Fix building with Clang for Darwin (OS X 10.6+ and iOS
authorZachary Waldowski <zwaldowski@gmail.com>
Thu, 5 Apr 2012 16:32:41 +0000 (12:32 -0400)
committerAnthony Green <green@moxielogic.com>
Thu, 5 Apr 2012 16:32:41 +0000 (12:32 -0400)
 4.0+)

14 files changed:
.pc/apple-fixes/.timestamp [new file with mode: 0644]
.pc/apple-fixes/ChangeLog [new file with mode: 0644]
.pc/apple-fixes/include/ffi.h.in [new file with mode: 0644]
.pc/apple-fixes/src/arm/sysv.S [new file with mode: 0644]
.pc/apple-fixes/src/x86/ffi.c [new file with mode: 0644]
.pc/applied-patches
ChangeLog
include/ffi.h.in
patches/apple-fixes [new file with mode: 0644]
patches/series
src/arm/sysv.S
src/x86/ffi.c
testsuite/libffi.call/a.out [new file with mode: 0755]
testsuite/libffi.call/promotion.c

diff --git a/.pc/apple-fixes/.timestamp b/.pc/apple-fixes/.timestamp
new file mode 100644 (file)
index 0000000..e69de29
diff --git a/.pc/apple-fixes/ChangeLog b/.pc/apple-fixes/ChangeLog
new file mode 100644 (file)
index 0000000..e1892aa
--- /dev/null
@@ -0,0 +1,5010 @@
+2012-04-02  Peter Rosin  <peda@lysator.liu.se>
+
+       * src/x86/win32.S (ffi_call_win32): Sign/zero extend the return
+       value in the Intel version as is already done for the AT&T version.
+       (ffi_closure_SYSV): Likewise.
+       (ffi_closure_raw_SYSV): Likewise.
+       (ffi_closure_STDCALL): Likewise.
+
+2012-03-29  Peter Rosin  <peda@lysator.liu.se>
+
+       * src/x86/win32.S (ffi_closure_raw_THISCALL): Unify the frame
+       generation, fix the ENDP label and remove the surplus third arg
+       from the 'lea' insn.
+
+2012-03-29  Peter Rosin  <peda@lysator.liu.se>
+
+       * src/x86/win32.S (ffi_closure_raw_SYSV): Make the 'stubraw' label
+       visible outside the PROC, so that ffi_closure_raw_THISCALL can see
+       it.  Also instruct the assembler to add a frame to the function.
+
+2012-03-23  Peter Rosin  <peda@lysator.liu.se>
+
+       * Makefile.am (AM_CPPFLAGS): Add -DFFI_BUILDING.
+       * Makefile.in: Rebuilt.
+       * include/ffi.h.in [MSVC]: Add __declspec(dllimport) decorations
+       to all data exports, when compiling libffi clients using MSVC.
+
+2012-03-29  Peter Rosin  <peda@lysator.liu.se>
+
+       * src/x86/ffitarget.h (ffi_abi): Add new ABI FFI_MS_CDECL and
+       make it the default for MSVC.
+       (FFI_TYPE_MS_STRUCT): New structure return convention.
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Tweak the structure
+       return convention for FFI_MS_CDECL to be FFI_TYPE_MS_STRUCT
+       instead of an ordinary FFI_TYPE_STRUCT.
+       (ffi_prep_args): Treat FFI_TYPE_MS_STRUCT as FFI_TYPE_STRUCT.
+       (ffi_call): Likewise.
+       (ffi_prep_incoming_args_SYSV): Likewise.
+       (ffi_raw_call): Likewise.
+       (ffi_prep_closure_loc): Treat FFI_MS_CDECL as FFI_SYSV.
+       * src/x86/win32.S (ffi_closure_SYSV): For FFI_TYPE_MS_STRUCT,
+       return a pointer to the result structure in eax and don't pop
+       that pointer from the stack, the caller takes care of it.
+       (ffi_call_win32): Treat FFI_TYPE_MS_STRUCT as FFI_TYPE_STRUCT.
+       (ffi_closure_raw_SYSV): Likewise.
+
+2012-03-22  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/closure_stdcall.c [MSVC]: Add inline
+       assembly version with Intel syntax.
+       * testsuite/libffi.call/closure_thiscall.c [MSVC]: Likewise.
+
+2012-03-23  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/ffitest.h: Provide abstration of
+       __attribute__((fastcall)) in the form of a __FASTCALL__
+       define.  Define it to __fastcall for MSVC.
+       * testsuite/libffi.call/fastthis1_win32.c: Use the above.
+       * testsuite/libffi.call/fastthis2_win32.c: Likewise.
+       * testsuite/libffi.call/fastthis3_win32.c: Likewise.
+       * testsuite/libffi.call/strlen2_win32.c: Likewise.
+       * testsuite/libffi.call/struct1_win32.c: Likewise.
+       * testsuite/libffi.call/struct2_win32.c: Likewise.
+
+2012-03-22  Peter Rosin  <peda@lysator.liu.se>
+
+       * src/x86/win32.S [MSVC] (ffi_closure_THISCALL): Remove the manual
+       frame on function entry, MASM adds one automatically.
+
+2012-03-22  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/ffitest.h [MSVC]: Add kludge for missing
+       bits in the MSVC headers.
+
+2012-03-22  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/cls_12byte.c: Adjust to the C89 style
+       with no declarations after statements.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_18byte.c: Likewise.
+       * testsuite/libffi.call/cls_19byte.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte1.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_5_1_byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+       * testsuite/libffi.call/cls_64byte.c: Likewise.
+       * testsuite/libffi.call/cls_6_1_byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7_1_byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_9byte1.c: Likewise.
+       * testsuite/libffi.call/cls_9byte2.c: Likewise.
+       * testsuite/libffi.call/cls_align_double.c: Likewise.
+       * testsuite/libffi.call/cls_align_float.c: Likewise.
+       * testsuite/libffi.call/cls_align_longdouble.c: Likewise.
+       * testsuite/libffi.call/cls_align_longdouble_split.c: Likewise.
+       * testsuite/libffi.call/cls_align_longdouble_split2.c: Likewise.
+       * testsuite/libffi.call/cls_align_pointer.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint64.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint64.c: Likewise.
+       * testsuite/libffi.call/cls_dbls_struct.c: Likewise.
+       * testsuite/libffi.call/cls_pointer_stack.c: Likewise.
+       * testsuite/libffi.call/err_bad_typedef.c: Likewise.
+       * testsuite/libffi.call/huge_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct1.c: Likewise.
+       * testsuite/libffi.call/nested_struct10.c: Likewise.
+       * testsuite/libffi.call/nested_struct2.c: Likewise.
+       * testsuite/libffi.call/nested_struct3.c: Likewise.
+       * testsuite/libffi.call/nested_struct4.c: Likewise.
+       * testsuite/libffi.call/nested_struct5.c: Likewise.
+       * testsuite/libffi.call/nested_struct6.c: Likewise.
+       * testsuite/libffi.call/nested_struct7.c: Likewise.
+       * testsuite/libffi.call/nested_struct8.c: Likewise.
+       * testsuite/libffi.call/nested_struct9.c: Likewise.
+       * testsuite/libffi.call/stret_large.c: Likewise.
+       * testsuite/libffi.call/stret_large2.c: Likewise.
+       * testsuite/libffi.call/stret_medium.c: Likewise.
+       * testsuite/libffi.call/stret_medium2.c: Likewise.
+       * testsuite/libffi.call/struct1.c: Likewise.
+       * testsuite/libffi.call/struct1_win32.c: Likewise.
+       * testsuite/libffi.call/struct2.c: Likewise.
+       * testsuite/libffi.call/struct2_win32.c: Likewise.
+       * testsuite/libffi.call/struct3.c: Likewise.
+       * testsuite/libffi.call/struct4.c: Likewise.
+       * testsuite/libffi.call/struct5.c: Likewise.
+       * testsuite/libffi.call/struct6.c: Likewise.
+       * testsuite/libffi.call/struct7.c: Likewise.
+       * testsuite/libffi.call/struct8.c: Likewise.
+       * testsuite/libffi.call/struct9.c: Likewise.
+       * testsuite/libffi.call/testclosure.c: Likewise.
+
+2012-03-06  Chung-Lin Tang  <cltang@codesourcery.com>
+
+       * src/arm/ffi.c (ffi_call): Add __ARM_EABI__ guard around call to
+       ffi_call_VFP().
+       (ffi_prep_closure_loc): Add __ARM_EABI__ guard around use of
+       ffi_closure_VFP.
+       * src/arm/sysv.S: Add __ARM_EABI__ guard around VFP code.
+
+2012-03-21  Peter Rosin         <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/float_va.c (float_va_fn): Use %f when
+       printing doubles (%lf is for long doubles).
+       (main): Likewise.
+
+2012-03-21  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/lib/target-libpath.exp [*-*-cygwin*, *-*-mingw*]
+       (set_ld_library_path_env_vars): Add the library search dir to PATH
+       (and save PATH for later).
+       (restore_ld_library_path_env_vars): Restore PATH.
+
+2012-03-20  Peter Rosin  <peda@lysator.liu.se>
+
+       * testsuite/libffi.call/strlen2_win32.c (main): Remove bug.
+       * src/x86/win32.S [MSVC] (ffi_closure_SYSV): Make the 'stub' label
+       visible outside the PROC, so that ffi_closure_THISCALL can see it.
+
+2012-03-19  Alan Hourihane  <alanh@fairlite.co.uk>
+
+       * src/m68k/ffi.c: Add MINT support.
+       * src/m68k/sysv.S: Ditto.
+
+2012-03-19  chennam  <csit@axway.com>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_closure_loc): Fix AIX closure
+       support.
+
+2012-04-02  Peter Bergner  <bergner@vnet.ibm.com>
+
+       * src/powerpc/ffi.c (ffi_prep_args_SYSV): Declare double_tmp.
+       Silence casting pointer to integer of different size warning.
+       Delete goto to previously deleted label.
+       (ffi_call): Silence possibly undefined warning.
+       (ffi_closure_helper_SYSV): Declare variable type.
+
+2012-03-13  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_prep_closure_loc): Don't ASSERT ABI test,
+       just return FFI_BAD_ABI when things are wrong.
+       * src/sh64/ffi.c (ffi_prep_closure_loc): Ditto.
+
+2012-03-09  David Edelsohn  <dje.gcc@gmail.com>
+
+       * src/powerpc/aix_closure.S (ffi_closure_ASM): Adjust for Darwin64
+       change to return value of ffi_closure_helper_DARWIN and load type
+       from return type.
+
+2012-03-03  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * src/x86/ffi64.c (ffi_call): Cast the return value to unsigned
+       long.
+       (ffi_prep_closure_loc): Cast to 64bit address in trampoline.
+       (ffi_closure_unix64_inner): Cast return pointer to unsigned long
+       first.
+
+       * src/x86/ffitarget.h (FFI_SIZEOF_ARG): Defined to 8 for x32.
+       (ffi_arg): Set to unsigned long long for x32.
+       (ffi_sarg): Set to long long for x32.
+
+2012-03-03  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * src/prep_cif.c (ffi_prep_cif_core): Properly check bad ABI.
+
+2012-03-03  Andoni Morales Alastruey  <ylatuya@gmail.com>
+
+       * configure.ac: Add -no-undefined for both 32- and 64-bit x86
+       windows-like hosts.
+       * configure: Rebuilt.
+
+2012-02-27  Mikael Pettersson  <mikpe@it.uu.se>
+
+       PR libffi/52223
+       * Makefile.am (FLAGS_TO_PASS): Define.
+       * Makefile.in: Regenerate.
+
+2012-02-23  Anthony Green  <green@moxielogic.com>
+
+       * src/*/ffitarget.h: Ensure that users never include ffitarget.h
+       directly.
+
+2012-02-23  Kai Tietz  <ktietz@redhat.com>
+
+       PR libffi/52221
+       * src/x86/ffi.c (ffi_closure_raw_THISCALL): New
+       prototype.
+       (ffi_prep_raw_closure_loc): Use ffi_closure_raw_THISCALL for
+       thiscall-convention.
+       (ffi_raw_call): Use ffi_prep_args_raw.
+       * src/x86/win32.S (ffi_closure_raw_THISCALL): Add
+       implementation for stub.
+
+2012-02-10  Kai Tietz  <ktietz@redhat.com>
+
+       * configure.ac (AM_LTLDFLAGS): Add -no-undefine for x64
+       windows target.
+       * configure: Regenerated.
+
+2012-02-08  Kai Tietz  <ktietz@redhat.com>
+
+       * src/prep_cif.c (ffi_prep_cif): Allow for X86_WIN32
+       also FFI_THISCALL.
+       * src/x86/ffi.c (ffi_closure_THISCALL): Add prototype.
+       (FFI_INIT_TRAMPOLINE_THISCALL): New trampoline code.
+       (ffi_prep_closure_loc): Add FFI_THISCALL support.
+       * src/x86/ffitarget.h (FFI_TRAMPOLINE_SIZE): Adjust size.
+       * src/x86/win32.S (ffi_closure_THISCALL): New closure code
+       for thiscall-calling convention.
+       * testsuite/libffi.call/closure_thiscall.c: New test.
+
+2012-01-28  Kai Tietz  <ktietz@redhat.com>
+
+       * src/libffi/src/x86/ffi.c (ffi_call_win32): Add new
+       argument to prototype for specify calling-convention.
+       (ffi_call): Add support for stdcall/thiscall convention.
+       (ffi_prep_args): Likewise.
+       (ffi_raw_call): Likewise.
+       * src/x86/ffitarget.h (ffi_abi): Add FFI_THISCALL and
+       FFI_FASTCALL.
+       * src/x86/win32.S (_ffi_call_win32): Add support for
+       fastcall/thiscall calling-convention calls.
+       * testsuite/libffi.call/fastthis1_win32.c: New test.
+       * testsuite/libffi.call/fastthis2_win32.c: New test.
+       * testsuite/libffi.call/fastthis3_win32.c: New test.
+       * testsuite/libffi.call/strlen2_win32.c: New test.
+       * testsuite/libffi.call/many2_win32.c: New test.
+       * testsuite/libffi.call/struct1_win32.c: New test.
+       * testsuite/libffi.call/struct2_win32.c: New test.
+
+2012-01-23  Uros Bizjak  <ubizjak@gmail.com>
+
+       * src/alpha/ffi.c (ffi_prep_closure_loc): Check for bad ABI.
+
+2012-01-23  Anthony Green  <green@moxielogic.com>
+           Chris Young  <cdyoung@ntlworld.com>
+
+       * configure.ac: Add Amiga support.
+       * configure: Rebuilt.
+
+2012-01-23  Dmitry Nadezhin  <dmitry.nadezhin@gmail.com>
+
+       * include/ffi_common.h (LIKELY, UNLIKELY): Fix definitions.
+
+2012-01-23  Andreas Schwab  <schwab@linux-m68k.org>
+
+       * src/m68k/sysv.S (ffi_call_SYSV): Properly test for plain
+       mc68000.  Test for __HAVE_68881__ in addition to __MC68881__.
+
+2012-01-19  Jakub Jelinek  <jakub@redhat.com>
+
+       PR rtl-optimization/48496
+       * src/ia64/ffi.c (ffi_call): Fix up aliasing violations.
+
+2012-01-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * configure.ac (i?86-*-*): Set TARGET to X86_64.
+       * configure: Regenerate.
+
+2011-12-07  Andrew Pinski  <apinski@cavium.com>
+
+       PR libffi/50051
+       * src/mips/n32.S: Add ".set mips4".
+
+2011-11-21  Andreas Tobler  <andreast@fgznet.ch>
+
+       * configure: Regenerate.
+
+2011-11-12  David Gilbert <david.gilbert@linaro.org>
+
+       * doc/libffi.texi, include/ffi.h.in, include/ffi_common.h,
+       man/Makefile.am, man/ffi.3, man/ffi_prep_cif.3,
+       man/ffi_prep_cif_var.3, src/arm/ffi.c, src/arm/ffitarget.h,
+       src/cris/ffi.c, src/prep_cif.c,
+       testsuite/libffi.call/cls_double_va.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/float_va.c: Many changes to support variadic
+       function calls.
+
+2011-11-12  Kyle Moffett <Kyle.D.Moffett@boeing.com>
+
+       * src/powerpc/ffi.c, src/powerpc/ffitarget.h,
+       src/powerpc/ppc_closure.S, src/powerpc/sysv.S: Many changes for
+       softfloat powerpc variants.
+
+2011-11-12  Petr Salinger <Petr.Salinger@seznam.cz>
+
+       * configure.ac (FFI_EXEC_TRAMPOLINE_TABLE): Fix kfreebsd support.
+       * configure: Rebuilt.
+
+2011-11-12  Timothy Wall  <twall@users.sf.net>
+
+       * src/arm/ffi.c (ffi_prep_args, ffi_prep_incoming_args_SYSV): Max
+       alignment of 4 for wince on ARM.
+
+2011-11-12  Kyle Moffett <Kyle.D.Moffett@boeing.com>
+           Anthony Green <green@moxielogic.com>
+
+       * src/ppc/sysv.S, src/ppc/ffi.c: Remove use of ppc string
+       instructions (not available on some cores, like the PPC440).
+
+2011-11-12  Kimura Wataru  <kimuraw@i.nifty.jp>
+
+       * m4/ax_enable_builddir: Change from string comparison to numeric
+       comparison for wc output.
+       * configure.ac: Enable FFI_MMAP_EXEC_WRIT for darwin11 aka Mac OS
+       X 10.7.
+       * configure: Rebuilt.
+
+2011-11-12  Anthony Green  <green@moxielogic.com>
+
+       * Makefile.am (AM_CCASFLAGS): Add -g option to build assembly
+       files with debug info.
+       * Makefile.in: Rebuilt.
+
+2011-11-12  Jasper Lievisse Adriaanse <jasper@openbsd.org>
+
+       * README: Update list of supported OpenBSD systems.
+
+2011-11-12  Anthony Green  <green@moxielogic.com>
+
+       * libtool-version: Update.
+       * Makefile.am (nodist_libffi_la_SOURCES): Add src/debug.c if
+       FFI_DEBUG.
+       (libffi_la_SOURCES): Remove src/debug.c
+       (EXTRA_DIST): Add src/debug.c
+       * Makefile.in: Rebuilt.
+       * README: Update for 3.0.11.
+
+2011-11-10  Richard Henderson  <rth@redhat.com>
+
+       * configure.ac (GCC_AS_CFI_PSEUDO_OP): Use it instead of inline check.
+       * configure, aclocal.m4: Rebuild.
+
+2011-09-04  Iain Sandoe  <iains@gcc.gnu.org>
+
+       PR libffi/49594
+       * src/powerpc/darwin_closure.S (stubs): Make the stub binding
+       helper reference track the architecture pointer size.
+
+2011-08-25  Andrew Haley  <aph@redhat.com>
+
+       * src/arm/ffi.c (FFI_INIT_TRAMPOLINE): Remove hard-coded assembly
+       instructions.
+       * src/arm/sysv.S (ffi_arm_trampoline): Put them here instead.
+
+2011-07-11  Andrew Haley  <aph@redhat.com>
+
+        * src/arm/ffi.c (FFI_INIT_TRAMPOLINE): Clear icache.
+
+2011-06-29  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * testsuite/libffi.call/cls_double_va.c: Move PR number to comment.
+       * testsuite/libffi.call/cls_longdouble_va.c: Likewise.
+
+2011-06-29  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       PR libffi/46660
+       * testsuite/libffi.call/cls_double_va.c: xfail dg-output on
+       mips-sgi-irix6*.
+       * testsuite/libffi.call/cls_longdouble_va.c: Likewise.
+
+2011-06-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * testsuite/libffi.call/huge_struct.c (test_large_fn): Use PRIu8,
+       PRId8 instead of %hhu, %hhd.
+       * testsuite/libffi.call/ffitest.h [__alpha__ && __osf__] (PRId8,
+       PRIu8): Define.
+       [__sgi__] (PRId8, PRIu8): Define.
+
+2011-04-29  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * src/alpha/osf.S (UA_SI, FDE_ENCODING, FDE_ENCODE, FDE_ARANGE):
+       Define.
+       Use them to handle ELF vs. ECOFF differences.
+       [__osf__] (_GLOBAL__F_ffi_call_osf): Define.
+
+2011-03-30  Timothy Wall  <twall@users.sf.net>
+
+       * src/powerpc/darwin.S: Fix unknown FDE encoding.
+       * src/powerpc/darwin_closure.S: ditto.
+
+2011-02-25  Anthony Green  <green@moxielogic.com>
+
+       * src/powerpc/ffi.c (ffi_prep_closure_loc): Allow for more
+       32-bit ABIs.
+
+2011-02-15  Anthony Green  <green@moxielogic.com>
+
+       * m4/ax_cc_maxopt.m4: Don't -malign-double or use -ffast-math.
+       * configure: Rebuilt.
+
+2011-02-13  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure: Regenerate.
+
+2011-02-13  Anthony Green  <green@moxielogic.com>
+
+       * include/ffi_common.h (UNLIKELY, LIKELY): Define.
+       * src/x86/ffi64.c (UNLIKELY, LIKELY): Remove definition.
+       * src/prep_cif.c (UNLIKELY, LIKELY): Remove definition.
+
+       * src/prep_cif.c (initialize_aggregate): Convert assertion into
+       FFI_BAD_TYPEDEF return.  Initialize arg size and alignment to 0.
+
+       * src/pa/ffi.c (ffi_prep_closure_loc): Don't ASSERT ABI test,
+       just return FFI_BAD_ABI when things are wrong.
+       * src/arm/ffi.c (ffi_prep_closure_loc): Ditto.
+       * src/powerpc/ffi.c (ffi_prep_closure_loc): Ditto.
+       * src/mips/ffi.c (ffi_prep_closure_loc): Ditto.
+       * src/ia64/ffi.c (ffi_prep_closure_loc): Ditto.
+       * src/avr32/ffi.c (ffi_prep_closure_loc): Ditto.
+
+2011-02-11  Anthony Green  <green@moxielogic.com>
+
+       * src/sparc/ffi.c (ffi_prep_closure_loc): Don't ASSERT ABI test,
+       just return FFI_BAD_ABI when things are wrong.
+
+2012-02-11  Eric Botcazou  <ebotcazou@adacore.com>
+
+       * src/sparc/v9.S (STACKFRAME): Bump to 176.
