Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Commits on Sep 23, 2011
  1. @headius
  2. @headius
  3. Add setNativeHandle/getNativeHandle to RubyBasicObject and use that t…

    Wayne Meissner authored
    …o store/retrieve the native cext handle.
  4. Mark children of T_DATA and T_ARRAY objects from rb_gc_mark(), to be …

    Wayne Meissner authored
    …more explicit about it - both were handled incidentally previously.
  5. @headius
  6. @headius

    Move the slowest tests out of JRubyTestSuite into SlowTestSuite, to s…

    headius authored
    …peed up the day-to-day run a bit.
  7. @headius

    Encoding fixes for JRUBY-6033:

    headius authored
    * String#% was not setting base encoding of new string to same as format string, as in MRI
    * Array#to_s and Array#inspect did not use encoding-aware methods for building the string
    * Hash#inspect ditto
Commits on Sep 22, 2011
  1. @headius

    Hopefully fix jruby-base CI failure by fixing a command line switch t…

    headius authored
    …est that expects in-process launching.
  2. @headius
  3. @headius

    Move RString and RArray native handles to varTable.

    headius authored
    RString and RArray are cached in RubyString and RubyArray to avoid weak-caching
    handles or recreating handles frequently. However these fields add an additional
    32 or 64 bits to the size of all String and Array instances even when C exts are
    not in use. Since C ext use on JRuby is very uncommon, I've moved these fields
    into the lazily-allocated variable table, in the same manner as the rarely-used
    object_id. This reduces the size and allocation cost of all String and Array
    instances for the good of all JRubyists.
    
    The cost of retrieving the handles increases slightly due to hopping through
    getMetaClass.getRealClass, the accessor, and the array, but that cost likely
    pales in comparison to the already-high cost of C extensions, and only impacts
    Array and String instances that require native handles.
Commits on Sep 21, 2011
  1. @headius

    Turn off in-process launching by default.

    headius authored
    This feature was originally intended to blunt the high startup cost of JRuby
    by launching subprocess commands that look like "ruby" commands inside the
    same JVM using a new JRuby instance. While this worked, it also led to lots
    of peculiar artifacts, like false pids, odd stream behavior, and runnaway
    "subprocess" threads that would have died naturally if run as a new process.
    
    Because of these issues, and because JRuby's base startup time has improved
    greatly since the feature was added, I'm finally turning off in-process
    launching by default. We'll see how it goes.
    
    Possible problems that will arise:
    
    * Any subprocess command that literally uses "ruby" as the program to run
      will fail unless "ruby" has been aliased to "jruby" as in RVM. People may
      see those subprocesses start incorrectly launching the system "ruby"
      rather than JRuby if they don't use rvm.
    * Any tools or scripts that launch many subprocess "ruby" commands may slow
      down somewhat. They will be launching new JVM processes and those JVMs will
      take time to warm up.
  2. @headius

    Cext fixes:

    headius authored
    * rb_class_of should return immediate metaclass, not Ruby .class metaclass
    * Better error message for multi-runtime collision
    * Reorg jruby namespace fields a bit
Commits on Sep 20, 2011
  1. @enebo

    JRUBY-6073: "\x80".index("\x80")

    enebo authored
  2. @headius
  3. @headius
  4. @BanzaiMan
  5. @BanzaiMan
  6. @headius @BanzaiMan

    Up delay in MRI's test_signal.rb in hopes of making the signal more r…

    headius authored BanzaiMan committed
    …eliably fire within the assert_raises block.
  7. @BanzaiMan
  8. @subbuss @BanzaiMan
  9. @subbuss @BanzaiMan

    [IR] Fixed IR generation bug for OpElementAsgnNode

    subbuss authored BanzaiMan committed
  10. @subbuss @BanzaiMan

    [IR] Raise LocalJumpError for bad redo statements

    subbuss authored BanzaiMan committed
  11. @subbuss @BanzaiMan

    [IR] Turn off binding load/store pass since it is buggy and it is not…

    subbuss authored BanzaiMan committed
    … needed for the IR interpreter
  12. @subbuss @BanzaiMan

    [IR] Update IR spec file

    subbuss authored BanzaiMan committed
  13. @subbuss @BanzaiMan
  14. @enebo @BanzaiMan

    never ever set your parent to yourself (you don't want to be your own…

    enebo authored BanzaiMan committed
    … father do ya?)
  15. @enebo @BanzaiMan

    Add a non-array variant to enumerable bench

    enebo authored BanzaiMan committed
  16. @BanzaiMan
Commits on Sep 19, 2011
  1. @subbuss
  2. @subbuss
  3. @subbuss
  4. @subbuss

    [IR] Turn off binding load/store pass since it is buggy and it is not…

    subbuss authored
    … needed for the IR interpreter
  5. @subbuss

    [IR] Update IR spec file

    subbuss authored
  6. @subbuss
  7. @enebo
Something went wrong with that request. Please try again.