Permalink
Commits on Oct 25, 2011
  1. Revert "Refactor and cleanup RubyArgsFile."

    headius committed Oct 25, 2011
    This reverts commit a70668d.
  2. Use normal LoadService.load for booting Ruby kernel, so security mana…

    headius committed Oct 25, 2011
    …ger doesn't complain.
Commits on Oct 24, 2011
  1. Fix #parameters and #source_location behavior on curried procs (degra…

    headius committed Oct 24, 2011
    …de to MRI behavior).
  2. Build damn you

    enebo committed Oct 24, 2011
  3. profiler:spec rake task and make sure it's run with rake_targets

    kares committed with BanzaiMan Oct 20, 2011
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  4. make sure profiler specs (graph_profile_printer_spec & profile_data_s…

    kares committed with BanzaiMan Oct 18, 2011
    …pec) are green
    
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  5. Move Signal into Ruby kernel.

    headius committed Oct 24, 2011
Commits on Oct 23, 2011
  1. JRUBY-6141 - added RubySpec test.

    jpace committed with BanzaiMan Oct 23, 2011
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  2. JRUBY-6141 - fixed so that MatchData are cloned, so that references t…

    jpace committed with BanzaiMan Oct 22, 2011
    …o MatchData are not updated with subsequent matches.
    
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
Commits on Oct 21, 2011
  1. Purge invalid tags.

    headius committed Oct 21, 2011
  2. Prevent cache leakage in extreme method churn situations.

    headius committed Oct 21, 2011
    While working on a support ticket, it came to my attention that Rails was
    generating lots of new methods for templates due to some gap in caching.
    These methods were generated with a new name each time, using the __id__
    of the Template object that would eventually manage them.
    
    Because our per-class caching did not actively flush (only passively,
    by invalidating each class and letting subsequent calls recache), old
    method names never encountered again would stick around in the cache.
    This led to a leak, as more and more new method names were created,
    called, and then abandoned.
    
    By actively invaldating the per-class cache in response to an invalidation
    event, we avoid the leak. It does not appear to add significant overhead
    to the invalidation process.
Commits on Oct 20, 2011
  1. [IR] Fixed one more zsuper bug dealing with super getting access to

    subbuss committed Oct 20, 2011
    unnamed rest args of a method when nested in a block; some code
    cleanup.
  2. Revert "pack_sockaddr_in should convert a port string"

    BanzaiMan committed Oct 20, 2011
    Ooops. This introduced compile error.
    
    This reverts commit 18a02a9.
  3. use the latest version of the gemify-maven-plugin. fixed a typo in va…

    mkristian committed with BanzaiMan Oct 15, 2011
    …riable when using custom settings.
    
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  4. allow relative path in the PATH env

    mkristian committed with BanzaiMan Oct 14, 2011
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  5. pack_sockaddr_in should convert a port string

    geemus committed with BanzaiMan Oct 3, 2011
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  6. Fix for JRUBY-6031:

    ajuckel committed with BanzaiMan Sep 27, 2011
    When sending non-UTF-8 character strings into RubyString#getUnicodeValue,
    the decoder was bailing out upon MalformedInput, but not communicating
    that failure further up the stack.  Therefore, character strings were
    getting silently truncated when going through ARJDBC.  This fix simply
    sets CodingErrorAction.REPLACE instead, which means the entire string
    will be returned, though with replacements for the malformed input.
    
    Signed-off-by: Hiro Asari <asari.ruby@gmail.com>
  7. These are still failing

    BanzaiMan committed Oct 20, 2011