+
+2011-02-09  Stuart Shelton  <srcshelton@gmail.com>
+
+       http://bugs.gentoo.org/show_bug.cgi?id=286911
+       * src/mips/ffitarget.h: Clean up error messages.
+       * src/java_raw_api.c (ffi_java_translate_args): Cast raw arg to
+       ffi_raw*.
+       * include/ffi.h.in: Add pragma for SGI compiler.
+
+2011-02-09  Anthony Green  <green@moxielogic.com>
+
+       * configure.ac: Add powerpc64-*-darwin* support.
+
+2011-02-09  Anthony Green <green@moxielogic.com>
+
+       * README: Mention Interix.
+
+2011-02-09  Jonathan Callen  <abcd@gentoo.org>
+
+       * configure.ac: Add Interix to win32/cygwin/mingw case.
+       * configure: Ditto.
+       * src/closures.c: Treat Interix like Cygwin, instead of as a
+       generic win32.
+
+2011-02-09  Anthony Green <green@moxielogic.com>
+
+       * testsuite/libffi.call/err_bad_typedef.c: Remove xfail.
+       * testsuite/libffi.call/err_bad_abi.c: Remove xfail.
+       * src/x86/ffi64.c (UNLIKELY, LIKELY): Define.
+       (ffi_prep_closure_loc): Check for bad ABI.
+       * src/prep_cif.c (UNLIKELY, LIKELY): Define.
+       (initialize_aggregate): Check for bad types.
+
+2011-02-09  Landon Fuller <landonf@plausible.coop>
+
+       * Makefile.am (EXTRA_DIST): Add build-ios.sh, src/arm/gentramp.sh,
+       src/arm/trampoline.S.
+       (nodist_libffi_la_SOURCES): Add src/arc/trampoline.S.
+       * configure.ac (FFI_EXEC_TRAMPOLINE_TABLE): Define.
+       * src/arm/ffi.c (ffi_trampoline_table)
+       (ffi_closure_trampoline_table_page, ffi_trampoline_table_entry)
+       (FFI_TRAMPOLINE_CODELOC_CONFIG, FFI_TRAMPOLINE_CONFIG_PAGE_OFFSET)
+       (FFI_TRAMPOLINE_COUNT, ffi_trampoline_lock, ffi_trampoline_tables)
+       (ffi_trampoline_table_alloc, ffi_closure_alloc, ffi_closure_free):
+       Define for FFI_EXEC_TRAMPOLINE_TABLE case (iOS).
+       (ffi_prep_closure_loc): Handl FFI_EXEC_TRAMPOLINE_TABLE case
+       separately.
+       * src/arm/sysv.S: Handle Apple iOS host.
+       * src/closures.c: Handle FFI_EXEC_TRAMPOLINE_TABLE case.
+       * build-ios.sh: New file.
+       * fficonfig.h.in, configure, Makefile.in: Rebuilt.
+       * README: Mention ARM iOS.
+
+2011-02-08  Oren Held  <orenhe@il.ibm.com>
+
+       * src/dlmalloc.c (_STRUCT_MALLINFO): Define in order to avoid
+       redefinition of mallinfo on HP-UX.
+
+2011-02-08  Ginn Chen  <ginn.chen@oracle.com>
+
+       * src/sparc/ffi.c (ffi_call): Make compatible with Solaris Studio
+       aggregate return ABI.  Flush cache.
+       (ffi_prep_closure_loc): Flush cache.
+
+2011-02-11  Anthony Green  <green@moxielogic.com>
+
+       From Tom Honermann <tom.honermann@oracle.com>:
+       * src/powerpc/aix.S (ffi_call_AIX): Support for xlc toolchain on
+       AIX.  Declare .ffi_prep_args.  Insert nops after branch
+       instructions so that the AIX linker can insert TOC reload
+       instructions.
+       * src/powerpc/aix_closure.S: Declare .ffi_closure_helper_DARWIN.
+
+2011-02-08  Ed  <ed@kdtc.net>
+
+       * src/powerpc/asm.h: Fix grammar nit in comment.
+
+2011-02-08  Uli Link  <ul.mcamafia@linkitup.de>
+
+       * include/ffi.h.in (FFI_64_BIT_MAX): Define and use.
+
+2011-02-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       PR libffi/46661
+       * testsuite/libffi.call/cls_pointer.c (main): Cast void * to
+       uintptr_t first.
+       * testsuite/libffi.call/cls_pointer_stack.c (main): Likewise.
+
+2011-02-08  Rafael Avila de Espindola  <respindola@mozilla.com>
+
+       * configure.ac: Fix x86 test for pc related relocs.
+       * configure: Rebuilt.
+
+2011-02-07  Joel Sherrill <joel.sherrill@oarcorp.com>
+
+       * libffi/src/m68k/ffi.c: Add RTEMS support for cache flushing.
+       Handle case when CPU variant does not have long double support.
+       * libffi/src/m68k/sysv.S: Add support for mc68000, Coldfire,
+       and cores with soft floating point.
+
+2011-02-07  Joel Sherrill <joel.sherrill@oarcorp.com>
+
+       * configure.ac: Add mips*-*-rtems* support.
+       * configure: Regenerate.
+       * src/mips/ffitarget.h: Ensure needed constants are available
+       for targets which do not have sgidefs.h.
+
+2011-01-26  Dave Korn  <dave.korn.cygwin@gmail.com>
+
+       PR target/40125
+       * configure.ac (AM_LTLDFLAGS): Add -bindir option for windows DLLs.
+       * configure: Regenerate.
+
+2010-12-18  Iain Sandoe  <iains@gcc.gnu.org>
+
+       PR libffi/29152
+       PR libffi/42378
+       * src/powerpc/darwin_closure.S: Provide Darwin64 implementation,
+       update comments.
+       * src/powerpc/ffitarget.h (POWERPC_DARWIN64): New,
+       (FFI_TRAMPOLINE_SIZE): Update for Darwin64.
+       * src/powerpc/darwin.S: Provide Darwin64 implementation,
+       update comments.
+       * src/powerpc/ffi_darwin.c: Likewise.
+
+2010-12-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * configure.ac (libffi_cv_as_ascii_pseudo_op): Use double
+       backslashes.
+       (libffi_cv_as_string_pseudo_op): Likewise.
+       * configure: Regenerate.
+
+2010-12-03  Chung-Lin Tang  <cltang@codesourcery.com>
+
+       * src/arm/sysv.S (ffi_closure_SYSV): Add UNWIND to .pad directive.
+       (ffi_closure_VFP): Same.
+       (ffi_call_VFP): Move down to before ffi_closure_VFP. Add '.fpu vfp'
+       directive.
+
+2010-12-01  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * testsuite/libffi.call/ffitest.h [__sgi] (PRId64, PRIu64): Define.
+       (PRIuPTR): Define.
+
+2010-11-29  Richard Henderson  <rth@redhat.com>
+           Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * src/x86/sysv.S (FDE_ENCODING, FDE_ENCODE): Define.
+       (.eh_frame): Use FDE_ENCODING.
+       (.LASFDE1, .LASFDE2, LASFDE3): Simplify with FDE_ENCODE.
+
+2010-11-22  Jacek Caban <jacek@codeweavers.com>
+
+       * configure.ac: Check for symbol underscores on mingw-w64.
+       * configure: Rebuilt.
+       * src/x86/win64.S: Correctly access extern symbols in respect to
+       underscores.
+
+2010-11-15  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * testsuite/lib/libffi-dg.exp: Rename ...
+       * testsuite/lib/libffi.exp: ... to this.
+       * libffi/testsuite/libffi.call/call.exp: Don't load libffi-dg.exp.
+       * libffi/testsuite/libffi.special/special.exp: Likewise.
+
+2010-10-28  Chung-Lin Tang  <cltang@codesourcery.com>
+
+       * src/arm/ffi.c (ffi_prep_args): Add VFP register argument handling
+       code, new parameter, and return value. Update comments.
+       (ffi_prep_cif_machdep): Add case for VFP struct return values. Add
+       call to layout_vfp_args().
+       (ffi_call_SYSV): Update declaration.
+       (ffi_call_VFP): New declaration.
+       (ffi_call): Add VFP struct return conditions. Call ffi_call_VFP()
+       when ABI is FFI_VFP.
+       (ffi_closure_VFP): New declaration.
+       (ffi_closure_SYSV_inner): Add new vfp_args parameter, update call to
+       ffi_prep_incoming_args_SYSV().
+       (ffi_prep_incoming_args_SYSV): Update parameters. Add VFP argument
+       case handling.
+       (ffi_prep_closure_loc): Pass ffi_closure_VFP to trampoline
+       construction under VFP hard-float.
+       (rec_vfp_type_p): New function.
+       (vfp_type_p): Same.
+       (place_vfp_arg): Same.
+       (layout_vfp_args): Same.
+       * src/arm/ffitarget.h (ffi_abi): Add FFI_VFP. Define FFI_DEFAULT_ABI
+       based on __ARM_PCS_VFP.
+       (FFI_EXTRA_CIF_FIELDS): Define for adding VFP hard-float specific
+       fields.
+       (FFI_TYPE_STRUCT_VFP_FLOAT): Define internally used type code.
+       (FFI_TYPE_STRUCT_VFP_DOUBLE): Same.
+       * src/arm/sysv.S (ffi_call_SYSV): Change call of ffi_prep_args() to
+       direct call. Move function pointer load upwards.
+       (ffi_call_VFP): New function.
+       (ffi_closure_VFP): Same.
+
+       * testsuite/lib/libffi-dg.exp (check-flags): New function.
+       (dg-skip-if): New function.
+       * testsuite/libffi.call/cls_double_va.c: Skip if target is arm*-*-*
+       and compiler options include -mfloat-abi=hard.
+       * testsuite/libffi.call/cls_longdouble_va.c: Same.
+
+2010-10-01  Jakub Jelinek  <jakub@redhat.com>
+
+       PR libffi/45677
+       * src/x86/ffi64.c (ffi_prep_cif_machdep): Ensure cif->bytes is
+       a multiple of 8.
+       * testsuite/libffi.call/many2.c: New test.
+
+2010-08-20  Mark Wielaard  <mjw@redhat.com>
+
+       * src/closures.c (open_temp_exec_file_mnt): Check if getmntent_r
+       returns NULL.
+
+2010-08-09  Andreas Tobler  <andreast@fgznet.ch>
+
+       * configure.ac: Add target powerpc64-*-freebsd*.
+       * configure: Regenerate.
+       * testsuite/libffi.call/cls_align_longdouble_split.c: Pass
+       -mlong-double-128 only to linux targets.
+       * testsuite/libffi.call/cls_align_longdouble_split2.c: Likewise.
+       * testsuite/libffi.call/cls_longdouble.c: Likewise.
+       * testsuite/libffi.call/huge_struct.c: Likewise.
+
+2010-08-05  Dan Witte  <dwitte@mozilla.com>
+
+        * Makefile.am: Pass FFI_DEBUG define to msvcc.sh for linking to the
+        debug CRT when --enable-debug is given.
+        * configure.ac: Define it.
+        * msvcc.sh: Translate -g and -DFFI_DEBUG appropriately.
+
+2010-08-04  Dan Witte  <dwitte@mozilla.com>
+
+       * src/x86/ffitarget.h: Add X86_ANY define for all x86/x86_64
+       platforms.
+       * src/x86/ffi.c: Remove redundant ifdef checks.
+       * src/prep_cif.c: Push stack space computation into src/x86/ffi.c
+       for X86_ANY so return value space doesn't get added twice.
+
+2010-08-03  Neil Rashbrooke <neil@parkwaycc.co.uk>
+
+       * msvcc.sh: Don't pass -safeseh to ml64 because behavior is buggy.
+
+2010-07-22  Dan Witte  <dwitte@mozilla.com>
+
+       * src/*/ffitarget.h: Make FFI_LAST_ABI one past the last valid ABI.
+       * src/prep_cif.c: Fix ABI assertion.
+        * src/cris/ffi.c: Ditto.
+
+2010-07-10  Evan Phoenix  <evan@fallingsnow.net>
+
+       * src/closures.c (selinux_enabled_check): Fix strncmp usage bug.
+
+2010-07-07  Dan Horák <dan@danny.cz>
+
+       * include/ffi.h.in: Protect #define with #ifndef.
+       * src/powerpc/ffitarget.h: Ditto.
+       * src/s390/ffitarget.h: Ditto.
+       * src/sparc/ffitarget.h: Ditto.
+
+2010-07-07   Neil Roberts <neil@linux.intel.com>
+
+       * src/x86/sysv.S (ffi_call_SYSV): Align the stack pointer to
+       16-bytes.
+
+2010-07-02  Jakub Jelinek  <jakub@redhat.com>
+
+       * Makefile.am (AM_MAKEFLAGS): Pass also mandir to submakes.
+       * Makefile.in: Regenerated.
+
+2010-05-19  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * configure.ac (libffi_cv_as_x86_pcrel): Check for illegal in as
+       output, too.
+       (libffi_cv_as_ascii_pseudo_op): Check for .ascii.
+       (libffi_cv_as_string_pseudo_op): Check for .string.
+       * configure: Regenerate.
+       * fficonfig.h.in: Regenerate.
+       * src/x86/sysv.S (.eh_frame): Use .ascii, .string or error.
+
+2010-05-11  Dan Witte  <dwitte@mozilla.com>
+
+       * doc/libffi.tex: Document previous change.
+
+2010-05-11  Makoto Kato <m_kato@ga2.so-net.ne.jp>
+
+       * src/x86/ffi.c (ffi_call): Don't copy structs passed by value.
+
+2010-05-05  Michael Kohler <michaelkohler@live.com>
+
+       * src/dlmalloc.c (dlfree): Fix spelling.
+       * src/ia64/ffi.c (ffi_prep_cif_machdep): Ditto.
+       * configure.ac: Ditto.
+       * configure: Rebuilt.
+
+2010-04-13  Dan Witte  <dwitte@mozilla.com>
+
+       * msvcc.sh: Build with -W3 instead of -Wall.
+       * src/powerpc/ffi_darwin.c: Remove build warnings.
+       * src/x86/ffi.c: Ditto.
+       * src/x86/ffitarget.h: Ditto.
+
+2010-04-12  Dan Witte  <dwitte@mozilla.com>
+           Walter Meinl <wuno@lsvw.de>
+
+       * configure.ac: Add OS/2 support.
+       * configure: Rebuilt.
+       * src/closures.c: Ditto.
+       * src/dlmalloc.c: Ditto.
+       * src/x86/win32.S: Ditto.
+
+2010-04-07  Jakub Jelinek  <jakub@redhat.com>
+
+       * testsuite/libffi.call/err_bad_abi.c: Remove unused args variable.
+
+2010-04-02  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * Makefile.in: Regenerate.
+       * aclocal.m4: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * man/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2010-03-30  Dan Witte  <dwitte@mozilla.com>
+
+       * msvcc.sh: Disable build warnings.
+       * README (tested): Clarify windows build procedure.
+
+2010-03-15  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * configure.ac (libffi_cv_as_x86_64_unwind_section_type): New test.
+       * configure: Regenerate.
+       * fficonfig.h.in: Regenerate.
+       * libffi/src/x86/unix64.S (.eh_frame)
+       [HAVE_AS_X86_64_UNWIND_SECTION_TYPE]: Use @unwind section type.
+
+2010-03-14  Matthias Klose  <doko@ubuntu.com>
+
+       * src/x86/ffi64.c: Fix typo in comment.
+       * src/x86/ffi.c: Use /* ... */ comment style.
+
+2010-02-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * doc/libffi.texi (The Closure API): Fix typo.
+       * doc/libffi.info: Remove.
+
+2010-02-15  Matthias Klose  <doko@ubuntu.com>
+
+       * src/arm/sysv.S (__ARM_ARCH__): Define for processor
+       __ARM_ARCH_7EM__.
+
+2010-01-15  Anthony Green  <green@redhat.com>
+
+       * README: Add notes on building with Microsoft Visual C++.
+
+2010-01-15  Daniel Witte  <dwitte@mozilla.com>
+
+       * msvcc.sh: New file.
+
+       * src/x86/win32.S: Port assembly routines to MSVC and #ifdef.
+       * src/x86/ffi.c: Tweak function declaration and remove excess
+       parens.
+       * include/ffi.h.in: Add __declspec(align(8)) to typedef struct
+       ffi_closure.
+
+       * src/x86/ffi.c: Merge ffi_call_SYSV and ffi_call_STDCALL into new
+       function ffi_call_win32 on X86_WIN32.
+       * src/x86/win32.S (ffi_call_SYSV): Rename to ffi_call_win32.
+       (ffi_call_STDCALL): Remove.
+
+       * src/prep_cif.c (ffi_prep_cif): Move stack space allocation code
+       to ffi_prep_cif_machdep for x86.
+       * src/x86/ffi.c (ffi_prep_cif_machdep): To here.
+
+2010-01-15  Oliver Kiddle  <okiddle@yahoo.co.uk>
+
+       * src/x86/ffitarget.h (ffi_abi): Check for __i386 and __amd64 for
+       Sun Studio compiler compatibility.
+
+2010-01-12  Conrad Irwin <conrad.irwin@gmail.com>
+
+       * doc/libffi.texi: Add closure example.
+
+2010-01-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       PR libffi/40701
+       * testsuite/libffi.call/ffitest.h [__alpha__ && __osf__] (PRIdLL,
+       PRIuLL, PRId64, PRIu64, PRIuPTR): Define.
+       * testsuite/libffi.call/cls_align_sint64.c: Add -Wno-format on
+       alpha*-dec-osf*.
+       * testsuite/libffi.call/cls_align_uint64.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/return_ll1.c: Likewise.
+       * testsuite/libffi.call/stret_medium2.c: Likewise.
+       * testsuite/libffi.special/ffitestcxx.h (allocate_mmap): Cast
+       MAP_FAILED to char *.
+
+2010-01-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       * src/mips/n32.S: Use .abicalls and .eh_frame with __GNUC__.
+
+2009-12-31  Anthony Green  <green@redhat.com>
+
+       * README: Update for libffi 3.0.9.
+
+2009-12-27  Matthias Klose  <doko@ubuntu.com>
+
+       * configure.ac (HAVE_LONG_DOUBLE): Define for mips when
+       appropriate.
+       * configure: Rebuilt.
+
+2009-12-26  Anthony Green  <green@redhat.com>
+
+       * testsuite/libffi.call/cls_longdouble_va.c: Mark as xfail for
+       avr32*-*-*.
+       * testsuite/libffi.call/cls_double_va.c: Ditto.
+
+2009-12-26  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * testsuite/libffi.call/ffitest.h: Conditionally include stdint.h
+       and inttypes.h.
+       * testsuite/libffi.special/unwindtest.cc: Ditto.
+
+2009-12-26  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * configure.ac: Add amd64-*-openbsd*.
+       * configure: Rebuilt.
+       * testsuite/lib/libffi-dg.exp (libffi_target_compile): Link
+       openbsd programs with -lpthread.
+
+2009-12-26  Anthony Green  <green@redhat.com>
+
+       * testsuite/libffi.call/cls_double_va.c,
+       testsuite/libffi.call/cls_longdouble.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/cls_pointer.c,
+       testsuite/libffi.call/cls_pointer_stack.c: Remove xfail for
+       mips*-*-* and arm*-*-*.
+       * testsuite/libffi.call/cls_align_longdouble_split.c,
+       testsuite/libffi.call/cls_align_longdouble_split2.c,
+       testsuite/libffi.call/stret_medium2.c,
+       testsuite/libffi.call/stret_medium.c,
+       testsuite/libffi.call/stret_large.c,
+       testsuite/libffi.call/stret_large2.c: Remove xfail for arm*-*-*.
+
+2009-12-31  Kay Tietz  <ktietz70@googlemail.com>
+
+       * testsuite/libffi.call/ffitest.h,
+       testsuite/libffi.special/ffitestcxx.h (PRIdLL, PRuLL): Fix
+       definitions.
+
+2009-12-31  Carlo Bramini  <carlo.bramix@libero.it>
+
+       * configure.ac (AM_LTLDFLAGS): Define for windows hosts.
+       * Makefile.am (libffi_la_LDFLAGS): Add AM_LTLDFLAGS.
+       * configure: Rebuilt.
+       * Makefile.in: Rebuilt.
+
+2009-12-31  Anthony Green  <green@redhat.com>
+           Blake Chaffin.
+
+       * testsuite/libffi.call/huge_struct.c: New test case from Blake
+       Chaffin @ Apple.
+
+2009-12-28  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Copy abi and nargs to
+       local variables.
+       (aix_adjust_aggregate_sizes): New function.
+       (ffi_prep_cif_machdep): Call it.
+
+2009-12-26  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * configure.ac: Define FFI_MMAP_EXEC_WRIT for the given targets.
+       * configure: Regenerate.
+       * fficonfig.h.in: Likewise.
+       * src/closures.c: Remove the FFI_MMAP_EXEC_WRIT definition for
+       Solaris/x86.
+
+2009-12-26  Andreas Schwab  <schwab@linux-m68k.org>
+
+       * src/powerpc/ffi.c (ffi_prep_args_SYSV): Advance intarg_count
+       when a float arguments is passed in memory.
+       (ffi_closure_helper_SYSV): Mark general registers as used up when
+       a 64bit or soft-float long double argument is passed in memory.
+
+2009-12-25  Matthias Klose  <doko@ubuntu.com>
+
+       * man/ffi_call.3: Fix #include in examples.
+       * doc/libffi.texi: Add dircategory.
+
+2009-12-25  Frank Everdij <f.p.x.everdij@tudelft.nl>
+
+       * include/ffi.h.in: Placed '__GNUC__' ifdef around
+       '__attribute__((aligned(8)))' in ffi_closure, fixes compile for
+       IRIX MIPSPro c99.
+       * include/ffi_common.h: Added '__sgi' define to non
+       '__attribute__((__mode__()))' integer typedefs.
+       * src/mips/ffi.c (ffi_call, ffi_closure_mips_inner_O32,
+       ffi_closure_mips_inner_N32): Added 'defined(_MIPSEB)' to BE check.
+       (ffi_closure_mips_inner_O32, ffi_closure_mips_inner_N32): Added
+       FFI_LONGDOUBLE support and alignment(N32 only).
