Commits on Oct 24, 2011
  1. Temporal fix for long directory name support

    * src/code/win32-pathname.lisp (unparse-native-win32-namestring):
    add "fudge factor" of 12 characters for long directory names.
    committed Oct 22, 2011
  2. Improved GC lossage

      * Change size_lose() and scav_lose() to indicate the widetag of the
    header of the losing object rather than of the pointer to the losing
      * Patch from Paul Khuong.
    Alastair Bridgewater committed with Feb 20, 2010
  3. Merge branch 'master' into mswinmt

    committed Oct 23, 2011
Commits on Oct 18, 2011
  1. Replace (speed 0) with (speed 3) in some inline _functions_ of load.lisp

    * src/code/load.lisp: we have (optimize (speed 0)) in the expansion code of some macros, and in inline functions (read-byte-arg and friends). Changed to (speed 3) for the latter case.
    committed Oct 18, 2011
Commits on Oct 12, 2011
Commits on Oct 11, 2011
  1. @stassats

    (format t "a~0&b") shouldn't print any newlines.

    Fixes lp#867684.
    stassats committed Oct 12, 2011
Commits on Oct 9, 2011
  1. @jsnell
  2. @jsnell
  3. @jsnell
Commits on Oct 7, 2011
  1. @csrhodes

    really fix float rounding on 32-bit platforms

    Now, as well as not giving completely the wrong answer for most
    larger-than-fixnum double floats, we additionally don't try to perform
    arithmetic on NIL for larger-than-fixnum single-floats.
    csrhodes committed Oct 7, 2011
Commits on Sep 30, 2011
  1. @lichtblau

    fix run-program misaligned access

    Thanks to Antoni Grzymala for the report.
    committed with lichtblau Sep 30, 2011
  2. @lichtblau

    Fix the build on Windows

    lichtblau committed Sep 30, 2011
Commits on Sep 29, 2011
  1. @lichtblau

    Ignore a known failure in on Darwin includes a test case that fails on Darwin because of
    a bug in SBCL.  However, this is the first release to run this test at
    all.  Lacking a proper mechanism to mark the test case as a known
    failure on Darwin, skip it entirely on that platform for now.
    Thanks to Fare for testing.
    lichtblau committed Sep 30, 2011
  2. Merge branch 'master' into mswinmt

    committed Sep 29, 2011
  3. @jsnell
  4. @jsnell
Commits on Sep 23, 2011
  1. Some fixes for win32 stack overflow patch Fix toplevel.lisp: Add an i…

    …gnore declaration
    Fix handler-case syntax in target-thread.lisp.
    committed Sep 23, 2011
  2. Handle win32 control stack overflow sensibly

    * src/code/target-thread.lisp (make-thread): terminate thread on control-stack-exhausted by default
    * src/code/toplevel.lisp (toplevel-repl): throw toplevel-catcher on control-stack-exhausted; reset guard page on win32 as well
    * src/runtime/win32-os.c (handle_exception): call control-stack-exhausted-error on stack overflow
    * src/runtime/interrupt.c (reset_control_stack_guard_page): windows-specific code now calls _resetstkoflw
    committed Sep 23, 2011
  3. @jimwise

    One missed contrib subdir.

    jimwise committed Sep 23, 2011
Commits on Sep 20, 2011
  1. @nikodemus

    handle non-standard slot allocations when updating classes

     * Refactor layout comparison to work on the effective slot definition list(s)
       directly -- easier to understand.
     * When new slots with custom allocation are added, add their names to the
       "added" list for UPDATE-INSTANCE-FOR-REDEFINED-CLASS. This is not specified
       by ANSI, but unless we do this those slots don't get initialized.
       Removing custom slots is hairier, as is changing a custom slot into
       a normal slot. Add some tests that poke in this area as well...
     * Replace wrapper-instance-slot-layout and wrapper-class-slots with the
       CLASS-SLOTS lists -- saves space and makes things easier to understand.
       Has a small performance cost for updating instances and SLOT-MISSING. Will
       refactor again if this is critical in the real world.
    nikodemus committed Sep 20, 2011
Commits on Sep 10, 2011
Commits on Sep 8, 2011
  1. @csrhodes

    cleanup DESCRIBE of symbols naming type specifiers a bit

    Firstly, don't indent the "... names a type specifier".  But then
    wander around in confusion for a while trying to debug where the extra
    indentation comes from, before working out that the pprint-newline at
    the end of describe-class introduces indentation as well as a newline,
    causing the subsequent logical block to be opened at an unexpected
    horizontal position.  Fix fix fix.
    csrhodes committed Sep 8, 2011
  2. More cleanup after sb-auto-fpu-switch removal

    * src/runtime/x86-fpu-insn.S: deleted
    * src/runtime/x86-assem.S: don't include x86-fpu-insn.S
    committed Sep 8, 2011
  3. Cleaned some ugly ifdeffery that was intended as temporary.

    * src/runtime/win32-os.c (handle_exception): unified 32/64-bit code at try_recommit label.
    committed Sep 8, 2011
  4. Clean up some leftovers from sb-auto-fpu-switch

    * src/runtime/thread.c: no establish_c_fpu_world
    * src/runtime/win32-os.h: no need to declare it
    committed Sep 8, 2011
  5. Removed intermediate win32-os.s that shouldn't be version-controlled …

    …at all.
    * src/runtime/win32-os.s: deleted
    committed Sep 8, 2011