Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Commits on Oct 8, 2011
  1. Evan Phoenix

    Sort out for in 1.9 mode

    evanphx authored
Commits on Oct 6, 2011
  1. Evan Phoenix

    Fix 1.9 parser crash

    evanphx authored
  2. Brian Shirai

    Fixed Symbol VM test in default 1.9 mode.

    brixen authored
    This one is for you Steve Jobs 1955-2011 RIP. :'( What an incredible
    gift life gave us.
Commits on Oct 5, 2011
  1. Brian Shirai
  2. Dirkjan Bussink
  3. Dirkjan Bussink
  4. Evan Phoenix

    Convert signal handling to use a normal Thread

    evanphx authored
    Previously the code would pass vm_ (ie, the main thread that the handler
    wakes up) into things like Channel::send as the state argument, which is
    not allowed.
    
    By creating a normal Thread, we'v got a unique VM* to use just for the
    handler.
  5. Evan Phoenix

    Guard exc since wakeup can GC

    evanphx authored
  6. Brian Shirai
  7. Brian Shirai
  8. Dirkjan Bussink

    In 1.9, re.h needs to be in the ruby/ directory.

    dbussink authored
    Fixes installation of the JSON gem in 1.9 mode
  9. Brian Shirai
  10. Brian Shirai

    Added Pathname from 1.9.2p290.

    brixen authored
  11. Brian Shirai

    Move 1.8 Pathname to lib/18.

    brixen authored
  12. Evan Phoenix
  13. Evan Phoenix
  14. Evan Phoenix

    Fix GC bug caused by layered GCLockGuard

    evanphx authored
    Channel::send in IOBuffer::create caused a GC would would cause the new
    IO* object to not be seen and be cleaned up, leaving a bad GC ref on the
    stack.
  15. Evan Phoenix
Commits on Oct 4, 2011
  1. Brian Shirai

    Merge pull request #1189 from yipdw/issue-1185

    brixen authored
    [PATCH] Zlib::GzipWriter#write segfaults when given large (2^23 byte) strings
  2. yipdw

    Allocate strings for FFI on the heap.

    yipdw authored
    Strings passed to foreign functions can be quite large -- too large to
    fit on the stack, and it's not always possible to detect an alloca(3)
    failure.
    
    Fixes #1185.
  3. yipdw

    Large-string example for Zlib::GzipWriter#write.

    yipdw authored
    The expected size was generated by the following Ruby program running
    under Ruby 1.8.7-p352:
    
        require 'stringio'
        require 'zlib'
    
        io = StringIO.new
        input = '.' * (2 ** 23)
    
        Zlib::GzipWriter.wrap(io) do |gzio|
          gzio.write(input)
        end
    
        print io.string
    
    This program was then run:
    
        $ ruby gen.rb | wc -c
        8176
    
    References #1185.
  4. Shane Becker
  5. Dirkjan Bussink

    Use Object if no class is given to rb_data_object_alloc

    dbussink authored
    MRI allows for creating a very strange object that has
    no klass pointer set. This means this object can't for example
    be inspected, since it will given the following result:
    
    NotImplementedError: method `class' called on terminated object (0xXXXXXXXX)
    
    In Rubinius we just use Object, so it at least works. It is
    definitely recommended to use a proper actual class to prevent
    potential issues on MRI.
    
    Fixes #1187
  6. Dirkjan Bussink
  7. Dirkjan Bussink

    Handle case where there isn't a call stack yet

    dbussink authored
    If you use call_til_compile=0 there is no call stack yet
    and it would crash.
    
    Fixes #1188
  8. Brian Shirai
  9. Brian Shirai

    Spec for Kernel.load on 1.9.

    brixen authored
  10. Brian Shirai
  11. Steve Klabnik

    Adding support for require_relative.

    steveklabnik authored
    This is based on my require_relative gem for MRI. Passes
    all of the RubySpec specs.
Commits on Oct 3, 2011
  1. Brian Shirai
  2. Brian Shirai
  3. Dirkjan Bussink
Commits on Sep 30, 2011
  1. Evan Phoenix
  2. Evan Phoenix
Commits on Sep 27, 2011
  1. Evan Phoenix

    Fix incorrect running of finalizers bug

    evanphx authored
    Baker::check_finalize checks if an object is forwarded to know if it has
    still alive or not therefore whether or not run it's finalizer.
    Therefore we can not check finalizers more than once because the 2nd
    time, we'll see objects that we have fixed up (and therefore are not
    forwarded) and finalizer them.
Something went wrong with that request. Please try again.