+       * src/mips/ffitarget.h: Corrected '#include <sgidefs.h>' for IRIX and
+       fixed non '__attribute__((__mode__()))' integer typedefs.
+       * src/mips/n32.S: Put '#ifdef linux' around '.abicalls' and '.eh_frame'
+       since they are Linux/GNU Assembler specific.
+
+2009-12-25  Bradley Smith  <brad@brad-smith.co.uk>
+
+       * configure.ac, Makefile.am, src/avr32/ffi.c,
+       src/avr32/ffitarget.h,
+       src/avr32/sysv.S: Add AVR32 port.
+       * configure, Makefile.in: Rebuilt.
+
+2009-12-21  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * configure.ac: Make i?86 build on FreeBSD and OpenBSD.
+       * configure: Regenerate.
+
+2009-12-15  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
+
+       * testsuite/libffi.call/ffitest.h: Define PRIuPTR on PA HP-UX.
+
+2009-12-13  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
+
+       * src/pa/ffi.c (ffi_closure_inner_pa32): Handle FFI_TYPE_LONGDOUBLE
+       type on HP-UX.
+
+2012-02-13  Kai Tietz  <ktietz@redhat.com>
+
+       PR libffi/52221
+       * src/x86/ffi.c (ffi_prep_raw_closure_loc): Add thiscall
+       support for X86_WIN32.
+       (FFI_INIT_TRAMPOLINE_THISCALL): Fix displacement.
+
+2009-12-11  Eric Botcazou  <ebotcazou@adacore.com>
+
+       * src/sparc/ffi.c (ffi_closure_sparc_inner_v9): Properly align 'long
+       double' arguments.
+
+2009-12-11  Eric Botcazou  <ebotcazou@adacore.com>
+
+       * testsuite/libffi.call/ffitest.h: Define PRIuPTR on Solaris < 10.
+
+2009-12-10  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
+
+       PR libffi/40700
+       * src/closures.c [X86_64 && __sun__ && __svr4__]
+       (FFI_MMAP_EXEC_WRIT): Define.
+
+2009-12-08  David Daney  <ddaney@caviumnetworks.com>
+
+       * testsuite/libffi.call/stret_medium.c: Remove xfail for mips*-*-*
+       * testsuite/libffi.call/cls_align_longdouble_split2.c: Same.
+       * testsuite/libffi.call/stret_large.c: Same.
+       * testsuite/libffi.call/cls_align_longdouble_split.c: Same.
+       * testsuite/libffi.call/stret_large2.c: Same.
+       * testsuite/libffi.call/stret_medium2.c: Same.
+
+2009-12-07  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/aix_closure.S (libffi_closure_ASM): Fix tablejump
+       typo.
+
+2009-12-05  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/aix.S: Update AIX32 code to be consistent with AIX64
+       code.
+       * src/powerpc/aix_closure.S: Same.
+
+2009-12-05  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * Makefile.in: Regenerate.
+       * configure: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * man/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2009-12-04  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/aix_closure.S: Reorganize 64-bit code to match
+       linux64_closure.S.
+
+2009-12-04  Uros Bizjak  <ubizjak@gmail.com>
+
+       PR libffi/41908
+       * src/x86/ffi64.c (classify_argument): Update from
+       gcc/config/i386/i386.c.
+       (ffi_closure_unix64_inner): Do not use the address of two consecutive
+       SSE registers directly.
+       * testsuite/libffi.call/cls_dbls_struct.c (main): Remove xfail
+       for x86_64 linux targets.
+
+2009-12-04  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/ffi_darwin.c (ffi_closure_helper_DARWIN): Increment
+       pfr for long double split between fpr13 and stack.
+
+2009-12-03  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Increment next_arg and
+       fparg_count twice for long double.
+
+2009-12-03  David Edelsohn  <edelsohn@gnu.org>
+
+       PR libffi/42243
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Remove extra parentheses.
+
+2009-12-03  Uros Bizjak  <ubizjak@gmail.com>
+
+       * testsuite/libffi.call/cls_longdouble_va.c (main): Fix format string.
+       Remove xfails for x86 linux targets.
+
+2009-12-02  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Fix typo in INT64
+       case.
+
+2009-12-01  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/aix.S (ffi_call_AIX): Convert to more standard
+       register usage.  Call ffi_prep_args directly.  Add long double
+       return value support.
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Double arg increment
+       applies to FFI_TYPE_DOUBLE.  Correct fpr_base increment typo.
+       Separate FFI_TYPE_SINT32 and FFI_TYPE_UINT32 cases.
+       (ffi_prep_cif_machdep): Only 16 byte stack alignment in 64 bit
+       mode.
+       (ffi_closure_helper_DARWIN): Remove nf and ng counters.  Move temp
+       into case.
+       * src/powerpc/aix_closure.S: Maintain 16 byte stack alignment.
+       Allocate result area between params and FPRs.
+
+2009-11-30  David Edelsohn  <edelsohn@gnu.org>
+
+       PR target/35484
+       * src/powerpc/ffitarget.h (POWERPC64): Define for PPC64 Linux and
+       AIX64.
+       * src/powerpc/aix.S: Implement AIX64 version.
+       * src/powerpc/aix_closure.S: Implement AIX64 version.
+       (ffi_closure_ASM): Use extsb, lha and displament addresses.
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Implement AIX64
+       support.
+       (ffi_prep_cif_machdep): Same.
+       (ffi_call): Same.
+       (ffi_closure_helper_DARWIN): Same.
+
+2009-11-02  Andreas Tobler  <a.tobler@schweiz.org>
+
+       PR libffi/41908
+       * testsuite/libffi.call/testclosure.c: New test.
+
+2009-09-28  Kai Tietz  <kai.tietz@onevision.com>
+
+       * src/x86/win64.S (_ffi_call_win64 stack): Remove for gnu
+       assembly version use of ___chkstk.
+
+2009-09-23  Matthias Klose  <doko@ubuntu.com>
+
+       PR libffi/40242, PR libffi/41443
+       * src/arm/sysv.S (__ARM_ARCH__): Define for processors
+       __ARM_ARCH_6T2__, __ARM_ARCH_6M__, __ARM_ARCH_7__,
+       __ARM_ARCH_7A__, __ARM_ARCH_7R__, __ARM_ARCH_7M__.
+       Change the conditionals to __SOFTFP__ || __ARM_EABI__
+       for -mfloat-abi=softfp to work.
+
+2009-09-17  Loren J. Rittle  <ljrittle@acm.org>
+
+       PR testsuite/32843 (strikes again)
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Add X86_FREEBSD to
+       enable proper extension on char and short.
+
+2009-09-15  David Daney  <ddaney@caviumnetworks.com>
+
+       * src/java_raw_api.c (ffi_java_raw_to_rvalue): Remove special
+       handling for FFI_TYPE_POINTER.
+       * src/mips/ffitarget.h (FFI_TYPE_STRUCT_D_SOFT,
+       FFI_TYPE_STRUCT_F_SOFT, FFI_TYPE_STRUCT_DD_SOFT,
+       FFI_TYPE_STRUCT_FF_SOFT, FFI_TYPE_STRUCT_FD_SOFT,
+       FFI_TYPE_STRUCT_DF_SOFT, FFI_TYPE_STRUCT_SOFT): New defines.
+       (FFI_N32_SOFT_FLOAT, FFI_N64_SOFT_FLOAT): New ffi_abi enumerations.
+       (enum ffi_abi): Set FFI_DEFAULT_ABI for soft-float.
+       * src/mips/n32.S (ffi_call_N32): Add handling for soft-float
+       structure and pointer returns.
+       (ffi_closure_N32): Add handling for pointer returns.
+       * src/mips/ffi.c (ffi_prep_args, calc_n32_struct_flags,
+       calc_n32_return_struct_flags): Handle soft-float.
+       (ffi_prep_cif_machdep):  Handle soft-float, fix pointer handling.
+       (ffi_call_N32): Declare proper argument types.
+       (ffi_call, copy_struct_N32, ffi_closure_mips_inner_N32): Handle
+       soft-float.
+
+2009-08-24  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure.ac (AC_PREREQ): Bump to 2.64.
+
+2009-08-22  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * Makefile.am (install-html, install-pdf): Remove.
+       * Makefile.in: Regenerate.
+
+       * Makefile.in: Regenerate.
+       * aclocal.m4: Regenerate.
+       * configure: Regenerate.
+       * fficonfig.h.in: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * man/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2011-08-22  Jasper Lievisse Adriaanse <jasper@openbsd.org>
+
+       * configure.ac: Add OpenBSD/hppa and OpenBSD/powerpc support.
+       * configure: Rebuilt.
+
+2009-07-30  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure.ac (_AC_ARG_VAR_PRECIOUS): Use m4_rename_force.
+
+2009-07-24  Dave Korn  <dave.korn.cygwin@gmail.com>
+
+       PR libffi/40807
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Also use sign/zero-extending
+       return types for X86_WIN32.
+       * src/x86/win32.S (_ffi_call_SYSV): Handle omitted return types.
+       (_ffi_call_STDCALL, _ffi_closure_SYSV, _ffi_closure_raw_SYSV,
+       _ffi_closure_STDCALL): Likewise.
+
+       * src/closures.c (is_selinux_enabled): Define to const 0 for Cygwin.
+       (dlmmap, dlmunmap): Also use these functions on Cygwin.
+
+2009-07-11  Richard Sandiford  <rdsandiford@googlemail.com>
+
+       PR testsuite/40699
+       PR testsuite/40707
+       PR testsuite/40709
+       * testsuite/lib/libffi-dg.exp: Revert 2009-07-02, 2009-07-01 and
+       2009-06-30 commits.
+
+2009-07-01  Richard Sandiford  <r.sandiford@uk.ibm.com>
+
+       * testsuite/lib/libffi-dg.exp (libffi-init): Set ld_library_path
+       to "" before adding paths.  (This reinstates an assignment that
+       was removed by my 2009-06-30 commit, but changes the initial
+       value from "." to "".)
+
+2009-07-01  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR testsuite/40601
+       * testsuite/lib/libffi-dg.exp (libffi-init): Properly set
+       gccdir.  Adjust ld_library_path for gcc only if gccdir isn't
+       empty.
+
+2009-06-30  Richard Sandiford  <r.sandiford@uk.ibm.com>
+
+       * testsuite/lib/libffi-dg.exp (libffi-init): Don't add "."
+       to ld_library_path.  Use add_path.  Add just find_libgcc_s
+       to ld_library_path, not every libgcc multilib directory.
+
+2009-06-16  Wim Lewis  <wiml@hhhh.org>
+
+       * src/powerpc/ffi.c: Avoid clobbering cr3 and cr4, which are
+       supposed to be callee-saved.
+       * src/powerpc/sysv.S (small_struct_return_value): Fix overrun of
+       return buffer for odd-size structs.
+
+2009-06-16  Andreas Tobler  <a.tobler@schweiz.org>
+
+       PR libffi/40444
+       * testsuite/lib/libffi-dg.exp (libffi_target_compile): Add
+       allow_stack_execute for Darwin.
+
+2009-06-16  Andrew Haley  <aph@redhat.com>
+
+       * configure.ac (TARGETDIR): Add missing blank lines.
+       * configure: Regenerate.
+
+2009-06-16  Andrew Haley  <aph@redhat.com>
+
+       * testsuite/libffi.call/cls_align_sint64.c,
+       testsuite/libffi.call/cls_align_uint64.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/cls_ulonglong.c,
+       testsuite/libffi.call/return_ll1.c,
+       testsuite/libffi.call/stret_medium2.c: Fix printf format
+       specifiers.
+       * testsuite/libffi.call/ffitest.h,
+       testsuite/libffi.special/ffitestcxx.h (PRIdLL, PRIuLL): Define.
+
+2009-06-15  Andrew Haley  <aph@redhat.com>
+
+       * testsuite/libffi.call/err_bad_typedef.c: xfail everywhere.
+       * testsuite/libffi.call/err_bad_abi.c: Likewise.
+
+2009-06-12  Andrew Haley  <aph@redhat.com>
+
+       * Makefile.am: Remove info_TEXINFOS.
+
+2009-06-12  Andrew Haley  <aph@redhat.com>
+
+       * ChangeLog.libffi: testsuite/libffi.call/cls_align_sint64.c,
+       testsuite/libffi.call/cls_align_uint64.c,
+       testsuite/libffi.call/cls_ulonglong.c,
+       testsuite/libffi.call/return_ll1.c,
+       testsuite/libffi.call/stret_medium2.c: Fix printf format
+       specifiers.
+       testsuite/libffi.special/unwindtest.cc: include stdint.h.
+
+2009-06-11  Timothy Wall  <twall@users.sf.net>
+
+       * Makefile.am,
+       configure.ac,
+       include/ffi.h.in,
+       include/ffi_common.h,
+       src/closures.c,
+       src/dlmalloc.c,
+       src/x86/ffi.c,
+       src/x86/ffitarget.h,
+       src/x86/win64.S (new),
+       README: Added win64 support (mingw or MSVC)
+       * Makefile.in,
+       include/Makefile.in,
+       man/Makefile.in,
+       testsuite/Makefile.in,
+       configure,
+       aclocal.m4: Regenerated
+       * ltcf-c.sh: properly escape cygwin/w32 path
+       * man/ffi_call.3: Clarify size requirements for return value.
+       * src/x86/ffi64.c: Fix filename in comment.
+       * src/x86/win32.S: Remove unused extern.
+
+       * testsuite/libffi.call/closure_fn0.c,
+       testsuite/libffi.call/closure_fn1.c,
+       testsuite/libffi.call/closure_fn2.c,
+       testsuite/libffi.call/closure_fn3.c,
+       testsuite/libffi.call/closure_fn4.c,
+       testsuite/libffi.call/closure_fn5.c,
+       testsuite/libffi.call/closure_fn6.c,
+       testsuite/libffi.call/closure_stdcall.c,
+       testsuite/libffi.call/cls_12byte.c,
+       testsuite/libffi.call/cls_16byte.c,
+       testsuite/libffi.call/cls_18byte.c,
+       testsuite/libffi.call/cls_19byte.c,
+       testsuite/libffi.call/cls_1_1byte.c,
+       testsuite/libffi.call/cls_20byte.c,
+       testsuite/libffi.call/cls_20byte1.c,
+       testsuite/libffi.call/cls_24byte.c,
+       testsuite/libffi.call/cls_2byte.c,
+       testsuite/libffi.call/cls_3_1byte.c,
+       testsuite/libffi.call/cls_3byte1.c,
+       testsuite/libffi.call/cls_3byte2.c,
+       testsuite/libffi.call/cls_4_1byte.c,
+       testsuite/libffi.call/cls_4byte.c,
+       testsuite/libffi.call/cls_5_1_byte.c,
+       testsuite/libffi.call/cls_5byte.c,
+       testsuite/libffi.call/cls_64byte.c,
+       testsuite/libffi.call/cls_6_1_byte.c,
+       testsuite/libffi.call/cls_6byte.c,
+       testsuite/libffi.call/cls_7_1_byte.c,
+       testsuite/libffi.call/cls_7byte.c,
+       testsuite/libffi.call/cls_8byte.c,
+       testsuite/libffi.call/cls_9byte1.c,
+       testsuite/libffi.call/cls_9byte2.c,
+       testsuite/libffi.call/cls_align_double.c,
+       testsuite/libffi.call/cls_align_float.c,
+       testsuite/libffi.call/cls_align_longdouble.c,
+       testsuite/libffi.call/cls_align_longdouble_split.c,
+       testsuite/libffi.call/cls_align_longdouble_split2.c,
+       testsuite/libffi.call/cls_align_pointer.c,
+       testsuite/libffi.call/cls_align_sint16.c,
+       testsuite/libffi.call/cls_align_sint32.c,
+       testsuite/libffi.call/cls_align_sint64.c,
+       testsuite/libffi.call/cls_align_uint16.c,
+       testsuite/libffi.call/cls_align_uint32.c,
+       testsuite/libffi.call/cls_align_uint64.c,
+       testsuite/libffi.call/cls_dbls_struct.c,
+       testsuite/libffi.call/cls_double.c,
+       testsuite/libffi.call/cls_double_va.c,
+       testsuite/libffi.call/cls_float.c,
+       testsuite/libffi.call/cls_longdouble.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/cls_multi_schar.c,
+       testsuite/libffi.call/cls_multi_sshort.c,
+       testsuite/libffi.call/cls_multi_sshortchar.c,
+       testsuite/libffi.call/cls_multi_uchar.c,
+       testsuite/libffi.call/cls_multi_ushort.c,
+       testsuite/libffi.call/cls_multi_ushortchar.c,
+       testsuite/libffi.call/cls_pointer.c,
+       testsuite/libffi.call/cls_pointer_stack.c,
+       testsuite/libffi.call/cls_schar.c,
+       testsuite/libffi.call/cls_sint.c,
+       testsuite/libffi.call/cls_sshort.c,
+       testsuite/libffi.call/cls_uchar.c,
+       testsuite/libffi.call/cls_uint.c,
+       testsuite/libffi.call/cls_ulonglong.c,
+       testsuite/libffi.call/cls_ushort.c,
+       testsuite/libffi.call/err_bad_abi.c,
+       testsuite/libffi.call/err_bad_typedef.c,
+       testsuite/libffi.call/float2.c,
+       testsuite/libffi.call/huge_struct.c,
+       testsuite/libffi.call/nested_struct.c,
+       testsuite/libffi.call/nested_struct1.c,
+       testsuite/libffi.call/nested_struct10.c,
+       testsuite/libffi.call/nested_struct2.c,
+       testsuite/libffi.call/nested_struct3.c,
+       testsuite/libffi.call/nested_struct4.c,
+       testsuite/libffi.call/nested_struct5.c,
+       testsuite/libffi.call/nested_struct6.c,
+       testsuite/libffi.call/nested_struct7.c,
+       testsuite/libffi.call/nested_struct8.c,
+       testsuite/libffi.call/nested_struct9.c,
+       testsuite/libffi.call/problem1.c,
+       testsuite/libffi.call/return_ldl.c,
+       testsuite/libffi.call/return_ll1.c,
+       testsuite/libffi.call/stret_large.c,
+       testsuite/libffi.call/stret_large2.c,
+       testsuite/libffi.call/stret_medium.c,
+       testsuite/libffi.call/stret_medium2.c,
+       testsuite/libffi.special/unwindtest.cc: use ffi_closure_alloc instead
+       of checking for MMAP.  Use intptr_t instead of long casts.
+
+2009-06-11  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * testsuite/libffi.call/cls_longdouble_va.c: Add xfail sh*-*-linux-*.
+       * testsuite/libffi.call/err_bad_abi.c: Add xfail sh*-*-*.
+       * testsuite/libffi.call/err_bad_typedef.c: Likewise.
+
+2009-06-09  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/freebsd.S: Add missing file.
+
+2009-06-08  Andrew Haley  <aph@redhat.com>
+
+       Import from libffi 3.0.8:
+
+       * doc/libffi.texi: New file.
+       * doc/libffi.info: Likewise.
+       * doc/stamp-vti: Likewise.
+       * man/Makefile.am: New file.
+       * man/ffi_call.3: New file.
+
+       * Makefile.am (EXTRA_DIST): Add src/x86/darwin64.S,
+       src/dlmalloc.c.
+       (nodist_libffi_la_SOURCES): Add X86_FREEBSD.
+
+       * configure.ac: Bump version to 3.0.8.
+       parisc*-*-linux*: Add.
+       i386-*-freebsd* | i386-*-openbsd*: Add.
+       powerpc-*-beos*: Add.
+       AM_CONDITIONAL X86_FREEBSD: Add.
+       AC_CONFIG_FILES: Add man/Makefile.
+
+       * include/ffi.h.in (FFI_FN): Change void (*)() to void (*)(void).
+
+2009-06-08  Andrew Haley  <aph@redhat.com>
+
+       * README: Import from libffi 3.0.8.
+
+2009-06-08  Andrew Haley  <aph@redhat.com>
+
+       * testsuite/libffi.call/err_bad_abi.c: Add xfails.
+       * testsuite/libffi.call/cls_longdouble_va.c: Add xfails.
+       * testsuite/libffi.call/cls_dbls_struct.c: Add xfail x86_64-*-linux-*.
+       * testsuite/libffi.call/err_bad_typedef.c: Add xfails.
+
+       * testsuite/libffi.call/stret_medium2.c: Add __UNUSED__ to args.
+       * testsuite/libffi.call/stret_medium.c: Likewise.
+       * testsuite/libffi.call/stret_large2.c: Likewise.
+       * testsuite/libffi.call/stret_large.c:  Likewise.
+
+2008-12-26  Timothy Wall  <twall@users.sf.net>
+
+       * testsuite/libffi.call/cls_longdouble.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/cls_align_longdouble.c,
+       testsuite/libffi.call/cls_align_longdouble_split.c,
+       testsuite/libffi.call/cls_align_longdouble_split2.c: mark expected
+       failures on x86_64 cygwin/mingw.
+
+2008-12-22  Timothy Wall  <twall@users.sf.net>
+
+       * testsuite/libffi.call/closure_fn0.c,
+       testsuite/libffi.call/closure_fn1.c,
+       testsuite/libffi.call/closure_fn2.c,
+       testsuite/libffi.call/closure_fn3.c,
+       testsuite/libffi.call/closure_fn4.c,
+       testsuite/libffi.call/closure_fn5.c,
+       testsuite/libffi.call/closure_fn6.c,
+       testsuite/libffi.call/closure_loc_fn0.c,
+       testsuite/libffi.call/closure_stdcall.c,
+       testsuite/libffi.call/cls_align_pointer.c,
+       testsuite/libffi.call/cls_pointer.c,
+       testsuite/libffi.call/cls_pointer_stack.c: use portable cast from
+       pointer to integer (intptr_t).
+       * testsuite/libffi.call/cls_longdouble.c: disable for win64.
+
+2008-07-24  Anthony Green  <green@redhat.com>
+
+       * testsuite/libffi.call/cls_dbls_struct.c,
+       testsuite/libffi.call/cls_double_va.c,
+       testsuite/libffi.call/cls_longdouble.c,
+       testsuite/libffi.call/cls_longdouble_va.c,
+       testsuite/libffi.call/cls_pointer.c,
+       testsuite/libffi.call/cls_pointer_stack.c,
+       testsuite/libffi.call/err_bad_abi.c: Clean up failures from
+       compiler warnings.
