Permalink
Commits on Dec 5, 2016
  1. remove staging repo (for jruby-openssl) regenerate pom.xml

    kares committed Dec 5, 2016
Commits on Dec 2, 2016
  1. update jruby-openssl to 0.9.19

    kares committed Dec 2, 2016
Commits on Nov 24, 2016
  1. there's no need to clone backtrace array (JRuby 9K hasn't been doing so)

    kares committed Nov 24, 2016
Commits on Nov 21, 2016
  1. update jruby-openssl to 0.9.17

    kares committed Nov 20, 2016
Commits on Oct 26, 2016
  1. Merge pull request #3383 from kml/public_o_j_r_profile_builtin_Method…

    …Data
    
    public org.jruby.Runtime.Profile.Builtin::MethodData
    kares committed on GitHub Oct 26, 2016
Commits on Oct 18, 2016
  1. Merge pull request #4226 from iconara/fix-4218-1.7

    Don't dup every element in Enumerator#drop (1.7)
    headius committed on GitHub Oct 18, 2016
Commits on Oct 13, 2016
  1. Don't dup every element in Enumerator#drop

    In 06f0441 (JRUBY-6892) Enumerable#drop was changed to call #dup on every element that ended up in the result, but that might not work. Not all objects correctly implement #dup, or are even allocatable.
    
    #take was not changed, and looking at the differences between #take and #drop the only thing that stood out was that the signature of the #each call was different. Changing from Arity.NO_ARGUMENTS to Arity.ONE_REQUIRED make the issue that 06f0441 tried to fix go away.
    iconara committed Oct 13, 2016
Commits on Sep 14, 2016
  1. Don't define 1.9 stuff on 1.8.

    headius committed Sep 14, 2016
  2. Merge pull request #4149 from headius/backport_securerandom

    Backport securerandom
    headius committed on GitHub Sep 14, 2016
  3. Add a spec for #4155.

    headius committed Sep 14, 2016
  4. Remove remaining uses of WrapperMethod.

    headius committed Sep 14, 2016
  5. Stop using WrapperMethod, since it messes up super logic.

    This is for #3869 and relates to the module_function change from
    and redefine it with a new visibility and implementation class.
    However the impl class never passed through to the contained
    method, preventing it from being used in super. This affected, for
    example, module_fuction singleton methods that need to super or
    methods transplanted using defined_method with a Method instance.
    The new logic always tries to dup the target method so it can
    be truly populated with the altered fields. This change fixed
    
    The previous commit, using cloning instead of construction for
    IR methods, works around the fact that there's no semi-transparent
    WrapperMethod to delegate its serial number to the wrapped method.
    Since in that case and in this one, the method's serial number
    was expected to be the same after duplication, the clone
    technique seems acceptable.
    
    jruby-1_7: Fixes #4155
    headius committed Aug 22, 2016
Commits on Sep 13, 2016
  1. Incorporate remaining diffs from master.

    Only diff is calling RubyArray.newArrayNoCopy instead of the
    9k-only newArrayMayCopy.
    headius committed Sep 13, 2016
  2. Synchronize all of Etc because UNIX is dumb. Fixes #4057.

    This also incorporates diffs from master, aligning both
    Etc implementations.
    headius committed Sep 13, 2016
  3. Provide a help string that makes sense for PRNG option.

    headius committed Sep 13, 2016
Commits on Sep 12, 2016
  1. Actually settle on SHA1PRNG if it's available. #1405

    I botched the previous patch a bit by unconditionally re-assigning
    the secureRandom local to a default JDK new SecureRandom. This
    could cause systems without the default preferred PRNG
    (NativePRNGNonBlocking, Java 8+) to have slower thread startup and/or
    random number generation.
    headius committed Sep 12, 2016
  2. avoid eager secure random initialization on every new Ruby thread all…

    …ocation
    
    shaves off 10-15% of Thread.new/start :
    
