Commits on Oct 5, 2012
  1. nci/libffi.c: check FLOATTYPE, not NUMVAL_SIZE

    NUMVAL_SIZE is not good enough, esp. wrong for x86_64
    rurban committed Oct 5, 2012
  2. Merge remote-tracking branch 'origin/master' into native_pbc2

    Fixed conflicts:
    rurban committed Oct 5, 2012
  3. Merge pull request #857 from jjuran/has-thread

    config/auto/thread:  Correct "has thread" grammar
    jkeenan committed Oct 5, 2012
Commits on Oct 4, 2012
Commits on Oct 3, 2012
  1. [cage] silence @noinline@ warning on unknown compilers

    e.g. solaris cc
    rurban committed Oct 3, 2012
  2. [GH #855] utf8 comparison problem

    rurban committed Oct 3, 2012
  3. [GH #854] ChangeLog line

    rurban committed Oct 3, 2012
  4. Fix a bug that causes parrot to fail on platforms where size_t is an …

    …unsigned type.
    Signed-off-by: Reini Urban <>
    Martin Vorländer committed with rurban Oct 3, 2012
Commits on Oct 2, 2012
  1. [GH #813, #814] Fix tests for big-endian

    Also check for has_icu to check with composed unicode strings
    rurban committed Oct 2, 2012
  2. Revised [GH #813 + #814] ChangeLog

    rurban committed Oct 2, 2012
  3. Merge branch 'master' into rurban/mime-base64-utf8-gh813+gh814

    Fixed merge conflicts:
    rurban committed Oct 2, 2012
  4. Fix t/steps/auto/alignof-01.t warning

    Thanks to jkeenan++ for adding the test.
    rurban committed Oct 2, 2012
  5. Add minimal tests for auto::alignof.

    Configuration step auto::alignof was added to master without a corresponding
    test file under t/steps.  Add that file; update MANIFEST.  Make indentation in
    config/auto/ consistent.
    Note: Running the newly added steps test shows that internal subroutine
    test_if_needed throws a warning, at least when we are not trying to compile
    with clang.  The warning is:
      value for '@TEMP_type@' in config/auto/alignof/ is undef at
        lib/Parrot/Configure/ line 584, <$in> line 55.
    This indicates a possible problem with config/auto/alignof/
    jkeenan committed Oct 2, 2012
  6. '1' added to result message is superfluous; revert.

    The second argument passed to auto::llvm::_handle_result() is merely a Boolean
    used to determine whether the result message displayed as part of's output should be set to 'yes' or 'no'.  Consequently, a result
    of 'yes, 1' conveys nothing above or beyond 'yes'.  Reverting the modification
    (which was also causing failures in t/steps/auto/llvm-01.t).
    jkeenan committed Oct 2, 2012
Commits on Oct 1, 2012
  1. [GH #813 + #814] Use Bytebuffer for MIME::Base64, add 2nd enc arg to …

    Use bytebuffer representations of the encoded string, not the encoded ord value.
    Also fix the tests to match this conformant behaviour.
    The problem is now that base64 encoded files are endian dependent, and the multibyte
    tests need to be skipped on big-endian.
    rurban committed Oct 1, 2012
  2. [tools] better string output for multi-byte

    Use errors=replace instead of strict. See
    Prepend encoding when not ascii and not latin1, add strlen and bufused fields.
    rurban committed Oct 1, 2012
  3. [cage] Improve llvm detection

    Probe for llvm-config-3.0 llvm-config-2.9 llvm-config-2.8 which do exist on debian,
    where llvm-config does not exist.
    Print found version, like "yes, 3.0" as with gcc and other libs.
    If not requested, print "not requested" instead of "no" to give a hint,
    that --with-llvm is wanted.
    Add llvm and inline probes to ChangeLog
    rurban committed Oct 1, 2012
  4. [cage] fix auto::inline for g++

    add missing return type to function
    error: ISO C++ forbids declaration of ‘f’ with no type [-fpermissive]
    rurban committed Oct 1, 2012
  5. remove hard tabs

    bubaflub committed with rurban Sep 30, 2012
  6. Revise [tools] Support more encodings

    Use native endianness
    rurban committed Sep 30, 2012
  7. [tools] Support more encodings

    Also fix a python gdb strlen error. The number of char is needed, not the number of codepoints.
    rurban committed Sep 30, 2012
  8. On second read through, it doesn't appear that the old IO system had …

    …a write buffer on stdout. Take that away in the new system to try and fix some rakudo issues reported by pmichaud++
    Whiteknight committed with rurban Sep 28, 2012
  9. Undo a prospective fix from rurban++. We don't want to flush stdout e…

    …very time we read from any filehandle.
    Whiteknight committed with rurban Sep 28, 2012