+
+2008-03-04  Anthony Green  <green@redhat.com>
+           Blake Chaffin
+           hos@tamanegi.org
+
+       * testsuite/libffi.call/cls_align_longdouble_split2.c
+         testsuite/libffi.call/cls_align_longdouble_split.c
+         testsuite/libffi.call/cls_dbls_struct.c
+         testsuite/libffi.call/cls_double_va.c
+         testsuite/libffi.call/cls_longdouble.c
+         testsuite/libffi.call/cls_longdouble_va.c
+         testsuite/libffi.call/cls_pointer.c
+         testsuite/libffi.call/cls_pointer_stack.c
+         testsuite/libffi.call/err_bad_abi.c
+         testsuite/libffi.call/err_bad_typedef.c
+         testsuite/libffi.call/stret_large2.c
+         testsuite/libffi.call/stret_large.c
+         testsuite/libffi.call/stret_medium2.c
+         testsuite/libffi.call/stret_medium.c: New tests from Apple.
+
+2009-06-05  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/ffitarget.h, src/x86/ffi.c: Merge stdcall changes from
+       libffi.
+
+2009-06-04  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/ffitarget.h, src/x86/win32.S, src/x86/ffi.c: Back out
+       stdcall changes.
+
+2008-02-26  Anthony Green  <green@redhat.com>
+           Thomas Heller  <theller@ctypes.org>
+
+       * src/x86/ffi.c (ffi_closure_SYSV_inner): Change C++ comment to C
+       comment.
+
+2008-02-03  Timothy Wall  <twall@users.sf.net>
+
+       * src/x86/ffi.c (FFI_INIT_TRAMPOLINE_STDCALL): Calculate jump return
+         offset based on code pointer, not data pointer.
+
+2008-01-31  Timothy Wall <twall@users.sf.net>
+
+       * testsuite/libffi.call/closure_stdcall.c: Add test for stdcall
+       closures.
+       * src/x86/ffitarget.h: Increase size of trampoline for stdcall
+       closures.
+       * src/x86/win32.S: Add assembly for stdcall closure.
+       * src/x86/ffi.c: Initialize stdcall closure trampoline.
+
+2009-06-04  Andrew Haley  <aph@redhat.com>
+
+       * include/ffi.h.in: Change void (*)() to void (*)(void).
+       * src/x86/ffi.c: Likewise.
+
+2009-06-04  Andrew Haley  <aph@redhat.com>
+
+       * src/powerpc/ppc_closure.S: Insert licence header.
+       * src/powerpc/linux64_closure.S: Likewise.
+       * src/m68k/sysv.S: Likewise.
+
+       * src/sh64/ffi.c: Change void (*)() to void (*)(void).
+       * src/powerpc/ffi.c: Likewise.
+       * src/powerpc/ffi_darwin.c: Likewise.
+       * src/m32r/ffi.c: Likewise.
+       * src/sh64/ffi.c: Likewise.
+       * src/x86/ffi64.c: Likewise.
+       * src/alpha/ffi.c: Likewise.
+       * src/alpha/osf.S: Likewise.
+       * src/frv/ffi.c: Likewise.
+       * src/s390/ffi.c: Likewise.
+       * src/pa/ffi.c: Likewise.
+       * src/pa/hpux32.S: Likewise.
+       * src/ia64/unix.S: Likewise.
+       * src/ia64/ffi.c: Likewise.
+       * src/sparc/ffi.c: Likewise.
+       * src/mips/ffi.c: Likewise.
+       * src/sh/ffi.c: Likewise.
+
+2008-02-15  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/ffi.c (USE__BUILTIN___CLEAR_CACHE):
+       Define (conditionally), and use it to include cachectl.h.
+       (ffi_prep_closure_loc): Fix cache flushing.
+       * src/mips/ffitarget.h (_ABIN32, _ABI64, _ABIO32): Define.
+
+2009-06-04  Andrew Haley  <aph@redhat.com>
+
+       include/ffi.h.in,
+       src/arm/ffitarget.h,
+       src/arm/ffi.c,
+       src/arm/sysv.S,
+       src/powerpc/ffitarget.h,
+       src/closures.c,
+       src/sh64/ffitarget.h,
+       src/sh64/ffi.c,
+       src/sh64/sysv.S,
+       src/types.c,
+       src/x86/ffi64.c,
+       src/x86/ffitarget.h,
+       src/x86/win32.S,
+       src/x86/darwin.S,
+       src/x86/ffi.c,
+       src/x86/sysv.S,
+       src/x86/unix64.S,
+       src/alpha/ffitarget.h,
+       src/alpha/ffi.c,
+       src/alpha/osf.S,
+       src/m68k/ffitarget.h,
+       src/frv/ffitarget.h,
+       src/frv/ffi.c,
+       src/s390/ffitarget.h,
+       src/s390/sysv.S,
+       src/cris/ffitarget.h,
+       src/pa/linux.S,
+       src/pa/ffitarget.h,
+       src/pa/ffi.c,
+       src/raw_api.c,
+       src/ia64/ffitarget.h,
+       src/ia64/unix.S,
+       src/ia64/ffi.c,
+       src/ia64/ia64_flags.h,
+       src/java_raw_api.c,
+       src/debug.c,
+       src/sparc/v9.S,
+       src/sparc/ffitarget.h,
+       src/sparc/ffi.c,
+       src/sparc/v8.S,
+       src/mips/ffitarget.h,
+       src/mips/n32.S,
+       src/mips/o32.S,
+       src/mips/ffi.c,
+       src/prep_cif.c,
+       src/sh/ffitarget.h,
+       src/sh/ffi.c,
+       src/sh/sysv.S: Update license text.
+
+2009-05-22  Dave Korn  <dave.korn.cygwin@gmail.com>
+
+       * src/x86/win32.S (_ffi_closure_STDCALL):  New function.
+       (.eh_frame):  Add FDE for it.
+
+2009-05-22  Dave Korn  <dave.korn.cygwin@gmail.com>
+
+       * configure.ac:  Also check if assembler supports pc-relative
+       relocs on X86_WIN32 targets.
+       * configure:  Regenerate.
+       * src/x86/win32.S (ffi_prep_args):  Declare extern, not global.
+       (_ffi_call_SYSV):  Add missing function type symbol .def and
+       add EH markup labels.
+       (_ffi_call_STDCALL):  Likewise.
+       (_ffi_closure_SYSV):  Likewise.
+       (_ffi_closure_raw_SYSV):  Likewise.
+       (.eh_frame):  Add hand-crafted EH data.
+
+2009-04-09  Jakub Jelinek  <jakub@redhat.com>
+
+       * testsuite/lib/libffi-dg.exp: Change copyright header to refer to
+       version 3 of the GNU General Public License and to point readers
+       at the COPYING3 file and the FSF's license web page.
+       * testsuite/libffi.call/call.exp: Likewise.
+       * testsuite/libffi.special/special.exp: Likewise.
+
+2009-03-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure: Regenerate.
+
+2008-12-18  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       PR libffi/26048
+       * configure.ac (HAVE_AS_X86_PCREL): New test.
+       * configure: Regenerate.
+       * fficonfig.h.in: Regenerate.
+       * src/x86/sysv.S [!FFI_NO_RAW_API]: Precalculate
+       RAW_CLOSURE_CIF_OFFSET, RAW_CLOSURE_FUN_OFFSET,
+       RAW_CLOSURE_USER_DATA_OFFSET for the Solaris 10/x86 assembler.
+       (.eh_frame): Only use SYMBOL-. iff HAVE_AS_X86_PCREL.
+       * src/x86/unix64.S (.Lstore_table): Move to .text section.
+       (.Lload_table): Likewise.
+       (.eh_frame): Only use SYMBOL-. iff HAVE_AS_X86_PCREL.
+
+2008-12-18  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure: Regenerate.
+
+2008-11-21  Eric Botcazou  <ebotcazou@adacore.com>
+
+       * src/sparc/ffi.c (ffi_prep_cif_machdep): Add support for
+       signed/unsigned int8/16 return values.
+       * src/sparc/v8.S (ffi_call_v8): Likewise.
+       (ffi_closure_v8): Likewise.
+
+2008-09-26  Peter O'Gorman  <pogma@thewrittenword.com>
+           Steve Ellcey  <sje@cup.hp.com>
+
+       * configure: Regenerate for new libtool.
+       * Makefile.in: Ditto.
+       * include/Makefile.in: Ditto.
+       * aclocal.m4: Ditto.
+
+2008-08-25  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * src/powerpc/ffitarget.h (ffi_abi): Add FFI_LINUX and
+       FFI_LINUX_SOFT_FLOAT to the POWERPC_FREEBSD enum.
+       Add note about flag bits used for FFI_SYSV_TYPE_SMALL_STRUCT.
+       Adjust copyright notice.
+       * src/powerpc/ffi.c: Add two new flags to indicate if we have one
+       register or two register to use for FFI_SYSV structs.
+       (ffi_prep_cif_machdep): Pass the right register flag introduced above.
+       (ffi_closure_helper_SYSV): Fix the return type for
+       FFI_SYSV_TYPE_SMALL_STRUCT. Comment.
+       Adjust copyright notice.
+
+2008-07-16  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_prep_closure_loc): Turn INSN into an unsigned
+       int.
+
+2008-06-17  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * configure: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2008-06-07  Joseph Myers  <joseph@codesourcery.com>
+
+       * configure.ac (parisc*-*-linux*, powerpc-*-sysv*,
+       powerpc-*-beos*): Remove.
+       * configure: Regenerate.
+
+2008-05-09  Julian Brown  <julian@codesourcery.com>
+
+       * Makefile.am (LTLDFLAGS): New.
+       (libffi_la_LDFLAGS): Use above.
+       * Makefile.in: Regenerate.
+
+2008-04-18  Paolo Bonzini  <bonzini@gnu.org>
+
+       PR bootstrap/35457
+       * aclocal.m4: Regenerate.
+       * configure: Regenerate.
+
+2008-03-26  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/sysv.S: Add .note.GNU-stack on Linux.
+       * src/sh64/sysv.S: Likewise.
+
+2008-03-26  Daniel Jacobowitz  <dan@debian.org>
+
+       * src/arm/sysv.S: Fix ARM comment marker.
+
+2008-03-26  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/alpha/osf.S: Add .note.GNU-stack on Linux.
+       * src/s390/sysv.S: Likewise.
+       * src/powerpc/ppc_closure.S: Likewise.
+       * src/powerpc/sysv.S: Likewise.
+       * src/x86/unix64.S: Likewise.
+       * src/x86/sysv.S: Likewise.
+       * src/sparc/v8.S: Likewise.
+       * src/sparc/v9.S: Likewise.
+       * src/m68k/sysv.S: Likewise.
+       * src/arm/sysv.S: Likewise.
+
+2008-03-16  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
+
+       * aclocal.m4: Regenerate.
+       * configure: Likewise.
+       * Makefile.in: Likewise.
+       * include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+
+2008-02-12  Bjoern Koenig  <bkoenig@alpha-tierchen.de>
+           Andreas Tobler  <a.tobler@schweiz.org>
+
+       * configure.ac: Add amd64-*-freebsd* target.
+       * configure: Regenerate.
+
+2008-01-30  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR libffi/34612
+       * src/x86/sysv.S (ffi_closure_SYSV): Pop 4 byte from stack when
+       returning struct.
+
+       * testsuite/libffi.call/call.exp: Add "-O2 -fomit-frame-pointer"
+       tests.
+
+2008-01-24  David Edelsohn  <edelsohn@gnu.org>
+
+       * configure: Regenerate.
+
+2008-01-06  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Fix thinko.
+
+2008-01-05  Andreas Tobler  <a.tobler@schweiz.org>
+
+       PR testsuite/32843
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Add code for
+       signed/unsigned int8/16 for X86_DARWIN.
+       Updated copyright info.
+       Handle one and two byte structs with special cif->flags.
+       * src/x86/ffitarget.h: Add special types for one and two byte structs.
+       Updated copyright info.
+       * src/x86/darwin.S (ffi_call_SYSV): Rewrite to use a jump table like
+       sysv.S
+       Remove code to pop args from the stack after call.
+       Special-case signed/unsigned for int8/16, one and two byte structs.
+       (ffi_closure_raw_SYSV): Handle FFI_TYPE_UINT8,
+       FFI_TYPE_SINT8, FFI_TYPE_UINT16, FFI_TYPE_SINT16, FFI_TYPE_UINT32,
+       FFI_TYPE_SINT32.
+       Updated copyright info.
+
+2007-12-08  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/n32.S (ffi_call_N32):  Replace dadd with ADDU, dsub with
+       SUBU, add with ADDU and use smaller code sequences.
+
+2007-12-07  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/ffi.c (ffi_prep_cif_machdep): Handle long double return
+       type.
+
+2007-12-06  David Daney  <ddaney@avtrex.com>
+
+       * include/ffi.h.in (FFI_SIZEOF_JAVA_RAW): Define if not already
+       defined.
+       (ffi_java_raw): New typedef.
+       (ffi_java_raw_call, ffi_java_ptrarray_to_raw,
+       ffi_java_raw_to_ptrarray): Change parameter types from ffi_raw to
+       ffi_java_raw.
+       (ffi_java_raw_closure) : Same.
+       (ffi_prep_java_raw_closure, ffi_prep_java_raw_closure_loc): Change
+       parameter types.
+       * src/java_raw_api.c (ffi_java_raw_size):  Replace FFI_SIZEOF_ARG with
+       FFI_SIZEOF_JAVA_RAW.
+       (ffi_java_raw_to_ptrarray): Change type of raw to ffi_java_raw.
+       Replace FFI_SIZEOF_ARG with FFI_SIZEOF_JAVA_RAW. Use
+       sizeof(ffi_java_raw) for alignment calculations.
+       (ffi_java_ptrarray_to_raw): Same.
+       (ffi_java_rvalue_to_raw): Add special handling for FFI_TYPE_POINTER
+       if FFI_SIZEOF_JAVA_RAW == 4.
+       (ffi_java_raw_to_rvalue): Same.
+       (ffi_java_raw_call): Change type of raw to ffi_java_raw.
+       (ffi_java_translate_args): Same.
+       (ffi_prep_java_raw_closure_loc, ffi_prep_java_raw_closure): Change
+       parameter types.
+       * src/mips/ffitarget.h (FFI_SIZEOF_JAVA_RAW): Define for N32 ABI.
+
+2007-12-06  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/n32.S (ffi_closure_N32): Use 64-bit add instruction on
+       pointer values.
+
+2007-12-01  Andreas Tobler  <a.tobler@schweiz.org>
+
+       PR libffi/31937
+       * src/powerpc/ffitarget.h: Introduce new ABI FFI_LINUX_SOFT_FLOAT.
+       Add local FFI_TYPE_UINT128 to handle soft-float long-double-128.
+       * src/powerpc/ffi.c: Distinguish between __NO_FPRS__ and not and
+       set the NUM_FPR_ARG_REGISTERS according to.
+       Add support for potential soft-float support under hard-float
+       architecture.
+       (ffi_prep_args_SYSV): Set NUM_FPR_ARG_REGISTERS to 0 in case of
+       FFI_LINUX_SOFT_FLOAT, handle float, doubles and long-doubles according
+       to the FFI_LINUX_SOFT_FLOAT ABI.
+       (ffi_prep_cif_machdep): Likewise.
+       (ffi_closure_helper_SYSV): Likewise.
+       * src/powerpc/ppc_closure.S: Make sure not to store float/double
+       on archs where __NO_FPRS__ is true.
+       Add FFI_TYPE_UINT128 support.
+       * src/powerpc/sysv.S: Add support for soft-float long-double-128.
+       Adjust copyright notice.
+
+2007-11-25  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * src/closures.c: Move defintion of MAYBE_UNUSED from here to ...
+       * include/ffi_common.h: ... here.
+       Update copyright.
+
+2007-11-17  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * src/powerpc/sysv.S: Load correct cr to compare if we have long double.
+       * src/powerpc/linux64.S: Likewise.
+       * src/powerpc/ffi.c: Add a comment to show which part goes into cr6.
+       * testsuite/libffi.call/return_ldl.c: New test.
+
+2007-09-04    <aph@redhat.com>
+
+       * src/arm/sysv.S (UNWIND): New.
+       (Whole file): Conditionally compile unwinder directives.
+       * src/arm/sysv.S: Add unwinder directives.
+
+       * src/arm/ffi.c (ffi_prep_args): Align structs by at least 4 bytes.
+       Only treat r0 as a struct address if we're actually returning a
+       struct by address.
+       Only copy the bytes that are actually within a struct.
+       (ffi_prep_cif_machdep): A Composite Type not larger than 4 bytes
+       is returned in r0, not passed by address.
+       (ffi_call): Allocate a word-sized temporary for the case where
+       a composite is returned in r0.
+       (ffi_prep_incoming_args_SYSV): Align as necessary.
+
+2007-08-05  Steven Newbury  <s_j_newbury@yahoo.co.uk>
+
+       * src/arm/ffi.c (FFI_INIT_TRAMPOLINE): Use __clear_cache instead of
+       directly using the sys_cacheflush syscall.
+
+2007-07-27  Andrew Haley  <aph@redhat.com>
+
+       * src/arm/sysv.S (ffi_closure_SYSV): Add soft-float.
+
+2007-09-03  Maciej W. Rozycki  <macro@linux-mips.org>
+
+       * Makefile.am: Unify MIPS_IRIX and MIPS_LINUX into MIPS.
+       * configure.ac: Likewise.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+       * configure: Likewise.
+
+2007-08-24  David Daney  <ddaney@avtrex.com>
+
+       * testsuite/libffi.call/return_sl.c: New test.
+
+2007-08-10  David Daney  <ddaney@avtrex.com>
+
+       * testsuite/libffi.call/cls_multi_ushort.c,
+       testsuite/libffi.call/cls_align_uint16.c,
+       testsuite/libffi.call/nested_struct1.c,
+       testsuite/libffi.call/nested_struct3.c,
+       testsuite/libffi.call/cls_7_1_byte.c,
+       testsuite/libffi.call/nested_struct5.c,
+       testsuite/libffi.call/cls_double.c,
+       testsuite/libffi.call/nested_struct7.c,
+       testsuite/libffi.call/cls_sint.c,
+       testsuite/libffi.call/nested_struct9.c,
+       testsuite/libffi.call/cls_20byte1.c,
+       testsuite/libffi.call/cls_multi_sshortchar.c,
+       testsuite/libffi.call/cls_align_sint64.c,
+       testsuite/libffi.call/cls_3byte2.c,
+       testsuite/libffi.call/cls_multi_schar.c,
+       testsuite/libffi.call/cls_multi_uchar.c,
+       testsuite/libffi.call/cls_19byte.c,
+       testsuite/libffi.call/cls_9byte1.c,
+       testsuite/libffi.call/cls_align_float.c,
+       testsuite/libffi.call/closure_fn1.c,
+       testsuite/libffi.call/problem1.c,
+       testsuite/libffi.call/closure_fn3.c,
+       testsuite/libffi.call/cls_sshort.c,
+       testsuite/libffi.call/closure_fn5.c,
+       testsuite/libffi.call/cls_align_double.c,
+       testsuite/libffi.call/nested_struct.c,
+       testsuite/libffi.call/cls_2byte.c,
+       testsuite/libffi.call/nested_struct10.c,
+       testsuite/libffi.call/cls_4byte.c,
+       testsuite/libffi.call/cls_6byte.c,
+       testsuite/libffi.call/cls_8byte.c,
+       testsuite/libffi.call/cls_multi_sshort.c,
+       testsuite/libffi.call/cls_align_sint16.c,
+       testsuite/libffi.call/cls_align_uint32.c,
+       testsuite/libffi.call/cls_20byte.c,
+       testsuite/libffi.call/cls_float.c,
+       testsuite/libffi.call/nested_struct2.c,
+       testsuite/libffi.call/cls_5_1_byte.c,
+       testsuite/libffi.call/nested_struct4.c,
+       testsuite/libffi.call/cls_24byte.c,
+       testsuite/libffi.call/nested_struct6.c,
+       testsuite/libffi.call/cls_64byte.c,
+       testsuite/libffi.call/nested_struct8.c,
+       testsuite/libffi.call/cls_uint.c,
+       testsuite/libffi.call/cls_multi_ushortchar.c,
+       testsuite/libffi.call/cls_schar.c,
+       testsuite/libffi.call/cls_uchar.c,
+       testsuite/libffi.call/cls_align_uint64.c,
+       testsuite/libffi.call/cls_ulonglong.c,
+       testsuite/libffi.call/cls_align_longdouble.c,
+       testsuite/libffi.call/cls_1_1byte.c,
+       testsuite/libffi.call/cls_12byte.c,
+       testsuite/libffi.call/cls_3_1byte.c,
+       testsuite/libffi.call/cls_3byte1.c,
+       testsuite/libffi.call/cls_4_1byte.c,
+       testsuite/libffi.call/cls_6_1_byte.c,
+       testsuite/libffi.call/cls_16byte.c,
+       testsuite/libffi.call/cls_18byte.c,
+       testsuite/libffi.call/closure_fn0.c,
+       testsuite/libffi.call/cls_9byte2.c,
+       testsuite/libffi.call/closure_fn2.c,
+       testsuite/libffi.call/closure_fn4.c,
+       testsuite/libffi.call/cls_ushort.c,
+       testsuite/libffi.call/closure_fn6.c,
+       testsuite/libffi.call/cls_5byte.c,
+       testsuite/libffi.call/cls_align_pointer.c,
+       testsuite/libffi.call/cls_7byte.c,
+       testsuite/libffi.call/cls_align_sint32.c,
+       testsuite/libffi.special/unwindtest_ffi_call.cc,
+       testsuite/libffi.special/unwindtest.cc: Remove xfail for mips64*-*-*.
+
+2007-08-10  David Daney  <ddaney@avtrex.com>
+
+       PR libffi/28313
+       * configure.ac: Don't treat mips64 as a special case.
+       * Makefile.am (nodist_libffi_la_SOURCES): Add n32.S.
+       * configure: Regenerate
+       * Makefile.in: Ditto.
+       * fficonfig.h.in: Ditto.
+       * src/mips/ffitarget.h (REG_L, REG_S, SUBU, ADDU, SRL, LI): Indent.
+       (LA, EH_FRAME_ALIGN, FDE_ADDR_BYTES): New preprocessor macros.