    ```
    Rehearsal ----------------------------------------------------------
    Thread.new [100000x]    12.800000   6.850000  19.650000 ( 10.392600)
    Thread.start [100000x]  12.670000   5.680000  18.350000 (  9.625846)
    ------------------------------------------------ total: 38.000000sec
    
                                 user     system      total        real
    Thread.new [100000x]    12.780000   5.760000  18.540000 (  9.620000)
    Thread.start [100000x]  12.490000   5.810000  18.300000 (  9.524975)
    ```
    
    ```
    Rehearsal ----------------------------------------------------------
    Thread.new [100000x]    11.040000   5.560000  16.600000 (  8.708792)
    Thread.start [100000x]  10.740000   5.520000  16.260000 (  8.400990)
    ------------------------------------------------ total: 32.860000sec
    
                                 user     system      total        real
    Thread.new [100000x]    10.840000   5.620000  16.460000 (  8.506339)
    Thread.start [100000x]  10.460000   5.560000  16.020000 (  8.323819)
    ```
    kares committed with headius Mar 10, 2016
Commits on Aug 29, 2016
  1. Update for next dev cycle

    enebo committed Aug 29, 2016
Commits on Aug 26, 2016
  1. Bump version to 1.7.26

    enebo committed Aug 26, 2016
  2. Merge pull request #4115 from jellymann/fix-stdin-bug-1_7

    Backport "Change script file restriction" from 9000
    kares committed on GitHub Aug 26, 2016
Commits on Aug 25, 2016
  1. Backport Change script file restriction from 9000

    jellymann committed Aug 25, 2016
Commits on Aug 24, 2016
  1. Revert "update getConstant lookup to avoid auto-loading constant on a…

    … const_defined? check"
    
    This reverts commit 5cdaa48.
    
    See #3920.
    headius committed Aug 24, 2016
  2. Super-gross hack to catch Java 9's InaccessibleObjectException.

    This new exception is raised by Jigsaw when you attempt to make
    accessible the inaccessible elements of a module's classes.
    Unfortunately, since the exception doesn't exist on Java 8,
    there's no way to catch it (in code that needs to build on Java 8)
    except for this hack.
    
    This allows `jruby -e 1` to run properly on Java 9.
    
    This fixes part of #4111.
    headius committed Aug 24, 2016
  3. Make sure we are green with this bugger excluded on appveyor

    enebo committed Aug 24, 2016
  4. Implement warn/warning in ripper

    enebo committed Aug 24, 2016
  5. update getConstant lookup to avoid auto-loading constant on a const_d…

    …efined? check
    kares committed with headius Jan 12, 2016
  6. Hopefully fix appveyor for good on 1.7 getting maven+ant

    enebo committed Aug 24, 2016
Commits on Aug 23, 2016
  1. Merge pull request #3927 from ahorek/backport_to_r_fix

    backport BigDecimal .to_r for #3728
    headius committed on GitHub Aug 23, 2016
Commits on Aug 18, 2016
  1. use the 1.8 way of making a float so we match on error for bad types …

    …in Math.log10
    enebo committed Aug 18, 2016
  2. Fixes #4089. Math.log10 should use Java's log10 method

    enebo committed Aug 18, 2016
Commits on Aug 16, 2016
  1. Merge pull request #4062 from camlow325/add-debug-log-messages-to-jru…

    …byclassloader-find-class
    
    Write a log message when findClass IOException is caught
    kares committed on GitHub Aug 16, 2016
Commits on Aug 11, 2016
  1. Use absolute classpath URL when loading jar-bootstrap. See #4000.

    headius committed Aug 11, 2016
  2. Synchronize around getgrgid call. Fixes #4057.

    getgrgid is thread-safe on OS X, but other platforms are not
    required to be thread-safe by the POSIX specification. It appears
    that at least some flavors of Linux do not have thread-safe
    getgrgid. To aid users who might be calling it across threads, we
    lock around the native getgrgid call.
    
    I chose to lock around the RubyEtc.class to ensure the native call
    is single-threaded for all JRuby instances in a given classloader.
    There's no clean way to make that guarantee across classloaders.
    headius committed Aug 11, 2016