Commits on Oct 3, 2012
  1. Cleaned up at_exit spec.

    brixen committed Oct 3, 2012
  2. Merge pull request #1898 from voxik/ignore-block-device-test

    Ignore block device test.
    Brian Ford committed Oct 3, 2012
  3. Merge pull request #1936 from ryoqun/raise-nil-false

    Correctly raise TypeError even if nil or false is given
    Brian Ford committed Oct 3, 2012
  4. Merge pull request #1937 from voxik/install-manpages

    Install man pages.
    Brian Ford committed Oct 3, 2012
  5. Changed configure --libdir to --appdir. Closes #1919.

    The --libdir option is retained for specifying where to put the Rubinius
    shared library. We don't build a shared or static library yet but we will
    when we have an embedding API.
    The --appdir option specifies where the application runtime and library
    files will be installed. The use of --libdir for this was misleading.
    brixen committed Oct 3, 2012
  6. Silence iconv warning.

    brixen committed Oct 1, 2012
Commits on Oct 2, 2012
  1. Trigger a wakeup after setting up the listening socket for the agent

    This fixes the problem on OS X where opening an agent console would just
    hang. Related to #1933.
    dbussink committed Oct 2, 2012
  2. Install man pages.

    voxik committed Oct 2, 2012
  3. Teach the JIT about fixnum_mul

    We have to work around a bug in LLVM here which makes smul.with.overflow
    not work correctly. It will not properly detect overflows if we use the
    i63 / i31 integer types as described here:
    We work around it by do multiplication with overflow detection for i64
    and i32 which does work and manually do a bounds check for Fixnum::MIN
    and Fixnum::MAX.
    dbussink committed Oct 2, 2012
  4. Use better descriptive name for comparison operations

    float_compare and fixnum_compare are actually the comparators for the
    Float#<=> and Fixnum#<=> behavior, so this name makes it less ambiguous
    what these wrappers are for.
    dbussink committed Oct 2, 2012
Commits on Oct 1, 2012
  1. Add missing file

    dbussink committed Oct 1, 2012
  2. Use a block around OnStack with setjmp / longjmp for C-API exceptions

    This way we ensure that the destructor for OnStack is executed, even
    when we encounter an exception. We use setjmp / longjmp in the C-API but
    that means we can't depend on RAII behavior here.
    By moving this into a separate block we make sure the destructor part of
    RAII runs before we do the longjmp.
    dbussink committed Oct 1, 2012
  3. OnStack can mark a location on the stack that contains NULL

    When we use OnStack, we might mark a location that holds the return
    value of a call. This return value can be NULL to indicate an exception
    so if we see a NULL here, that is a valid value and we just shouldn't do
    anything with it.
    Before this would cause a segmentation fault because we would deference
    a NULL pointer.
    dbussink committed Oct 1, 2012
  4. Only try to remove a GC root if it's actually used

    These are initialized with cUndef. We already guard the cases for adding
    them so we don't add cUndef, so we should also guard this case when
    trying to remove it.
    dbussink committed Oct 1, 2012
Commits on Sep 29, 2012
  1. Merge pull request #1935 from ryoqun/thread-kill-lock

    Fix newly-introduced Thread#kill deadlock
    dbussink committed Sep 29, 2012
  2. Fix newly-introduced Thread#kill deadlock

    The following commit introduced another deadlock. Threads MUST be
    Rubinius.synchronize-d, but they weren't in Thread#kill.
      ba7667b Correctly implement Thread#kill
    Previously, Thread#kill internally used Thread#raise, which had gotten
    completely deadlock-free. And there is no problem.
    But, the commit made Thread#kill use a primitive for its function. That must be
    inside Rubinius.synchronize as was before.
    ryoqun committed Sep 29, 2012
Commits on Sep 28, 2012
  1. Use Iconv extension from MRI

    Fixes #1924
    dbussink committed Sep 28, 2012