+       (FFI_DEFAULT_ABI): Set for n64 case.
+       (FFI_CLOSURES, FFI_TRAMPOLINE_SIZE): Define for n32 and n64 cases.
+       * src/mips/n32.S (ffi_call_N32): Add debug macros and labels for FDE.
+       (ffi_closure_N32): New function.
+       (.eh_frame): New section
+       * src/mips/o32.S: Clean up comments.
+       (ffi_closure_O32): Pass ffi_closure parameter in $12.
+       * src/mips/ffi.c: Use FFI_MIPS_N32 instead of
+       _MIPS_SIM == _ABIN32 throughout.
+       (FFI_MIPS_STOP_HERE): New, use in place of
+       ffi_stop_here.
+       (ffi_prep_args): Use unsigned long to hold pointer values.  Rewrite
+       to support n32/n64 ABIs.
+       (calc_n32_struct_flags): Rewrite.
+       (calc_n32_return_struct_flags): Remove unused variable.  Reverse
+       position of flag bits.
+       (ffi_prep_cif_machdep): Rewrite n32 portion.
+       (ffi_call): Enable for n64.  Add special handling for small structure
+       return values.
+       (ffi_prep_closure_loc): Add n32 and n64 support.
+       (ffi_closure_mips_inner_O32): Add cast to silence warning.
+       (copy_struct_N32, ffi_closure_mips_inner_N32): New functions.
+
+2007-08-08  David Daney  <ddaney@avtrex.com>
+
+       * testsuite/libffi.call/ffitest.h (ffi_type_mylong): Remove definition.
+       * testsuite/libffi.call/cls_align_uint16.c (main): Use correct type
+       specifiers.
+       * testsuite/libffi.call/nested_struct1.c (main): Ditto.
+       * testsuite/libffi.call/cls_sint.c (main): Ditto.
+       * testsuite/libffi.call/nested_struct9.c (main): Ditto.
+       * testsuite/libffi.call/cls_20byte1.c (main): Ditto.
+       * testsuite/libffi.call/cls_9byte1.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn1.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn3.c (main): Ditto.
+       * testsuite/libffi.call/return_dbl2.c (main): Ditto.
+       * testsuite/libffi.call/cls_sshort.c (main): Ditto.
+       * testsuite/libffi.call/return_fl3.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn5.c (main): Ditto.
+       * testsuite/libffi.call/nested_struct.c (main): Ditto.
+       * testsuite/libffi.call/nested_struct10.c (main): Ditto.
+       * testsuite/libffi.call/return_ll1.c (main): Ditto.
+       * testsuite/libffi.call/cls_8byte.c (main): Ditto.
+       * testsuite/libffi.call/cls_align_uint32.c (main): Ditto.
+       * testsuite/libffi.call/cls_align_sint16.c (main): Ditto.
+       * testsuite/libffi.call/cls_20byte.c (main): Ditto.
+       * testsuite/libffi.call/nested_struct2.c (main): Ditto.
+       * testsuite/libffi.call/cls_24byte.c (main): Ditto.
+       * testsuite/libffi.call/nested_struct6.c (main): Ditto.
+       * testsuite/libffi.call/cls_uint.c (main): Ditto.
+       * testsuite/libffi.call/cls_12byte.c (main): Ditto.
+       * testsuite/libffi.call/cls_16byte.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn0.c (main): Ditto.
+       * testsuite/libffi.call/cls_9byte2.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn2.c (main): Ditto.
+       * testsuite/libffi.call/return_dbl1.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn4.c (main): Ditto.
+       * testsuite/libffi.call/closure_fn6.c (main): Ditto.
+       * testsuite/libffi.call/cls_align_sint32.c (main): Ditto.
+
+2007-08-07  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/sysv.S (ffi_closure_raw_SYSV): Fix typo in previous
+       checkin.
+
+2007-08-06  Andrew Haley  <aph@redhat.com>
+
+       PR testsuite/32843
+       * src/x86/sysv.S (ffi_closure_raw_SYSV): Handle FFI_TYPE_UINT8,
+       FFI_TYPE_SINT8, FFI_TYPE_UINT16, FFI_TYPE_SINT16, FFI_TYPE_UINT32,
+       FFI_TYPE_SINT32.
+
+2007-08-02  David Daney  <ddaney@avtrex.com>
+
+       * testsuite/libffi.call/return_ul.c (main): Define return type as
+       ffi_arg.  Use proper printf conversion specifier.
+
+2007-07-30  Andrew Haley  <aph@redhat.com>
+
+       PR testsuite/32843
+       * src/x86/ffi.c (ffi_prep_cif_machdep): in x86 case, add code for
+       signed/unsigned int8/16.
+       * src/x86/sysv.S (ffi_call_SYSV): Rewrite to:
+       Use a jump table.
+       Remove code to pop args from the stack after call.
+       Special-case signed/unsigned int8/16.
+       * testsuite/libffi.call/return_sc.c (main): Revert.
+
+2007-07-26  Richard Guenther  <rguenther@suse.de>
+
+       PR testsuite/32843
+       * testsuite/libffi.call/return_sc.c (main): Verify call
+       result as signed char, not ffi_arg.
+
+2007-07-16  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * configure.ac (i?86-*-solaris2.1[0-9]): Set TARGET to X86_64.
+       * configure: Regenerate.
+
+2007-07-11  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/ffi.c: Don't include sys/cachectl.h.
+       (ffi_prep_closure_loc): Use __builtin___clear_cache() instead of
+       cacheflush().
+
+2007-05-18  Aurelien Jarno  <aurelien@aurel32.net>
+
+       * src/arm/ffi.c (ffi_prep_closure_loc): Renamed and ajusted
+       from (ffi_prep_closure): ... this.
+       (FFI_INIT_TRAMPOLINE): Adjust.
+
+2005-12-31  Phil Blundell  <pb@reciva.com>
+
+       * src/arm/ffi.c (ffi_prep_incoming_args_SYSV,
+       ffi_closure_SYSV_inner, ffi_prep_closure): New, add closure support.
+       * src/arm/sysv.S(ffi_closure_SYSV): Likewise.
+       * src/arm/ffitarget.h (FFI_TRAMPOLINE_SIZE): Likewise.
+       (FFI_CLOSURES): Enable closure support.
+
+2007-07-03  Andrew Haley  <aph@hedges.billgatliff.com>
+
+       * testsuite/libffi.call/cls_multi_ushort.c,
+       testsuite/libffi.call/cls_align_uint16.c,
+       testsuite/libffi.call/nested_struct1.c,
+       testsuite/libffi.call/nested_struct3.c,
+       testsuite/libffi.call/cls_7_1_byte.c,
+       testsuite/libffi.call/cls_double.c,
+       testsuite/libffi.call/nested_struct5.c,
+       testsuite/libffi.call/nested_struct7.c,
+       testsuite/libffi.call/cls_sint.c,
+       testsuite/libffi.call/nested_struct9.c,
+       testsuite/libffi.call/cls_20byte1.c,
+       testsuite/libffi.call/cls_multi_sshortchar.c,
+       testsuite/libffi.call/cls_align_sint64.c,
+       testsuite/libffi.call/cls_3byte2.c,
+       testsuite/libffi.call/cls_multi_schar.c,
+       testsuite/libffi.call/cls_multi_uchar.c,
+       testsuite/libffi.call/cls_19byte.c,
+       testsuite/libffi.call/cls_9byte1.c,
+       testsuite/libffi.call/cls_align_float.c,
+       testsuite/libffi.call/closure_fn1.c,
+       testsuite/libffi.call/problem1.c,
+       testsuite/libffi.call/closure_fn3.c,
+       testsuite/libffi.call/cls_sshort.c,
+       testsuite/libffi.call/closure_fn5.c,
+       testsuite/libffi.call/cls_align_double.c,
+       testsuite/libffi.call/cls_2byte.c,
+       testsuite/libffi.call/nested_struct.c,
+       testsuite/libffi.call/nested_struct10.c,
+       testsuite/libffi.call/cls_4byte.c,
+       testsuite/libffi.call/cls_6byte.c,
+       testsuite/libffi.call/cls_8byte.c,
+       testsuite/libffi.call/cls_multi_sshort.c,
+       testsuite/libffi.call/cls_align_uint32.c,
+       testsuite/libffi.call/cls_align_sint16.c,
+       testsuite/libffi.call/cls_float.c,
+       testsuite/libffi.call/cls_20byte.c,
+       testsuite/libffi.call/cls_5_1_byte.c,
+       testsuite/libffi.call/nested_struct2.c,
+       testsuite/libffi.call/cls_24byte.c,
+       testsuite/libffi.call/nested_struct4.c,
+       testsuite/libffi.call/nested_struct6.c,
+       testsuite/libffi.call/cls_64byte.c,
+       testsuite/libffi.call/nested_struct8.c,
+       testsuite/libffi.call/cls_uint.c,
+       testsuite/libffi.call/cls_multi_ushortchar.c,
+       testsuite/libffi.call/cls_schar.c,
+       testsuite/libffi.call/cls_uchar.c,
+       testsuite/libffi.call/cls_align_uint64.c,
+       testsuite/libffi.call/cls_ulonglong.c,
+       testsuite/libffi.call/cls_align_longdouble.c,
+       testsuite/libffi.call/cls_1_1byte.c,
+       testsuite/libffi.call/cls_12byte.c,
+       testsuite/libffi.call/cls_3_1byte.c,
+       testsuite/libffi.call/cls_3byte1.c,
+       testsuite/libffi.call/cls_4_1byte.c,
+       testsuite/libffi.call/cls_6_1_byte.c,
+       testsuite/libffi.call/cls_16byte.c,
+       testsuite/libffi.call/cls_18byte.c,
+       testsuite/libffi.call/closure_fn0.c,
+       testsuite/libffi.call/cls_9byte2.c,
+       testsuite/libffi.call/closure_fn2.c,
+       testsuite/libffi.call/closure_fn4.c,
+       testsuite/libffi.call/cls_ushort.c,
+       testsuite/libffi.call/closure_fn6.c,
+       testsuite/libffi.call/cls_5byte.c,
+       testsuite/libffi.call/cls_align_pointer.c,
+       testsuite/libffi.call/cls_7byte.c,
+       testsuite/libffi.call/cls_align_sint32.c,
+       testsuite/libffi.special/unwindtest_ffi_call.cc,
+       testsuite/libffi.special/unwindtest.cc: Enable for ARM.
+
+2007-07-05  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * aclocal.m4: Regenerated.
+
+2007-06-02  Paolo Bonzini  <bonzini@gnu.org>
+
+       * configure: Regenerate.
+
+2007-05-23  Steve Ellcey  <sje@cup.hp.com>
+
+       * Makefile.in: Regenerate.
+       * configure: Regenerate.
+       * aclocal.m4: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2007-05-10  Roman Zippel <zippel@linux-m68k.org>
+
+       * src/m68k/ffi.c (ffi_prep_incoming_args_SYSV,
+       ffi_closure_SYSV_inner,ffi_prep_closure): New, add closure support.
+       * src/m68k/sysv.S(ffi_closure_SYSV,ffi_closure_struct_SYSV): Likewise.
+       * src/m68k/ffitarget.h (FFI_TRAMPOLINE_SIZE): Likewise.
+       (FFI_CLOSURES): Enable closure support.
+
+2007-05-10  Roman Zippel <zippel@linux-m68k.org>
+
+       * configure.ac (HAVE_AS_CFI_PSEUDO_OP): New test.
+       * configure: Regenerate.
+       * fficonfig.h.in: Regenerate.
+       * src/m68k/sysv.S (CFI_STARTPROC,CFI_ENDPROC,
+       CFI_OFFSET,CFI_DEF_CFA): New macros.
+       (ffi_call_SYSV): Add callframe annotation.
+
+2007-05-10  Roman Zippel <zippel@linux-m68k.org>
+
+       * src/m68k/ffi.c (ffi_prep_args,ffi_prep_cif_machdep): Fix
+       numerous test suite failures.
+       * src/m68k/sysv.S (ffi_call_SYSV): Likewise.
+
+2007-04-11  Paolo Bonzini  <bonzini@gnu.org>
+
+       * Makefile.am (EXTRA_DIST): Bring up to date.
+       * Makefile.in: Regenerate.
+       * src/frv/eabi.S: Remove RCS keyword.
+
+2007-04-06  Richard Henderson  <rth@redhat.com>
+
+       * configure.ac: Tidy target case.
+       (HAVE_LONG_DOUBLE): Allow the target to override.
+       * configure: Regenerate.
+       * include/ffi.h.in: Don't define ffi_type_foo if
+       LIBFFI_HIDE_BASIC_TYPES is defined.
+       (ffi_type_longdouble): If not HAVE_LONG_DOUBLE, define
+       to ffi_type_double.
+       * types.c (LIBFFI_HIDE_BASIC_TYPES): Define.
+       (FFI_TYPEDEF, ffi_type_void): Mark the data const.
+       (ffi_type_longdouble): Special case for Alpha.  Don't define
+       if long double == double.
+
+       * src/alpha/ffi.c (FFI_TYPE_LONGDOUBLE): Assert unique value.
+       (ffi_prep_cif_machdep): Handle it as the 128-bit type.
+       (ffi_call, ffi_closure_osf_inner): Likewise.
+       (ffi_closure_osf_inner): Likewise.  Mark hidden.
+       (ffi_call_osf, ffi_closure_osf): Mark hidden.
+       * src/alpha/ffitarget.h (FFI_LAST_ABI): Tidy definition.
+       * src/alpha/osf.S (ffi_call_osf, ffi_closure_osf): Mark hidden.
+       (load_table): Handle 128-bit long double.
+
+       * testsuite/libffi.call/float4.c: Add -mieee for alpha.
+
+2007-04-06  Tom Tromey  <tromey@redhat.com>
+
+       PR libffi/31491:
+       * README: Fixed bug in example.
+
+2007-04-03  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/closures.c: Include sys/statfs.h.
+       (_GNU_SOURCE): Define on Linux.
+       (FFI_MMAP_EXEC_SELINUX): Define.
+       (selinux_enabled): New variable.
+       (selinux_enabled_check): New function.
+       (is_selinux_enabled): Define.
+       (dlmmap): Use it.
+
+2007-03-24  Uros Bizjak  <ubizjak@gmail.com>
+
+       * testsuite/libffi.call/return_fl2.c (return_fl): Mark as static.
+       Use 'volatile float sum' to create sum of floats to avoid false
+       negative due to excess precision on ix86 targets.
+       (main): Ditto.
+
+2007-03-08  Alexandre Oliva  <aoliva@redhat.com>
+
+       * src/powerpc/ffi.c (flush_icache): Fix left-over from previous
+       patch.
+       (ffi_prep_closure_loc): Remove unneeded casts.  Add needed ones.
+
+2007-03-07  Alexandre Oliva  <aoliva@redhat.com>
+
+       * include/ffi.h.in (ffi_closure_alloc, ffi_closure_free): New.
+       (ffi_prep_closure_loc): New.
+       (ffi_prep_raw_closure_loc): New.
+       (ffi_prep_java_raw_closure_loc): New.
+       * src/closures.c: New file.
+       * src/dlmalloc.c [FFI_MMAP_EXEC_WRIT] (struct malloc_segment):
+       Replace sflags with exec_offset.
+       [FFI_MMAP_EXEC_WRIT] (mmap_exec_offset, add_segment_exec_offset,
+       sub_segment_exec_offset): New macros.
+       (get_segment_flags, set_segment_flags, check_segment_merge): New
+       macros.
+       (is_mmapped_segment, is_extern_segment): Use get_segment_flags.
+       (add_segment, sys_alloc, create_mspace, create_mspace_with_base,
+       destroy_mspace): Use new macros.
+       (sys_alloc): Silence warning.
+       * Makefile.am (libffi_la_SOURCES): Add src/closures.c.
+       * Makefile.in: Rebuilt.
+       * src/prep_cif [FFI_CLOSURES] (ffi_prep_closure): Implement in
+       terms of ffi_prep_closure_loc.
+       * src/raw_api.c (ffi_prep_raw_closure_loc): Renamed and adjusted
+       from...
+       (ffi_prep_raw_closure): ... this.  Re-implement in terms of the
+       renamed version.
+       * src/java_raw_api (ffi_prep_java_raw_closure_loc): Renamed and
+       adjusted from...
+       (ffi_prep_java_raw_closure): ... this.  Re-implement in terms of
+       the renamed version.
+       * src/alpha/ffi.c (ffi_prep_closure_loc): Renamed from
+       (ffi_prep_closure): ... this.
+       * src/pa/ffi.c: Likewise.
+       * src/cris/ffi.c: Likewise.  Adjust.
+       * src/frv/ffi.c: Likewise.
+       * src/ia64/ffi.c: Likewise.
+       * src/mips/ffi.c: Likewise.
+       * src/powerpc/ffi_darwin.c: Likewise.
+       * src/s390/ffi.c: Likewise.
+       * src/sh/ffi.c: Likewise.
+       * src/sh64/ffi.c: Likewise.
+       * src/sparc/ffi.c: Likewise.
+       * src/x86/ffi64.c: Likewise.
+       * src/x86/ffi.c: Likewise.
+       (FFI_INIT_TRAMPOLINE): Adjust.
+       (ffi_prep_raw_closure_loc): Renamed and adjusted from...
+       (ffi_prep_raw_closure): ... this.
+       * src/powerpc/ffi.c (ffi_prep_closure_loc): Renamed from
+       (ffi_prep_closure): ... this.
+       (flush_icache): Adjust.
+
+2007-03-07  Alexandre Oliva  <aoliva@redhat.com>
+
+       * src/dlmalloc.c: New file, imported version 2.8.3 of Doug
+       Lea's malloc.
+
+2007-03-01  Brooks Moses  <brooks.moses@codesourcery.com>
+
+       * Makefile.am: Add dummy install-pdf target.
+       * Makefile.in: Regenerate
+
+2007-02-13  Andreas Krebbel  <krebbel1@de.ibm.com>
+
+       * src/s390/ffi.c (ffi_prep_args, ffi_prep_cif_machdep,
+       ffi_closure_helper_SYSV): Add long double handling.
+
+2007-02-02  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/powerpc/linux64.S (ffi_call_LINUX64): Move restore of r2
+       immediately after bctrl instruction.
+
+2007-01-18  Alexandre Oliva  <aoliva@redhat.com>
+
+       * Makefile.am (all-recursive, install-recursive,
+       mostlyclean-recursive, clean-recursive, distclean-recursive,
+       maintainer-clean-recursive): Add missing targets.
+       * Makefile.in: Rebuilt.
+
+2006-12-14  Andreas Tobler  <a.tobler@schweiz.org>
+
+       * configure.ac: Add TARGET for x86_64-*-darwin*.
+       * Makefile.am (nodist_libffi_la_SOURCES): Add rules for 64-bit sources
+       for X86_DARWIN.
+       * src/x86/ffitarget.h: Set trampoline size for x86_64-*-darwin*.
+       * src/x86/darwin64.S: New file for x86_64-*-darwin* support.
+       * configure: Regenerate.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+       * testsuite/libffi.special/unwindtest_ffi_call.cc: New test case for
+       ffi_call only.
+
+2006-12-13  Andreas Tobler <a.tobler@schweiz.org>
+
+       * aclocal.m4: Regenerate with aclocal -I .. as written in the
+       Makefile.am.
+
+2006-10-31  Geoffrey Keating  <geoffk@apple.com>
+
+       * src/powerpc/ffi_darwin.c (darwin_adjust_aggregate_sizes): New.
+       (ffi_prep_cif_machdep): Call darwin_adjust_aggregate_sizes for
+       Darwin.
+       * testsuite/libffi.call/nested_struct4.c: Remove Darwin XFAIL.
+       * testsuite/libffi.call/nested_struct6.c: Remove Darwin XFAIL.
+
+2006-10-10  Paolo Bonzini  <bonzini@gnu.org>
+           Sandro Tolaini  <tolaini@libero.it>
+
+       * configure.ac [i*86-*-darwin*]: Set X86_DARWIN symbol and
+       conditional.
+       * configure: Regenerated.
+       * Makefile.am (nodist_libffi_la_SOURCES) [X86_DARWIN]: New case.
+       (EXTRA_DIST): Add src/x86/darwin.S.
+       * Makefile.in: Regenerated.
+       * include/Makefile.in: Regenerated.
+       * testsuite/Makefile.in: Regenerated.
+
+       * src/x86/ffi.c (ffi_prep_cif_machdep) [X86_DARWIN]: Treat like
+       X86_WIN32, and additionally align stack to 16 bytes.
+       * src/x86/darwin.S: New, based on sysv.S.
+       * src/prep_cif.c (ffi_prep_cif) [X86_DARWIN]: Align > 8-byte structs.
+
+2006-09-12  David Daney  <ddaney@avtrex.com>
+
+       PR libffi/23935
+       * include/Makefile.am: Install both ffi.h and ffitarget.h in
+       $(libdir)/gcc/$(target_alias)/$(gcc_version)/include.
+       * aclocal.m4: Regenerated for automake 1.9.6.
+       * Makefile.in: Regenerated.
+       * include/Makefile.in: Regenerated.
+       * testsuite/Makefile.in: Regenerated.
+
+2006-08-17  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * include/ffi_common.h (struct): Revert accidental commit.
+
+2006-08-15  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * include/ffi_common.h: Remove lint directives.
+       * include/ffi.h.in: Likewise.
+
+2006-07-25  Torsten Schoenfeld  <kaffeetisch@gmx.de>
+
+       * include/ffi.h.in (ffi_type_ulong, ffi_type_slong): Define correctly
+       for 32-bit architectures.
+       * testsuite/libffi.call/return_ul.c: New test case.
+
+2006-07-19  David Daney  <ddaney@avtrex.com>
+
+       * testsuite/libffi.call/closure_fn6.c: Remove xfail for mips,
+       xfail remains for mips64.
+
+2006-05-23  Carlos O'Donell  <carlos@codesourcery.com>
+
+       * Makefile.am: Add install-html target. Add install-html to .PHONY
+       * Makefile.in: Regenerate.
+       * aclocal.m4: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2006-05-18  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
+
+       * pa/ffi.c (ffi_prep_args_pa32): Load floating point arguments from
+       stack slot.
+
+2006-04-22  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * README: Remove notice about 'Crazy Comments'.
+       * src/debug.c: Remove lint directives. Cleanup white spaces.
