket.pair
- Add Inliner::fixnum_le, and self recursion cuteness
- Simplify the Class#to_s of a metaclass (rails fix)
- Allow Modules to be dup’d, but not String
- Flip include order so that local LLVM wins
- Fix left shifting for fixnums
- Repair Bignum#divmod for cases with negative numbers. Fixes #319.
- Speed up String#[] and String#=~
- A Time is not eql? to a non-time.
- Raise the right exception when alias fails in a module.
- method_defined? shouldn’t search Object/Kernel.
- Print error messages to stderr instead of stdout.
- Ensure the proper Regexp constant is found when processing literals
- Introduce ‘push_rubinius’ instruction and use it like a rented mule
- Factor out some jit code from push_cpath_top and push_rubinius
- Quit as early as possible if we’re building on an unsupported platform
FFI
- Search all iconv FFI libs.
- Fix FFI::DynamicLibrary::Symbol#inspect
- Exclude info generation from all in libffi. Fixes #286.
- Add missing header files to scan for mode_t
- FFI API revamp, take 1.
C-API
- Added rb_ary_delete.
- Changed rb_set_errinfo to accept Qnil like 1.9.
- Added HAVE_STDARG_PROTOTYPES for C-API. Closes #300.
- Added OBJ_FREEZE for bluecloth gem.
- Added more C-API functions to get rjb to load.
- Only set status in rb_protect if it is not NULL.
- More C-API functions for DL.
- Added C-API HAVE_ macros for Hash.
- Moved C-API .h files to include directory.
- Moved C-API specs under spec/ruby/optional.
- Fixed rb_big2dbl.
- Add hacked rb_iterate/rb_each
- Add rb_reg_regcomp
- Annotate rb_raise as not returning to fix gherkin warnings. Fixes #309.
- Add stub of re.h
- Add missing symbols for curb
- Add simplicit rb_thread_blocking_region
- Provide builtin UBF macros
- Flesh out C-API to compile ImageMagick. Fixes #294.
- Make readline multithread aware and minor cleanup
- Check for async events when returning from a native method
- Check async events and raise an exception via rb_thread_select.
- Fix INT2NUM to actually take a long, not an int.
- Implement rb_is_{const,instance_class}_id() in CAPI.
- rb_protect’s callback takes a VALUE, not ANYARGS.
- Add apparently-missed prototype for rb_apply()
- Add rb_f_global_variables() and rb_str_new3().
- rb_define_module_under() shouldn’t return a parent’s module.
- Make rb_num2dbl handle all inputs correctly.
- Set rbconfig’s LIBS to “” to to appease mkrf
- Set rbconfig’s LIBRUBYARG_SHARED to “”, because we don’t have a shared library yet.
- Add NativeMethod functors up to a ten-argument version
New
- Added NKF ext from MRI 1.8.7. Closes #304.
- Imported DL extension from 1.8.7.
EXPERIMENTAL
- Rewire debugger infrastructure, add new simple debugger.
- Add defered breakpoints to the debugger
- Add backtrace and discover ability to QueryAgent
- Started on the console, pulled in our own copy of bert
- Revamp agent API, add a number of variables
- Get jit.inline.blocks running again
- Add Rubinius::Agent and agent loopback connection
- Add -Xjit.sync