Skip to content
Commits on Feb 13, 2016
  1. @hoelzro
  2. @hoelzro

    Allow users to disable multi-line input

    hoelzro committed
    Via RAKUDO_DISABLE_MULTILINE
  3. @hoelzro

    Change is-interactive to multi-line-enabled

    hoelzro committed
    For an upcoming change, multi-line input may not always be
    active when input is interactive, so I didn't want them to be
    viewed as synonymous
  4. @hoelzro

    Multi-line input: Don't segfault if an exception has no payload

    hoelzro committed
    We extract BOOTException's payloads via nqp::getpayload to detect
    X::Syntax::Missing errors; sometimes, an exception has no payload,
    which (until now) causes a segmentation fault
  5. @hoelzro

    Only turn on multi-line behavior if we're interactive

    hoelzro committed
    Multi-line behavior isn't needed for one-liners or scripts
  6. @hoelzro

    Override HLL::Compiler.eval for multi-line input

    hoelzro committed
    The way this works is pretty dodgy, but it's good enough for
    my experiment.  How this works is that we catch any exceptions
    from evaluating the code; if there's an exception, it's an
    X::Syntax::Missing, and it occurs at the end of the input string,
    that means that more input is required.  Instead of just throwing
    the exception at the user, we signal HLL::Compiler that we need
    more input via self.needs-more-input
  7. @hoelzro

    Add pos attribute to X::Syntax

    hoelzro committed
  8. @hoelzro
  9. @lizmat

    Remove superfluous line, sortiz++

    lizmat committed
Commits on Feb 12, 2016
  1. @lizmat

    Merge pull request #711 from salortiz/setty_hash

    lizmat committed
    Fix for RT#127402: Set.hash stringifies its objects
  2. @salortiz
Commits on Feb 11, 2016
  1. @lizmat
  2. @lizmat

    Remove hash- prefix from MappyIterator attributes

    lizmat committed
    For no other reason I got sick of them  :-)
  3. @lizmat
  4. @lizmat

    Make sure Failure.handled looks like a Bool

    lizmat committed
    Even though it's a native int internally on MoarVM.  b2gills++, sortiz++
  5. @lizmat

    Merge pull request #709 from Skarsnik/nom

    lizmat committed
    Change the orders of arguments documentation for MAIN display.
  6. @Skarsnik

    Change the orders of arguments documentation for MAIN display. Previo…

    Skarsnik committed
    …usly it was "random" (hash key order) now it the order on wich they are written
  7. @lizmat

    Merge pull request #693 from salortiz/failure.handled

    lizmat committed
    Implement Failure.handled as documented
  8. @hoelzro

    Install Rakudo jars in proper location

    hoelzro committed
    Stuff written in NQP should go under $prefix/share/nqp/lib (J_LIBPATH),
    and stuff written in Perl 6 should go under
    $prefix/share/perl6/runtime/ (PERL6_LANG_DIR), since that's where the
    respective module loaders look for it
  9. @timo
Commits on Feb 10, 2016
  1. @lizmat
  2. @lizmat
  3. @lizmat
  4. @lizmat

    Improve Int %h by 20% by removing code

    lizmat committed
    Why?  Because the candidates that handle the type constrained
    values Hash, were less optimised.  But the only difference really
    was how they tried to get at the Hash descriptor (which is the
    only thing that really separates a type constrained Hash from a
    "normal" Hash).  So removing the code, will revert back to the
    better optimised un-constrained candidates, and thus get better
    performance.
  5. @lizmat

    Fix for RT #127504

    lizmat committed
Commits on Feb 9, 2016
  1. @lizmat

    Refactor object hashes completely

    lizmat committed
    Before:
      Hash[Any,Any]
        |-- $!keys
        |     |-- object
        \-- $!storage
              |-- value
    
    After:
      Hash[Any,Any]
        \-- $!storage
              \-- Pair.new(object,value)
    
    Sample benchmarks showed about 10% less memory usage, and a few percent
    less CPU usage.  More importantly, methods such as kv, pairs, antipairs
    have become much simpler and faster, whereas methods such as keys/values
    didn't suffer much
    
    This should also make it easier to subclass Bags/Mixes from object hashes.
  2. @lizmat

    Scrape a few % off of Hash.DELETE-KEY

    lizmat committed
    - don't bother with the :SINK candidate, the current overhead kills it
    - don't bother allocating storage if there is none yet
    - don't bother getting value if the key isn't there
  3. @lizmat

    Bump NQP_REVISION

    lizmat committed
Commits on Feb 8, 2016
  1. @coke @niner

    mention point release

    coke committed with niner
  2. @coke @niner

    [release] bump VERSION

    coke committed with niner
  3. @coke @niner

    moar specific google

    coke committed with niner
  4. @coke @niner

    update copyright

    coke committed with niner
  5. @coke @niner

    actual release date

    coke committed with niner
  6. @coke @niner

    Thank author of last minute commit

    coke committed with niner
  7. @coke @niner

    [release] bump VERSION

    coke committed with niner
Something went wrong with that request. Please try again.