+       * src/java_raw_api.c: Likewise.
+       * src/prep_cif.c: Likewise.
+       * src/raw_api.c: Likewise.
+       * src/ffitest.c: Delete. No longer needed, all test cases migrated
+       to the testsuite.
+       * src/arm/ffi.c: Remove lint directives.
+       * src/m32r/ffi.c: Likewise.
+       * src/pa/ffi.c: Likewise.
+       * src/powerpc/ffi.c: Likewise.
+       * src/powerpc/ffi_darwin.c: Likewise.
+       * src/sh/ffi.c: Likewise.
+       * src/sh64/ffi.c: Likewise.
+       * src/x86/ffi.c: Likewise.
+       * testsuite/libffi.call/float2.c: Likewise.
+       * testsuite/libffi.call/promotion.c: Likewise.
+       * testsuite/libffi.call/struct1.c: Likewise.
+
+2006-04-13  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/pa/hpux32.S: Correct unwind offset calculation for
+       ffi_closure_pa32.
+       * src/pa/linux.S: Likewise.
+
+2006-04-12  James E Wilson  <wilson@specifix.com>
+
+       PR libgcj/26483
+       * src/ia64/ffi.c (stf_spill, ldf_fill): Rewrite as macros.
+       (hfa_type_load): Call stf_spill.
+       (hfa_type_store): Call ldf_fill.
+       (ffi_call): Adjust calls to above routines.  Add local temps for
+       macro result.
+
+2006-04-10  Matthias Klose  <doko@debian.org>
+
+       * testsuite/lib/libffi-dg.exp (libffi-init): Recognize multilib
+       directory names containing underscores.
+
+2006-04-07  James E Wilson  <wilson@specifix.com>
+
+       * testsuite/libffi.call/float4.c: New testcase.
+
+2006-04-05  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
+           Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * Makefile.am: Add PA_HPUX port.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+       * configure.ac: Add PA_HPUX rules.
+       * configure: Regenerate.
+       * src/pa/ffitarget.h: Rename linux target to PA_LINUX.
+       Add PA_HPUX and PA64_HPUX.
+       Rename FFI_LINUX ABI to FFI_PA32 ABI.
+       (FFI_TRAMPOLINE_SIZE): Define for 32-bit HP-UX targets.
+       (FFI_TYPE_SMALL_STRUCT2): Define.
+       (FFI_TYPE_SMALL_STRUCT4): Likewise.
+       (FFI_TYPE_SMALL_STRUCT8): Likewise.
+       (FFI_TYPE_SMALL_STRUCT3): Redefine.
+       (FFI_TYPE_SMALL_STRUCT5): Likewise.
+       (FFI_TYPE_SMALL_STRUCT6): Likewise.
+       (FFI_TYPE_SMALL_STRUCT7): Likewise.
+       * src/pa/ffi.c (ROUND_DOWN): Delete.
+       (fldw, fstw, fldd, fstd): Use '__asm__'.
+       (ffi_struct_type): Add support for FFI_TYPE_SMALL_STRUCT2,
+       FFI_TYPE_SMALL_STRUCT4 and FFI_TYPE_SMALL_STRUCT8.
+       (ffi_prep_args_LINUX): Rename to ffi_prep_args_pa32. Update comment.
+       Simplify incrementing of stack slot variable. Change type of local
+       'n' to unsigned int.
+       (ffi_size_stack_LINUX): Rename to ffi_size_stack_pa32. Handle long
+       double on PA_HPUX.
+       (ffi_prep_cif_machdep): Likewise.
+       (ffi_call): Likewise.
+       (ffi_closure_inner_LINUX): Rename to ffi_closure_inner_pa32. Change
+       return type to ffi_status. Simplify incrementing of stack slot
+       variable. Only copy floating point argument registers when PA_LINUX
+       is true. Reformat debug statement.
+       Add support for FFI_TYPE_SMALL_STRUCT2, FFI_TYPE_SMALL_STRUCT4 and
+       FFI_TYPE_SMALL_STRUCT8.
+       (ffi_closure_LINUX): Rename to ffi_closure_pa32. Add 'extern' to
+       declaration.
+       (ffi_prep_closure): Make linux trampoline conditional on PA_LINUX.
+       Add nops to cache flush.  Add trampoline for PA_HPUX.
+       * src/pa/hpux32.S: New file.
+       * src/pa/linux.S (ffi_call_LINUX): Rename to ffi_call_pa32. Rename
+       ffi_prep_args_LINUX to ffi_prep_args_pa32.
+       Localize labels. Add support for 2, 4 and 8-byte small structs. Handle
+       unaligned destinations in 3, 5, 6 and 7-byte small structs. Order
+       argument type checks so that common argument types appear first.
+       (ffi_closure_LINUX): Rename to ffi_closure_pa32. Rename
+       ffi_closure_inner_LINUX to ffi_closure_inner_pa32.
+
+2006-03-24  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/powerpc/ffitarget.h (enum ffi_abi): Add FFI_LINUX.  Default
+       for 32-bit using IBM extended double format.  Fix FFI_LAST_ABI.
+       * src/powerpc/ffi.c (ffi_prep_args_SYSV): Handle linux variant of
+       FFI_TYPE_LONGDOUBLE.
+       (ffi_prep_args64): Assert using IBM extended double.
+       (ffi_prep_cif_machdep): Don't munge FFI_TYPE_LONGDOUBLE type.
+       Handle FFI_LINUX FFI_TYPE_LONGDOUBLE return and args.
+       (ffi_call): Handle FFI_LINUX.
+       (ffi_closure_helper_SYSV): Non FFI_LINUX long double return needs
+       gpr3 return pointer as for struct return.  Handle FFI_LINUX
+       FFI_TYPE_LONGDOUBLE return and args.  Don't increment "nf"
+       unnecessarily.
+       * src/powerpc/ppc_closure.S (ffi_closure_SYSV): Load both f1 and f2
+       for FFI_TYPE_LONGDOUBLE.  Move epilogue insns into case table.
+       Don't use r6 as pointer to results, instead use sp offset.  Don't
+       make a special call to load lr with case table address, instead
+       use offset from previous call.
+       * src/powerpc/sysv.S (ffi_call_SYSV): Save long double return.
+       * src/powerpc/linux64.S (ffi_call_LINUX64): Simplify long double
+       return.
+
+2006-03-15  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh64/ffi.c (ffi_prep_cif_machdep): Handle float arguments
+       passed with FP registers correctly.
+       (ffi_closure_helper_SYSV): Likewise.
+       * src/sh64/sysv.S: Likewise.
+
+2006-03-01  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.special/unwindtest.cc (closure_test_fn): Mark cif,
+       args and userdata unused.
+       (closure_test_fn1): Mark cif and userdata unused.
+       (main): Remove unused res.
+
+2006-02-28  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/call.exp: Adjust FSF address. Add test runs for
+       -O2, -O3, -Os and the warning flags -W -Wall.
+       * testsuite/libffi.special/special.exp: Likewise.
+       * testsuite/libffi.call/ffitest.h: Add an __UNUSED__ macro to mark
+       unused parameter unused for gcc or else do nothing.
+       * testsuite/libffi.special/ffitestcxx.h: Likewise.
+       * testsuite/libffi.call/cls_12byte.c (cls_struct_12byte_gn): Mark cif
+       and userdata unused.
+       * testsuite/libffi.call/cls_16byte.c (cls_struct_16byte_gn): Likewise.
+       * testsuite/libffi.call/cls_18byte.c (cls_struct_18byte_gn): Likewise.
+       * testsuite/libffi.call/cls_19byte.c (cls_struct_19byte_gn): Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c (cls_struct_1_1byte_gn): Likewise.
+       * testsuite/libffi.call/cls_20byte.c (cls_struct_20byte_gn): Likewise.
+       * testsuite/libffi.call/cls_20byte1.c (cls_struct_20byte_gn): Likewise.
+       * testsuite/libffi.call/cls_24byte.c (cls_struct_24byte_gn): Likewise.
+       * testsuite/libffi.call/cls_2byte.c (cls_struct_2byte_gn): Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c (cls_struct_3_1byte_gn): Likewise.
+       * testsuite/libffi.call/cls_3byte1.c (cls_struct_3byte_gn): Likewise.
+       * testsuite/libffi.call/cls_3byte2.c (cls_struct_3byte_gn1): Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c (cls_struct_4_1byte_gn): Likewise.
+       * testsuite/libffi.call/cls_4byte.c (cls_struct_4byte_gn): Likewise.
+       * testsuite/libffi.call/cls_5_1_byte.c (cls_struct_5byte_gn): Likewise.
+       * testsuite/libffi.call/cls_5byte.c (cls_struct_5byte_gn): Likewise.
+       * testsuite/libffi.call/cls_64byte.c (cls_struct_64byte_gn): Likewise.
+       * testsuite/libffi.call/cls_6_1_byte.c (cls_struct_6byte_gn): Likewise.
+       * testsuite/libffi.call/cls_6byte.c (cls_struct_6byte_gn): Likewise.
+       * testsuite/libffi.call/cls_7_1_byte.c (cls_struct_7byte_gn): Likewise.
+       * testsuite/libffi.call/cls_7byte.c (cls_struct_7byte_gn): Likewise.
+       * testsuite/libffi.call/cls_8byte.c (cls_struct_8byte_gn): Likewise.
+       * testsuite/libffi.call/cls_9byte1.c (cls_struct_9byte_gn): Likewise.
+       * testsuite/libffi.call/cls_9byte2.c (cls_struct_9byte_gn): Likewise.
+       * testsuite/libffi.call/cls_align_double.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_float.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_longdouble.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_pointer.c (cls_struct_align_fn): Cast
+       void* to avoid compiler warning.
+       (main): Likewise.
+       (cls_struct_align_gn): Mark cif and userdata unused.
+       * testsuite/libffi.call/cls_align_sint16.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_sint32.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_sint64.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_uint16.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_align_uint32.c (cls_struct_align_gn):
+       Likewise.
+       * testsuite/libffi.call/cls_double.c (cls_ret_double_fn): Likewise.
+       * testsuite/libffi.call/cls_float.c (cls_ret_float_fn): Likewise.
+       * testsuite/libffi.call/cls_multi_schar.c (test_func_gn): Mark cif and
+       data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_multi_sshort.c (test_func_gn): Mark cif and
+       data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_multi_sshortchar.c (test_func_gn): Mark cif
+       and data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_multi_uchar.c (test_func_gn): Mark cif and
+       data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_multi_ushort.c (test_func_gn): Mark cif and
+       data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_multi_ushortchar.c (test_func_gn): Mark cif
+       and data unused.
+       (main): Cast res_call to silence gcc.
+       * testsuite/libffi.call/cls_schar.c (cls_ret_schar_fn): Mark cif and
+       userdata unused.
+       (cls_ret_schar_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/cls_sint.c (cls_ret_sint_fn): Mark cif and
+       userdata unused.
+       (cls_ret_sint_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/cls_sshort.c (cls_ret_sshort_fn): Mark cif and
+       userdata unused.
+       (cls_ret_sshort_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/cls_uchar.c (cls_ret_uchar_fn):  Mark cif and
+       userdata unused.
+       (cls_ret_uchar_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/cls_uint.c (cls_ret_uint_fn): Mark cif and
+       userdata unused.
+       (cls_ret_uint_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/cls_ulonglong.c (cls_ret_ulonglong_fn): Mark cif
+       and userdata unused.
+       * testsuite/libffi.call/cls_ushort.c (cls_ret_ushort_fn): Mark cif and
+       userdata unused.
+       (cls_ret_ushort_fn): Cast printf parameter to silence gcc.
+       * testsuite/libffi.call/float.c (floating): Remove unused parameter e.
+       * testsuite/libffi.call/float1.c (main): Remove unused variable i.
+       Cleanup white spaces.
+       * testsuite/libffi.call/negint.c (checking): Remove unused variable i.
+       * testsuite/libffi.call/nested_struct.c (cls_struct_combined_gn): Mark
+       cif and userdata unused.
+       * testsuite/libffi.call/nested_struct1.c (cls_struct_combined_gn):
+       Likewise.
+       * testsuite/libffi.call/nested_struct10.c (B_gn): Likewise.
+       * testsuite/libffi.call/nested_struct2.c (B_fn): Adjust printf
+       formatters to silence gcc.
+       (B_gn): Mark cif and userdata unused.
+       * testsuite/libffi.call/nested_struct3.c (B_gn): Mark cif and userdata
+       unused.
+       * testsuite/libffi.call/nested_struct4.c: Mention related PR.
+       (B_gn): Mark cif and userdata unused.
+       * testsuite/libffi.call/nested_struct5.c (B_gn): Mark cif and userdata
+       unused.
+       * testsuite/libffi.call/nested_struct6.c: Mention related PR.
+       (B_gn): Mark cif and userdata unused.
+       * testsuite/libffi.call/nested_struct7.c (B_gn): Mark cif and userdata
+       unused.
+       * testsuite/libffi.call/nested_struct8.c (B_gn): Likewise.
+       * testsuite/libffi.call/nested_struct9.c (B_gn): Likewise.
+       * testsuite/libffi.call/problem1.c (stub): Likewise.
+       * testsuite/libffi.call/pyobjc-tc.c (main): Cast the result to silence
+       gcc.
+       * testsuite/libffi.call/return_fl2.c (return_fl): Add the note mentioned
+       in the last commit for this test case in the test case itself.
+       * testsuite/libffi.call/closure_fn0.c (closure_test_fn0): Mark cif as
+       unused.
+       * testsuite/libffi.call/closure_fn1.c (closure_test_fn1): Likewise.
+       * testsuite/libffi.call/closure_fn2.c (closure_test_fn2): Likewise.
+       * testsuite/libffi.call/closure_fn3.c (closure_test_fn3): Likewise.
+       * testsuite/libffi.call/closure_fn4.c (closure_test_fn0): Likewise.
+       * testsuite/libffi.call/closure_fn5.c (closure_test_fn5): Likewise.
+       * testsuite/libffi.call/closure_fn6.c (closure_test_fn0): Likewise.
+
+2006-02-22  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/sysv.S: Fix register numbers in the FDE for
+       ffi_closure_SYSV.
+
+2006-02-20  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/return_fl2.c (return_fl): Remove static
+       declaration to avoid a false negative on ix86. See PR323.
+
+2006-02-18  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_closure_helper_SYSV): Remove unused variable
+       and cast integer to void * if needed.  Update the pointer to
+       the FP register saved area correctly.
+
+2006-02-17  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/nested_struct6.c: XFAIL this test until PR25630
+       is fixed.
+       * testsuite/libffi.call/nested_struct4.c: Likewise.
+
+2006-02-16  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/return_dbl.c: New test case.
+       * testsuite/libffi.call/return_dbl1.c: Likewise.
+       * testsuite/libffi.call/return_dbl2.c: Likewise.
+       * testsuite/libffi.call/return_fl.c: Likewise.
+       * testsuite/libffi.call/return_fl1.c: Likewise.
+       * testsuite/libffi.call/return_fl2.c: Likewise.
+       * testsuite/libffi.call/return_fl3.c: Likewise.
+       * testsuite/libffi.call/closure_fn6.c: Likewise.
+
+       * testsuite/libffi.call/nested_struct2.c: Remove ffi_type_mylong
+       definition.
+       * testsuite/libffi.call/ffitest.h: Add ffi_type_mylong definition
+       here to be used by other test cases too.
+
+       * testsuite/libffi.call/nested_struct10.c: New test case.
+       * testsuite/libffi.call/nested_struct9.c: Likewise.
+       * testsuite/libffi.call/nested_struct8.c: Likewise.
+       * testsuite/libffi.call/nested_struct7.c: Likewise.
+       * testsuite/libffi.call/nested_struct6.c: Likewise.
+       * testsuite/libffi.call/nested_struct5.c: Likewise.
+       * testsuite/libffi.call/nested_struct4.c: Likewise.
+
+2006-01-21  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * configure.ac: Enable libffi for sparc64-*-freebsd*.
+       * configure: Rebuilt.
+
+2006-01-18  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/powerpc/sysv.S (smst_two_register): Don't call __ashldi3,
+       instead do the shifting inline.
+       * src/powerpc/ppc_closure.S (ffi_closure_SYSV): Don't compute %r5
+       shift count unconditionally.  Simplify load sequences for 1, 2, 3, 4
+       and 8 byte structs, for the remaining struct sizes don't call
+       __lshrdi3, instead do the shifting inline.
+
+2005-12-07  Thiemo Seufer  <ths@networkno.de>
+
+       * src/mips/ffitarget.h: Remove obsolete sgidefs.h include. Add
+       missing parentheses.
+       * src/mips/o32.S (ffi_call_O32): Code formatting. Define
+       and use A3_OFF, FP_OFF, RA_OFF. Micro-optimizations.
+       (ffi_closure_O32): Likewise, but with newly defined A3_OFF2,
+       A2_OFF2, A1_OFF2, A0_OFF2, RA_OFF2, FP_OFF2, S0_OFF2, GP_OFF2,
+       V1_OFF2, V0_OFF2, FA_1_1_OFF2, FA_1_0_OFF2, FA_0_1_OFF2,
+       FA_0_0_OFF2.
+       * src/mips/ffi.c (ffi_prep_args): Code formatting. Fix
+       endianness bugs.
+       (ffi_prep_closure): Improve trampoline instruction scheduling.
+       (ffi_closure_mips_inner_O32): Fix endianness bugs.
+
+2005-12-03  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/powerpc/ffi.c: Formatting.
+       (ffi_prep_args_SYSV): Avoid possible aliasing problems by using unions.
+       (ffi_prep_args64): Likewise.
+
+2005-09-30  Geoffrey Keating  <geoffk@apple.com>
+
+       * testsuite/lib/libffi-dg.exp (libffi_target_compile): For
+       darwin, use -shared-libgcc not -lgcc_s, and explain why.
+
+2005-09-26  Tom Tromey  <tromey@redhat.com>
+
+       * testsuite/libffi.call/float1.c (value_type): New typedef.
+       (CANARY): New define.
+       (main): Check for result buffer overflow.
+       * src/powerpc/linux64.S: Handle linux64 long double returns.
+       * src/powerpc/ffi.c (FLAG_RETURNS_128BITS): New constant.
+       (ffi_prep_cif_machdep): Handle linux64 long double returns.
+
+2005-08-25  Alan Modra  <amodra@bigpond.net.au>
+
+       PR target/23404
+       * src/powerpc/ffi.c (ffi_prep_args_SYSV): Correct placement of stack
+       homed fp args.
+       (ffi_status ffi_prep_cif_machdep): Correct stack sizing for same.
+
+2005-08-11  Jakub Jelinek  <jakub@redhat.com>
+
+       * configure.ac (HAVE_HIDDEN_VISIBILITY_ATTRIBUTE): New test.
+       (AH_BOTTOM): Add FFI_HIDDEN definition.
+       * configure: Rebuilt.
+       * fficonfig.h.in: Rebuilt.
+       * src/powerpc/ffi.c (hidden): Remove.
+       (ffi_closure_LINUX64, ffi_prep_args64, ffi_call_LINUX64,
+       ffi_closure_helper_LINUX64): Use FFI_HIDDEN instead of hidden.
+       * src/powerpc/linux64_closure.S (ffi_closure_LINUX64,
+       .ffi_closure_LINUX64): Use FFI_HIDDEN instead of .hidden.
+       * src/x86/ffi.c (ffi_closure_SYSV, ffi_closure_raw_SYSV): Remove,
+       add FFI_HIDDEN to its prototype.
+       (ffi_closure_SYSV_inner): New.
+       * src/x86/sysv.S (ffi_closure_SYSV, ffi_closure_raw_SYSV): New.
+       * src/x86/win32.S (ffi_closure_SYSV, ffi_closure_raw_SYSV): New.
+
+2005-08-10  Alfred M. Szmidt  <ams@gnu.org>
+
+       PR libffi/21819:
+       * configure: Rebuilt.
+       * configure.ac: Handle i*86-*-gnu*.
+
+2005-08-09  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/powerpc/ppc_closure.S (ffi_closure_SYSV): Use
+       DW_CFA_offset_extended_sf rather than
+       DW_CFA_GNU_negative_offset_extended.
+       * src/powerpc/sysv.S (ffi_call_SYSV): Likewise.
+
+2005-07-22  SUGIOKA Toshinobu  <sugioka@itonet.co.jp>
+
+       * src/sh/sysv.S (ffi_call_SYSV): Stop argument popping correctly
+       on sh3.
+       (ffi_closure_SYSV): Change the stack layout for sh3 struct argument.
+       * src/sh/ffi.c (ffi_prep_args): Fix sh3 argument copy, when it is
+       partially on register.
+       (ffi_closure_helper_SYSV): Likewise.
+       (ffi_prep_cif_machdep): Don't set too many cif->flags.
+
+2005-07-20  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_call): Handle small structures correctly.
+       Remove empty line.
+       * src/sh64/ffi.c (simple_type): Remove.
+       (return_type): Handle small structures correctly.
+       (ffi_prep_args): Likewise.
+       (ffi_call): Likewise.
+       (ffi_closure_helper_SYSV): Likewise.
+       * src/sh64/sysv.S (ffi_call_SYSV): Handle 1, 2 and 4-byte return.
+       Emit position independent code if PIC and remove wrong datalabel
+       prefixes from EH data.
+
+2005-07-19  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * Makefile.am (nodist_libffi_la_SOURCES): Add POWERPC_FREEBSD.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+       * configure.ac: Add POWERPC_FREEBSD rules.
+       * configure: Regenerate.
+       * src/powerpc/ffitarget.h: Add POWERPC_FREEBSD rules.
+       (FFI_SYSV_TYPE_SMALL_STRUCT): Define.
+       * src/powerpc/ffi.c: Add flags to handle small structure returns
+       in ffi_call_SYSV.
+       (ffi_prep_cif_machdep): Handle small structures for SYSV 4 ABI.
+       Aka FFI_SYSV.
+       (ffi_closure_helper_SYSV): Likewise.
+       * src/powerpc/ppc_closure.S: Add return types for small structures.
+       * src/powerpc/sysv.S: Add bits to handle small structures for
+       final SYSV 4 ABI.
+
+2005-07-10  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/cls_5_1_byte.c: New test file.
+       * testsuite/libffi.call/cls_6_1_byte.c: Likewise.
+       * testsuite/libffi.call/cls_7_1_byte.c: Likewise.
+
+2005-07-05  Randolph Chung  <tausq@debian.org>
+
+       * src/pa/ffi.c (ffi_struct_type): Rename FFI_TYPE_SMALL_STRUCT1
+       as FFI_TYPE_SMALL_STRUCT3.  Break out handling for 5-7 byte
+       structures.  Kill compilation warnings.
+       (ffi_closure_inner_LINUX): Print return values as hex in debug
+       message.  Rename FFI_TYPE_SMALL_STRUCT1 as FFI_TYPE_SMALL_STRUCT3.
+       Properly handle 5-7 byte structure returns.
+       * src/pa/ffitarget.h (FFI_TYPE_SMALL_STRUCT1)
+       (FFI_TYPE_SMALL_STRUCT2): Remove.
+       (FFI_TYPE_SMALL_STRUCT3, FFI_TYPE_SMALL_STRUCT5)
+       (FFI_TYPE_SMALL_STRUCT6, FFI_TYPE_SMALL_STRUCT7): Define.
+       * src/pa/linux.S: Mark source file as using PA1.1 assembly.
+       (checksmst1, checksmst2): Remove.
+       (checksmst3): Optimize handling of 3-byte struct returns.
+       (checksmst567): Properly handle 5-7 byte struct returns.
+
+2005-06-15  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       PR libgcj/21943
+       * src/mips/n32.S: Enforce PIC code.
+       * src/mips/o32.S: Likewise.
+
+2005-06-15  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * configure.ac: Treat i*86-*-solaris2.10 and up as X86_64.
+       * configure: Regenerate.
+
+2005-06-01  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/powerpc/ppc_closure.S (ffi_closure_SYSV): Don't use JUMPTARGET
+       to call ffi_closure_helper_SYSV.  Append @local instead.
+       * src/powerpc/sysv.S (ffi_call_SYSV): Likewise for ffi_prep_args_SYSV.
+
+2005-05-17  Kelley Cook  <kcook@gcc.gnu.org>
+
+       * configure.ac: Use AC_C_BIGENDIAN instead of AC_C_BIGENDIAN_CROSS.
+       Use AC_CHECK_SIZEOF instead of AC_COMPILE_CHECK_SIZEOF.
+       * Makefile.am (ACLOCAL_AMFLAGS): Remove -I ../config.
+       * aclocal.m4, configure, fficonfig.h.in, Makefile.in,
+       include/Makefile.in, testsuite/Makefile.in: Regenerate.
+
+2005-05-09  Mike Stump  <mrs@apple.com>
+
+       * configure: Regenerate.
+
+2005-05-08  Richard Henderson  <rth@redhat.com>
+
+       PR libffi/21285
+       * src/alpha/osf.S: Update unwind into to match code.
+
+2005-05-04  Andreas Degert <ad@papyrus-gmbh.de>
+           Richard Henderson  <rth@redhat.com>
+
+       * src/x86/ffi64.c (ffi_prep_cif_machdep): Save sse-used flag in
+       bit 11 of flags.
+       (ffi_call): Mask return type field.  Pass ssecount to ffi_call_unix64.
+       (ffi_prep_closure): Set carry bit if sse-used flag set.
+       * src/x86/unix64.S (ffi_call_unix64): Add ssecount argument.
+       Only load sse registers if ssecount non-zero.
+       (ffi_closure_unix64): Only save sse registers if carry set on entry.
+
+2005-04-29  Ralf Corsepius  <ralf.corsepius@rtems.org>
+
+       * configure.ac: Add i*86-*-rtems*, sparc*-*-rtems*,
+       powerpc-*rtems*, arm*-*-rtems*, sh-*-rtems*.
+       * configure: Regenerate.
+
+2005-04-20  Hans-Peter Nilsson  <hp@axis.com>
+
+       * testsuite/lib/libffi-dg.exp (libffi-dg-test-1): In regsub use,
+       have Tcl8.3-compatible intermediate variable.
+
+2005-04-18  Simon Posnjak <simon.posnjak@siol.net>
+           Hans-Peter Nilsson  <hp@axis.com>
+
+       * Makefile.am: Add CRIS support.
+       * configure.ac: Likewise.
+       * Makefile.in, configure, testsuite/Makefile.in,
+       include/Makefile.in: Regenerate.
+       * src/cris: New directory.
+       * src/cris/ffi.c, src/cris/sysv.S, src/cris/ffitarget.h: New files.
+       * src/prep_cif.c (ffi_prep_cif): Wrap in #ifndef __CRIS__.
+
+       * testsuite/lib/libffi-dg.exp (libffi-dg-test-1): Replace \n with
+       \r?\n in output tests.
+
+2005-04-12  Mike Stump  <mrs@apple.com>
+
+       * configure: Regenerate.
+
+2005-03-30  Hans Boehm  <Hans.Boehm@hp.com>
+
+       * src/ia64/ffitarget.h (ffi_arg): Use long long instead of DI.
+
+2005-03-30  Steve Ellcey  <sje@cup.hp.com>
+
+       * src/ia64/ffitarget.h (ffi_arg) ADD DI attribute.
+       (ffi_sarg) Ditto.
+       * src/ia64/unix.S (ffi_closure_unix): Extend gp
+       to 64 bits in ILP32 mode.
+       Load 64 bits even for short data.
+
+2005-03-23  Mike Stump  <mrs@apple.com>
+
+       * src/powerpc/darwin.S: Update for -m64 multilib.
+       * src/powerpc/darwin_closure.S: Likewise.
+
+2005-03-21  Zack Weinberg  <zack@codesourcery.com>
+
+       * configure.ac: Do not invoke TL_AC_GCC_VERSION.
+       Do not set tool_include_dir.
+       * aclocal.m4, configure, Makefile.in, testsuite/Makefile.in:
+       Regenerate.
+       * include/Makefile.am: Set gcc_version and toollibffidir.
+       * include/Makefile.in: Regenerate.
+
+2005-02-22  Andrew Haley  <aph@redhat.com>
+
+       * src/powerpc/ffi.c (ffi_prep_cif_machdep): Bump alignment to
+       odd-numbered register pairs for 64-bit integer types.
+
+2005-02-23  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       PR libffi/20104
+       * testsuite/libffi.call/return_ll1.c: New test case.
+
+2005-02-11  Janis Johnson  <janis187@us.ibm.com>
+
+       * testsuite/libffi.call/cls_align_longdouble.c: Remove dg-options.
+       * testsuite/libffi.call/float.c: Ditto.
+       * testsuite/libffi.call/float2.c: Ditto.
+       * testsuite/libffi.call/float3.c: Ditto.
+
+2005-02-08  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/frv/ffitarget.h: Remove PPC stuff which does not belong to frv.
+
+2005-01-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
+
+       * testsuite/libffi.special/special.exp (cxx_options): Add
+       -shared-libgcc.
+
+2004-12-31  Richard Henderson  <rth@redhat.com>
+
+       * src/types.c (FFI_AGGREGATE_TYPEDEF): Remove.
+       (FFI_TYPEDEF): Rename from FFI_INTEGRAL_TYPEDEF.  Replace size and
+       offset parameters with a type parameter; deduce size and structure
+       alignment.  Update all users.
+
+2004-12-31  Richard Henderson  <rth@redhat.com>
+
+       * src/types.c (FFI_TYPE_POINTER): Define with sizeof.
+       (FFI_TYPE_LONGDOUBLE): Fix for ia64.
+       * src/ia64/ffitarget.h (struct ffi_ia64_trampoline_struct): Move
+       into ffi_prep_closure.
+       * src/ia64/ia64_flags.h, src/ia64/ffi.c, src/ia64/unix.S: Rewrite
+       from scratch.
+
+2004-12-27  Richard Henderson  <rth@redhat.com>
+
+       * src/x86/unix64.S: Fix typo in unwind info.
+
+2004-12-25  Richard Henderson  <rth@redhat.com>
+
+       * src/x86/ffi64.c (struct register_args): Rename from stackLayout.
+       (enum x86_64_reg_class): Add X86_64_COMPLEX_X87_CLASS.
+       (merge_classes): Check for it.
+       (SSE_CLASS_P): New.
+       (classify_argument): Pass byte_offset by value; perform all updates
+       inside struct case.
+       (examine_argument): Add classes argument; handle
+       X86_64_COMPLEX_X87_CLASS.
+       (ffi_prep_args): Merge into ...
+       (ffi_call): ... here.  Share stack frame with ffi_call_unix64.
+       (ffi_prep_cif_machdep): Setup cif->flags for proper structure return.
+       (ffi_fill_return_value): Remove.
+       (ffi_prep_closure): Remove dead assert.
+       (ffi_closure_unix64_inner): Rename from ffi_closure_UNIX64_inner.
+       Rewrite to use struct register_args instead of va_list.  Create
+       flags for handling structure returns.
+       * src/x86/unix64.S: Remove dead strings.
+       (ffi_call_unix64): Rename from ffi_call_UNIX64.  Rewrite to share
+       stack frame with ffi_call.  Handle structure returns properly.
+       (float2sse, floatfloat2sse, double2sse): Remove.
+       (sse2float, sse2double, sse2floatfloat): Remove.
+       (ffi_closure_unix64): Rename from ffi_closure_UNIX64.  Rewrite
+       to handle structure returns properly.
+
+2004-12-08  David Edelsohn  <edelsohn@gnu.org>
+
+       * Makefile.am (AM_MAKEFLAGS): Remove duplicate LIBCFLAGS and
+       PICFLAG.
+       * Makefile.in: Regenerated.
+
+2004-12-02  Richard Sandiford  <rsandifo@redhat.com>
+
+       * configure.ac: Use TL_AC_GCC_VERSION to set gcc_version.
+       * configure, aclocal.m4, Makefile.in: Regenerate.
+       * include/Makefile.in, testsuite/Makefile.in: Regenerate.
+
+2004-11-29  Kelley Cook  <kcook@gcc.gnu.org>
+
+       * configure: Regenerate for libtool change.
+
+2004-11-25  Kelley Cook  <kcook@gcc.gnu.org>
+
+       * configure: Regenerate for libtool reversion.
+
+2004-11-24  Kelley Cook  <kcook@gcc.gnu.org>
+
+       * configure: Regenerate for libtool change.
+
+2004-11-23  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
+
+       * testsuite/lib/libffi-dg.exp: Use new procs in target-libpath.exp.
+
+2004-11-23  Richard Sandiford  <rsandifo@redhat.com>
+
+       * src/mips/o32.S (ffi_call_O32, ffi_closure_O32): Use jalr instead
+       of jal.  Use an absolute encoding for the frame information.
+
+2004-11-23  Kelley Cook  <kcook@gcc.gnu.org>
+
+       * Makefile.am: Remove no-dependencies.  Add ACLOCAL_AMFLAGS.
+       * acinclude.m4: Delete logic for sincludes.
+       * aclocal.m4, Makefile.in, configure: Regenerate.
+       * include/Makefile: Likewise.
+       * testsuite/Makefile: Likewise.
+
+2004-11-22  Eric Botcazou  <ebotcazou@libertysurf.fr>
+
+       * src/sparc/ffi.c (ffi_prep_closure): Align doubles and 64-bit integers
+       on a 8-byte boundary.
+       * src/sparc/v8.S (ffi_closure_v8): Reserve frame space for arguments.
+
+2004-10-27  Richard Earnshaw  <rearnsha@arm.com>
+
+       * src/arm/ffi.c (ffi_prep_cif_machdep): Handle functions that return
+       long long values.  Round stack allocation to a multiple of 8 bytes
+       for ATPCS compatibility.
+       * src/arm/sysv.S (ffi_call_SYSV): Rework to avoid use of APCS register
+       names.  Handle returning long long types.  Add Thumb and interworking
+       support.  Improve soft-float code.
+
+2004-10-27  Richard Earnshaw  <rearnsha@arm.com>
+
+       * testsuite/lib/libffi-db.exp (load_gcc_lib): New function.
+       (libffi_exit): New function.
+       (libffi_init): Build the testglue wrapper if needed.
+
+2004-10-25  Eric Botcazou  <ebotcazou@libertysurf.fr>
+
+       PR other/18138
+       * testsuite/lib/libffi-dg.exp: Accept more than one multilib libgcc.
+
+2004-10-25  Kazuhiro Inaoka <inaoka.kazuhiro@renesas.com>
+
+       * src/m32r/libffitarget.h (FFI_CLOSURES): Set to 0.
+
+2004-10-20  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/sysv.S (ffi_call_SYSV): Don't align for double data.
+       * testsuite/libffi.call/float3.c: New test case.
+
+2004-10-18  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_prep_closure): Set T bit in trampoline for
+       the function returning a structure pointed with R2.
+       * src/sh/sysv.S (ffi_closure_SYSV): Use R2 as the pointer to
+       the structure return value if T bit set.  Emit position
+       independent code and EH data if PIC.
+
+2004-10-13  Kazuhiro Inaoka  <inaoka.kazuhiro@renesas.com>
+
+       * Makefile.am: Add m32r support.
+       * configure.ac: Likewise.
+       * Makefile.in: Regenerate.
+       * confiugre: Regenerate.
+       * src/types.c: Add m32r port to FFI_INTERNAL_TYPEDEF
+       (uint64, sint64, double, longdouble)
+       * src/m32r: New directory.
+       * src/m32r/ffi.c: New file.
+       * src/m32r/sysv.S: Likewise.
+       * src/m32r/ffitarget.h: Likewise.
+
+2004-10-02  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * testsuite/libffi.call/negint.c: New test case.
+
+2004-09-14  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR libgcj/17465
+       * testsuite/lib/libffi-dg.exp: Don't use global ld_library_path.
+       Set up LD_LIBRARY_PATH, SHLIB_PATH, LD_LIBRARYN32_PATH,
+       LD_LIBRARY64_PATH, LD_LIBRARY_PATH_32, LD_LIBRARY_PATH_64 and
+       DYLD_LIBRARY_PATH.
+
+2004-09-05  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/many_win32.c: Remove whitespaces.
+       * testsuite/libffi.call/promotion.c: Likewise.
+       * testsuite/libffi.call/return_ll.c: Remove unused var. Cleanup
+       whitespaces.
+       * testsuite/libffi.call/return_sc.c: Likewise.
+       * testsuite/libffi.call/return_uc.c: Likewise.
+
+2004-09-05  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/darwin.S: Fix comments and identation.
+       * src/powerpc/darwin_closure.S: Likewise.
+
+2004-09-02  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/ffi_darwin.c: Add flag for longdouble return values.
+       (ffi_prep_args): Handle longdouble arguments.
+       (ffi_prep_cif_machdep): Set flags for longdouble. Calculate space for
+       longdouble.
+       (ffi_closure_helper_DARWIN): Add closure handling for longdouble.
+       * src/powerpc/darwin.S (_ffi_call_DARWIN): Add handling of longdouble
+       values.
+       * src/powerpc/darwin_closure.S (_ffi_closure_ASM): Likewise.
+       * src/types.c: Defined longdouble size and alignment for darwin.
+
+2004-09-02  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/aix.S: Remove whitespaces.
+       * src/powerpc/aix_closure.S: Likewise.
+       * src/powerpc/asm.h: Likewise.
+       * src/powerpc/ffi.c: Likewise.
+       * src/powerpc/ffitarget.h: Likewise.
+       * src/powerpc/linux64.S: Likewise.
+       * src/powerpc/linux64_closure.S: Likewise.
+       * src/powerpc/ppc_closure.S: Likewise.
+       * src/powerpc/sysv.S: Likewise.
+
+2004-08-30  Anthony Green  <green@redhat.com>
+
+       * Makefile.am: Add frv support.
+       * Makefile.in, testsuite/Makefile.in: Rebuilt.
+       * configure.ac: Read configure.host.
+       * configure.in: Read configure.host.
+       * configure.host: New file.  frv-elf needs libgloss.
+       * include/ffi.h.in: Force ffi_closure to have a nice big (8)
+       alignment.  This is needed to frv and shouldn't harm the others.
+       * include/ffi_common.h (ALIGN_DOWN): New macro.
+       * src/frv/ffi.c, src/frv/ffitarget.h, src/frv/eabi.S: New files.
+
+2004-08-24  David Daney  <daney@avtrex.com>
+
+       * testsuite/libffi.call/closure_fn0.c: Xfail mips64* instead of mips*.
+       * testsuite/libffi.call/closure_fn1.c: Likewise.
+       * testsuite/libffi.call/closure_fn2.c  Likewise.
+       * testsuite/libffi.call/closure_fn3.c: Likewise.
+       * testsuite/libffi.call/closure_fn4.c: Likewise.
+       * testsuite/libffi.call/closure_fn5.c: Likewise.
+       * testsuite/libffi.call/cls_18byte.c: Likewise.
+       * testsuite/libffi.call/cls_19byte.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte1.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_64byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_9byte1.c: Likewise.
+       * testsuite/libffi.call/cls_9byte2.c: Likewise.
+       * testsuite/libffi.call/cls_align_double.c: Likewise.
+       * testsuite/libffi.call/cls_align_float.c: Likewise.
+       * testsuite/libffi.call/cls_align_longdouble.c: Likewise.
+       * testsuite/libffi.call/cls_align_pointer.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint64.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint64.c: Likewise.
+       * testsuite/libffi.call/cls_double.c: Likewise.
+       * testsuite/libffi.call/cls_float.c: Likewise.
+       * testsuite/libffi.call/cls_multi_schar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_sshort.c: Likewise.
+       * testsuite/libffi.call/cls_multi_sshortchar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_ushort.c: Likewise.
+       * testsuite/libffi.call/cls_multi_ushortchar.c: Likewise.
+       * testsuite/libffi.call/cls_schar.c: Likewise.
+       * testsuite/libffi.call/cls_sint.c: Likewise.
+       * testsuite/libffi.call/cls_sshort.c: Likewise.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/nested_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct1.c: Likewise.
+       * testsuite/libffi.call/nested_struct2.c: Likewise.
+       * testsuite/libffi.call/nested_struct3.c: Likewise.
+       * testsuite/libffi.call/problem1.c: Likewise.
+       * testsuite/libffi.special/unwindtest.cc: Likewise.
+       * testsuite/libffi.call/cls_12byte.c: Likewise and set return value
+       to zero.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+
+2004-08-23  David Daney <daney@avtrex.com>
+
+       PR libgcj/13141
+       * src/mips/ffitarget.h (FFI_O32_SOFT_FLOAT): New ABI.
+       * src/mips/ffi.c (ffi_prep_args): Fix alignment calculation.
+       (ffi_prep_cif_machdep): Handle FFI_O32_SOFT_FLOAT floating point
+       parameters and return types.
+       (ffi_call): Handle FFI_O32_SOFT_FLOAT ABI.
+       (ffi_prep_closure): Ditto.
+       (ffi_closure_mips_inner_O32): Handle FFI_O32_SOFT_FLOAT ABI, fix
+       alignment calculations.
+       * src/mips/o32.S (ffi_closure_O32): Don't use floating point
+       instructions if FFI_O32_SOFT_FLOAT, make stack frame ABI compliant.
+
+2004-08-14  Casey Marshall <csm@gnu.org>
+
+       * src/mips/ffi.c (ffi_pref_cif_machdep): set `cif->flags' to
+       contain `FFI_TYPE_UINT64' as return type for any 64-bit
+       integer (O32 ABI only).
+       (ffi_prep_closure): new function.
+       (ffi_closure_mips_inner_O32): new function.
+       * src/mips/ffitarget.h: Define `FFI_CLOSURES' and
+       `FFI_TRAMPOLINE_SIZE' appropriately if the ABI is o32.
+       * src/mips/o32.S (ffi_call_O32): add labels for .eh_frame. Return
+       64 bit integers correctly.
+       (ffi_closure_O32): new function.
+       Added DWARF-2 unwind info for both functions.
+
+2004-08-10  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/ffi64.c (ffi_prep_args ): 8-align all stack arguments.
+
+2004-08-01  Robert Millan  <robertmh@gnu.org>
+
+       * configure.ac: Detect knetbsd-gnu and kfreebsd-gnu.
+       * configure: Regenerate.
+
+2004-07-30  Maciej W. Rozycki  <macro@linux-mips.org>
+
+       * acinclude.m4 (AC_FUNC_MMAP_BLACKLIST): Check for <sys/mman.h>
+       and mmap() explicitly instead of relying on preset autoconf cache
+       variables.
+       * aclocal.m4: Regenerate.
+       * configure: Regenerate.
+
+2004-07-11  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * src/s390/ffi.c (ffi_prep_args): Fix C aliasing violation.
+       (ffi_check_float_struct): Remove unused prototype.
+
+2004-06-30  Geoffrey Keating  <geoffk@apple.com>
+
+       * src/powerpc/ffi_darwin.c (flush_icache): ';' is a comment
+       character on Darwin, use '\n\t' instead.
+
+2004-06-26  Matthias Klose  <doko@debian.org>
+
+       * libtool-version: Fix typo in revision/age.
+
+2004-06-17  Matthias Klose  <doko@debian.org>
+
+       * libtool-version: New.
+       * Makefile.am (libffi_la_LDFLAGS): Use -version-info for soname.
+       * Makefile.in: Regenerate.
+
+2004-06-15  Paolo Bonzini  <bonzini@gnu.org>
+
+       * Makefile.am: Remove useless multilib rules.
+       * Makefile.in: Regenerate.
+       * aclocal.m4: Regenerate with automake 1.8.5.
+       * configure.ac: Remove useless multilib configury.
+       * configure: Regenerate.
+
+2004-06-15  Paolo Bonzini  <bonzini@gnu.org>
+
+       * .cvsignore: New file.
+
+2004-06-10  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/ia64/unix.S (ffi_call_unix): Insert group barrier break
+       fp_done.
+       (ffi_closure_UNIX): Fix f14/f15 adjustment if FLOAT_SZ is ever
+       changed from 8.
+
+2004-06-06  Sean McNeil  <sean@mcneil.com>
+
+       * configure.ac: Add x86_64-*-freebsd* support.
+       * configure: Regenerate.
+
+2004-04-26  Joe Buck <jbuck@welsh-buck.org>
+
+       Bug 15093
+       * configure.ac: Test for existence of mmap and sys/mman.h before
+       checking blacklist.  Fix suggested by Jim Wilson.
+       * configure: Regenerate.
+
+2004-04-26  Matt Austern  <austern@apple.com>
+
+       * src/powerpc/darwin.S: Go through a non-lazy pointer for initial
+       FDE location.
+       * src/powerpc/darwin_closure.S: Likewise.
+
+2004-04-24  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/cls_multi_schar.c (main): Fix initialization
+       error. Reported by Thomas Heller <theller@python.net>.
+       * testsuite/libffi.call/cls_multi_sshort.c (main): Likewise.
+       * testsuite/libffi.call/cls_multi_ushort.c (main): Likewise.
+
+2004-03-20  Matthias Klose  <doko@debian.org>
+
+       * src/pa/linux.S: Fix typo.
+
+2004-03-19  Matthias Klose  <doko@debian.org>
+
+       * Makefile.am: Update.
+       * Makefile.in: Regenerate.
+       * src/pa/ffi.h.in: Remove.
+       * src/pa/ffitarget.h: New file.
+
+2004-02-10  Randolph Chung  <tausq@debian.org>
+
+       * Makefile.am: Add PA support.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * configure.ac: Add PA target.
+       * configure: Regenerate.
+       * src/pa/ffi.c: New file.
+       * src/pa/ffi.h.in: Add PA support.
+       * src/pa/linux.S: New file.
+       * prep_cif.c: Add PA support.
+
+2004-03-16  Hosaka Yuji  <hos@tamanegi.org>
+
+       * src/types.c: Fix alignment size of X86_WIN32 case int64 and
+       double.
+       * src/x86/ffi.c (ffi_prep_args): Replace ecif->cif->rtype->type
+       with ecif->cif->flags.
+       (ffi_call, ffi_prep_incoming_args_SYSV): Replace cif->rtype->type
+       with cif->flags.
+       (ffi_prep_cif_machdep): Add X86_WIN32 struct case.
+       (ffi_closure_SYSV): Add 1 or 2-bytes struct case for X86_WIN32.
+       * src/x86/win32.S (retstruct1b, retstruct2b, sc_retstruct1b,
+       sc_retstruct2b): Add for 1 or 2-bytes struct case.
+
+2004-03-15 Kelley Cook <kcook@gcc.gnu.org>
+
+       * configure.in: Rename file to ...
+       * configure.ac: ... this.
+       * fficonfig.h.in: Regenerate.
+       * Makefile.in: Regenerate.
+       * include/Makefile.in: Regenerate.
+       * testsuite/Makefile.in: Regenerate.
+
+2004-03-12  Matt Austern  <austern@apple.com>
+
+       * src/powerpc/darwin.S: Fix EH information so it corresponds to
+       changes in EH format resulting from addition of linkonce support.
+       * src/powerpc/darwin_closure.S: Likewise.
+
+2004-03-11  Andreas Tobler  <a.tobler@schweiz.ch>
+           Paolo Bonzini  <bonzini@gnu.org>
+
+       * Makefile.am (AUTOMAKE_OPTIONS): Set them.
+       Remove VPATH. Remove rules for object files. Remove multilib support.
+       (AM_CCASFLAGS): Add.
+       * configure.in (AC_CONFIG_HEADERS): Relace AM_CONFIG_HEADER.
+       (AC_PREREQ): Bump version to 2.59.
+       (AC_INIT): Fill with version info and bug address.
+       (ORIGINAL_LD_FOR_MULTILIBS): Remove.
+       (AM_ENABLE_MULTILIB): Use this instead of AC_ARG_ENABLE.
+       De-precious CC so that the right flags are passed down to multilibs.
+       (AC_MSG_ERROR): Replace obsolete macro AC_ERROR.
+       (AC_CONFIG_FILES): Replace obsolete macro AC_LINK_FILES.
+       (AC_OUTPUT): Reorganize the output with AC_CONFIG_COMMANDS.
+       * configure: Rebuilt.
+       * aclocal.m4: Likewise.
+       * Makefile.in, include/Makefile.in, testsuite/Makefile.in: Likewise.
+       * fficonfig.h.in: Likewise.
+
+2004-03-11  Andreas Schwab  <schwab@suse.de>
+
+       * src/ia64/ffi.c (ffi_prep_incoming_args_UNIX): Get floating point
+       arguments from fp registers only for the first 8 parameter slots.
+       Don't convert a float parameter when passed in memory.
+
+2004-03-09  Hans-Peter Nilsson  <hp@axis.com>
+
+       * configure: Regenerate for config/accross.m4 correction.
+
+2004-02-25  Matt Kraai  <kraai@alumni.cmu.edu>
+
+       * src/powerpc/ffi.c (ffi_prep_args_SYSV): Change
+       ecif->cif->bytes to bytes.
+       (ffi_prep_cif_machdep): Add braces around nested if statement.
+
+2004-02-09  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/types.c (pointer): POWERPC64 has 8 byte pointers.
+
+       * src/powerpc/ffi.c (ffi_prep_args64): Correct long double handling.
+       (ffi_closure_helper_LINUX64): Fix typo.
+       * testsuite/libffi.call/cls_align_longdouble.c: Pass -mlong-double-128
+       for powerpc64-*-*.
+       * testsuite/libffi.call/float.c: Likewise.
+       * testsuite/libffi.call/float2.c: Likewise.
+
+2004-02-08  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/powerpc/ffi.c (ffi_prep_cif_machdep <FFI_LINUX64>): Correct
+       long double function return and long double arg handling.
+       (ffi_closure_helper_LINUX64): Formatting.  Delete unused "ng" var.
+       Use "end_pfr" instead of "nf".  Correct long double handling.
+       Localise "temp".
+       * src/powerpc/linux64.S (ffi_call_LINUX64): Save f2 long double
+       return value.
+       * src/powerpc/linux64_closure.S (ffi_closure_LINUX64): Allocate
+       space for long double return value.  Adjust stack frame and offsets.
+       Load f2 long double return.
+
+2004-02-07  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/types.c: Use 16 byte long double for POWERPC64.
+
+2004-01-25  Eric Botcazou  <ebotcazou@libertysurf.fr>
+
+       * src/sparc/ffi.c (ffi_prep_args_v9): Shift the parameter array
+       when the structure return address is passed in %o0.
+       (ffi_V9_return_struct): Rename into ffi_v9_layout_struct.
+       (ffi_v9_layout_struct): Align the field following a nested structure
+       on a word boundary.  Use memmove instead of memcpy.
+       (ffi_call): Update call to ffi_V9_return_struct.
+       (ffi_prep_closure): Define 'ctx' only for V8.
+       (ffi_closure_sparc_inner): Clone into ffi_closure_sparc_inner_v8
+       and ffi_closure_sparc_inner_v9.
+       (ffi_closure_sparc_inner_v8): Return long doubles by reference.
+       Always skip the structure return address.  For structures and long
+       doubles, copy the argument directly.
+       (ffi_closure_sparc_inner_v9): Skip the structure return address only
+       if required.  Shift the maximum floating-point slot accordingly.  For
+       big structures, copy the argument directly; otherwise, left-justify the
+       argument and call ffi_v9_layout_struct to lay out the structure on
+       the stack.
+       * src/sparc/v8.S: Undef STACKFRAME before defining it.
+       (ffi_closure_v8): Pass the structure return address.  Update call to
+       ffi_closure_sparc_inner_v8.  Short-circuit FFI_TYPE_INT handling.
+       Skip the 'unimp' insn when returning long doubles and structures.
+       * src/sparc/v9.S: Undef STACKFRAME before defining it.
+       (ffi_closure_v9): Increase the frame size by 2 words.  Short-circuit
+       FFI_TYPE_INT handling.  Load structures both in integers and
+       floating-point registers on return.
+       * README: Update status of the SPARC port.
+
+2004-01-24  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/pyobjc-tc.c (main): Treat result value
+       as of type ffi_arg.
+       * testsuite/libffi.call/struct3.c (main): Fix CHECK.
+
+2004-01-22  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * testsuite/libffi.call/cls_uint.c (cls_ret_uint_fn): Treat result
+       value as of type ffi_arg, not unsigned int.
+
+2004-01-21  Michael Ritzert  <ritzert@t-online.de>
+
+       * ffi64.c (ffi_prep_args): Cast the RHS of an assignment instead
+       of the LHS.
+
+2004-01-12  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/lib/libffi-dg.exp: Set LD_LIBRARY_PATH_32 for
+       Solaris.
+
+2004-01-08  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * testsuite/libffi.call/ffitest.h (allocate_mmap): Cast MAP_FAILED
+       to void *.
+
+2003-12-10  Richard Henderson  <rth@redhat.com>
+
+       * testsuite/libffi.call/cls_align_pointer.c: Cast pointers to
+       size_t instead of int.
+
+2003-12-04  Hosaka Yuji  <hos@tamanegi.org>
+
+       * testsuite/libffi.call/many_win32.c: Include <float.h>.
+       * testsuite/libffi.call/many_win32.c (main): Replace variable
+       int i with unsigned long ul.
+
+       * testsuite/libffi.call/cls_align_uint64.c: New test case.
+       * testsuite/libffi.call/cls_align_sint64.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint32.c: Likewise.
+       * testsuite/libffi.call/cls_align_uint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_sint16.c: Likewise.
+       * testsuite/libffi.call/cls_align_float.c: Likewise.
+       * testsuite/libffi.call/cls_align_double.c: Likewise.
+       * testsuite/libffi.call/cls_align_longdouble.c: Likewise.
+       * testsuite/libffi.call/cls_align_pointer.c: Likewise.
+
+2003-12-02  Hosaka Yuji  <hos@tamanegi.org>
+
+       PR other/13221
+       * src/x86/ffi.c (ffi_prep_args, ffi_prep_incoming_args_SYSV):
+       Align arguments to 32 bits.
+
+2003-12-01  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       PR other/13221
+       * testsuite/libffi.call/cls_multi_sshort.c: New test case.
+       * testsuite/libffi.call/cls_multi_sshortchar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_schar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_ushortchar.c: Likewise.
+       * testsuite/libffi.call/cls_multi_ushort.c: Likewise.
+
+       * testsuite/libffi.special/unwindtest.cc: Cosmetics.
+
+2003-11-26  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
+
+       * testsuite/libffi.call/ffitest.h: Include <fcntl.h>.
+       * testsuite/libffi.special/ffitestcxx.h: Likewise.
+
+2003-11-22  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * Makefile.in: Rebuilt.
+       * configure: Likewise.
+       * testsuite/libffi.special/unwindtest.cc: Convert the mmap to
+       the right type.
+
+2003-11-21  Andreas Jaeger  <aj@suse.de>
+           Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * acinclude.m4: Add AC_FUNC_MMAP_BLACKLIST.
+       * configure.in: Call AC_FUNC_MMAP_BLACKLIST.
+       * Makefile.in: Rebuilt.
+       * aclocal.m4: Likewise.
+       * configure: Likewise.
+       * fficonfig.h.in: Likewise.
+       * testsuite/lib/libffi-dg.exp: Add include dir.
+       * testsuite/libffi.call/ffitest.h: Add MMAP definitions.
+       * testsuite/libffi.special/ffitestcxx.h: Likewise.
+       * testsuite/libffi.call/closure_fn0.c: Use MMAP functionality
+       for ffi_closure if available.
+       * testsuite/libffi.call/closure_fn1.c: Likewise.
+       * testsuite/libffi.call/closure_fn2.c: Likewise.
+       * testsuite/libffi.call/closure_fn3.c: Likewise.
+       * testsuite/libffi.call/closure_fn4.c: Likewise.
+       * testsuite/libffi.call/closure_fn5.c: Likewise.
+       * testsuite/libffi.call/cls_12byte.c: Likewise.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_18byte.c: Likewise.
+       * testsuite/libffi.call/cls_19byte.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte1.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+       * testsuite/libffi.call/cls_64byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_9byte1.c: Likewise.
+       * testsuite/libffi.call/cls_9byte2.c: Likewise.
+       * testsuite/libffi.call/cls_double.c: Likewise.
+       * testsuite/libffi.call/cls_float.c: Likewise.
+       * testsuite/libffi.call/cls_schar.c: Likewise.
+       * testsuite/libffi.call/cls_sint.c: Likewise.
+       * testsuite/libffi.call/cls_sshort.c: Likewise.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/nested_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct1.c: Likewise.
+       * testsuite/libffi.call/nested_struct2.c: Likewise.
+       * testsuite/libffi.call/nested_struct3.c: Likewise.
+       * testsuite/libffi.call/problem1.c: Likewise.
+       * testsuite/libffi.special/unwindtest.cc: Likewise.
+
+2003-11-20  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/lib/libffi-dg.exp: Make the -lgcc_s conditional.
+
+2003-11-19  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/lib/libffi-dg.exp: Add DYLD_LIBRARY_PATH for darwin.
+       Add -lgcc_s to additional flags.
+
+2003-11-12  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * configure.in, include/Makefile.am: PR libgcj/11147, install
+       the ffitarget.h header file in a gcc versioned and target
+       dependent place.
+       * configure: Regenerated.
+       * Makefile.in, include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+
+2003-11-09  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/closure_fn0.c: Print result and check
+       with dg-output to make debugging easier.
+       * testsuite/libffi.call/closure_fn1.c: Likewise.
+       * testsuite/libffi.call/closure_fn2.c: Likewise.
+       * testsuite/libffi.call/closure_fn3.c: Likewise.
+       * testsuite/libffi.call/closure_fn4.c: Likewise.
+       * testsuite/libffi.call/closure_fn5.c: Likewise.
+       * testsuite/libffi.call/cls_12byte.c: Likewise.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_18byte.c: Likewise.
+       * testsuite/libffi.call/cls_19byte.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte1.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+       * testsuite/libffi.call/cls_64byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_9byte1.c: Likewise.
+       * testsuite/libffi.call/cls_9byte2.c: Likewise.
+       * testsuite/libffi.call/cls_double.c: Likewise.
+       * testsuite/libffi.call/cls_float.c: Likewise.
+       * testsuite/libffi.call/cls_schar.c: Likewise.
+       * testsuite/libffi.call/cls_sint.c: Likewise.
+       * testsuite/libffi.call/cls_sshort.c: Likewise.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/problem1.c: Likewise.
+
+       * testsuite/libffi.special/unwindtest.cc: Make ffi_closure
+       static.
+
+2003-11-08  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/cls_9byte2.c: New test case.
+       * testsuite/libffi.call/cls_9byte1.c: Likewise.
+       * testsuite/libffi.call/cls_64byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte1.c: Likewise.
+       * testsuite/libffi.call/cls_19byte.c: Likewise.
+       * testsuite/libffi.call/cls_18byte.c: Likewise.
+       * testsuite/libffi.call/closure_fn4.c: Likewise.
+       * testsuite/libffi.call/closure_fn5.c: Likewise.
+       * testsuite/libffi.call/cls_schar.c: Likewise.
+       * testsuite/libffi.call/cls_sint.c: Likewise.
+       * testsuite/libffi.call/cls_sshort.c: Likewise.
+       * testsuite/libffi.call/nested_struct2.c: Likewise.
+       * testsuite/libffi.call/nested_struct3.c: Likewise.
+
+2003-11-08  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/cls_double.c: Do a check on the result.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/return_sc.c: Cleanup whitespaces.
+
+2003-11-06  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/prep_cif.c (ffi_prep_cif): Move the validity check after
+       the initialization.
+
+2003-10-23  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/java_raw_api.c (ffi_java_ptrarray_to_raw): Replace
+       FFI_ASSERT(FALSE) with FFI_ASSERT(0).
+
+2003-10-22  David Daney  <ddaney@avtrex.com>
+
+       * src/mips/ffitarget.h: Replace undefined UINT32 and friends with
+       __attribute__((__mode__(__SI__))) and friends.
+
+2003-10-22  Andreas Schwab  <schwab@suse.de>
+
+       * src/ia64/ffi.c: Replace FALSE/TRUE with false/true.
+
+2003-10-21  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * configure.in: AC_LINK_FILES(ffitarget.h).
+       * configure: Regenerate.
+       * Makefile.in: Likewise.
+       * include/Makefile.in: Likewise.
+       * testsuite/Makefile.in: Likewise.
+       * fficonfig.h.in: Likewise.
+
+2003-10-21  Paolo Bonzini  <bonzini@gnu.org>
+           Richard Henderson  <rth@redhat.com>
+
+       Avoid that ffi.h includes fficonfig.h.
+
+       * Makefile.am (EXTRA_DIST): Include ffitarget.h files
+       (TARGET_SRC_MIPS_GCC): Renamed to TARGET_SRC_MIPS_IRIX.
+       (TARGET_SRC_MIPS_SGI): Removed.
+       (MIPS_GCC): Renamed to TARGET_SRC_MIPS_IRIX.
+       (MIPS_SGI): Removed.
+       (CLEANFILES): Removed.
+       (mostlyclean-am, clean-am, mostlyclean-sub, clean-sub): New
+       targets.
+       * acconfig.h: Removed.
+       * configure.in: Compute sizeofs only for double and long double.
+       Use them to define and subst HAVE_LONG_DOUBLE.  Include comments
+       into AC_DEFINE instead of using acconfig.h.  Create
+       include/ffitarget.h instead of include/fficonfig.h.  Rename
+       MIPS_GCC to MIPS_IRIX, drop MIPS_SGI since we are in gcc's tree.
+       AC_DEFINE EH_FRAME_FLAGS.
+       * include/Makefile.am (DISTCLEANFILES): New automake macro.
+       (hack_DATA): Add ffitarget.h.
+       * include/ffi.h.in: Remove all system specific definitions.
+       Declare raw API even if it is not installed, why bother?
+       Use limits.h instead of SIZEOF_* to define ffi_type_*.  Do
+       not define EH_FRAME_FLAGS, it is in fficonfig.h now.  Include
+       ffitarget.h instead of fficonfig.h.  Remove ALIGN macro.
+       (UINT_ARG, INT_ARG): Removed, use ffi_arg and ffi_sarg instead.
+       * include/ffi_common.h (bool): Do not define.
+       (ffi_assert): Accept failed assertion.
+       (ffi_type_test): Return void and accept file/line.
+       (FFI_ASSERT): Pass stringized failed assertion.
+       (FFI_ASSERT_AT): New macro.
+       (FFI_ASSERT_VALID_TYPE): New macro.
+       (UINT8, SINT8, UINT16, SINT16, UINT32, SINT32,
+       UINT64, SINT64): Define here with gcc's __attribute__ macro
+       instead of in ffi.h
+       (FLOAT32, ALIGN): Define here instead of in ffi.h
+       * include/ffi-mips.h: Removed.  Its content moved to
+       src/mips/ffitarget.h after separating assembly and C sections.
+       * src/alpha/ffi.c, src/alpha/ffi.c, src/java_raw_api.c
+       src/prep_cif.c, src/raw_api.c, src/ia64/ffi.c,
+       src/mips/ffi.c, src/mips/n32.S, src/mips/o32.S,
+       src/mips/ffitarget.h, src/sparc/ffi.c, src/x86/ffi64.c:
+       SIZEOF_ARG -> FFI_SIZEOF_ARG.
+       * src/ia64/ffi.c: Include stdbool.h (provided by GCC 2.95+).
+       * src/debug.c (ffi_assert): Accept stringized failed assertion.
+       (ffi_type_test): Rewritten.
+       * src/prep-cif.c (initialize_aggregate, ffi_prep_cif): Call
+       FFI_ASSERT_VALID_TYPE.
+       * src/alpha/ffitarget.h, src/arm/ffitarget.h,
+       src/ia64/ffitarget.h, src/m68k/ffitarget.h,
+       src/mips/ffitarget.h, src/powerpc/ffitarget.h,
+       src/s390/ffitarget.h, src/sh/ffitarget.h,
+       src/sh64/ffitarget.h, src/sparc/ffitarget.h,
+       src/x86/ffitarget.h: New files.
+       * src/alpha/osf.S, src/arm/sysv.S, src/ia64/unix.S,
+       src/m68k/sysv.S, src/mips/n32.S, src/mips/o32.S,
+       src/powerpc/aix.S, src/powerpc/darwin.S,
+       src/powerpc/ffi_darwin.c, src/powerpc/linux64.S,
+       src/powerpc/linux64_closure.S, src/powerpc/ppc_closure.S,
+       src/powerpc/sysv.S, src/s390/sysv.S, src/sh/sysv.S,
+       src/sh64/sysv.S, src/sparc/v8.S, src/sparc/v9.S,
+       src/x86/sysv.S, src/x86/unix64.S, src/x86/win32.S:
+       include fficonfig.h
+
+2003-10-20  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * src/mips/ffi.c: Use _ABIN32, _ABIO32 instead of external
+       _MIPS_SIM_NABI32, _MIPS_SIM_ABI32.
+
+2003-10-19  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Declare bytes again.
+       Used when FFI_DEBUG = 1.
+
+2003-10-14  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/types.c (double, longdouble): Default POWERPC64 to 8 byte size
+       and align.
+
+2003-10-06  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * include/ffi_mips.h: Define FFI_MIPS_N32 for N32/N64 ABIs,
+       FFI_MIPS_O32 for O32 ABI.
+
+2003-10-01  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/lib/libffi-dg.exp: Set LD_LIBRARY_PATH_64 for
+       SPARC64. Cleanup whitespaces.
+
+2003-09-19  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * testsuite/libffi.call/closure_fn0.c: Xfail mips, arm,
+       strongarm, xscale. Cleanup whitespaces.
+       * testsuite/libffi.call/closure_fn1.c: Likewise.
+       * testsuite/libffi.call/closure_fn2.c: Likewise.
+       * testsuite/libffi.call/closure_fn3.c: Likewise.
+       * testsuite/libffi.call/cls_12byte.c: Likewise.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_double.c: Likewise.
+       * testsuite/libffi.call/cls_float.c: Likewise.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/nested_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct1.c: Likewise.
+       * testsuite/libffi.call/problem1.c: Likewise.
+       * testsuite/libffi.special/unwindtest.cc: Likewise.
+       * testsuite/libffi.call/pyobjc-tc.c: Cleanup whitespaces.
+
+2003-09-18  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/aix.S: Cleanup whitespaces.
+       * src/powerpc/aix_closure.S: Likewise.
+
+2003-09-18  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/darwin.S: Cleanup whitespaces, comment formatting.
+       * src/powerpc/darwin_closure.S: Likewise.
+       * src/powerpc/ffi_darwin.c: Likewise.
+
+2003-09-18  Andreas Tobler  <a.tobler@schweiz.ch>
+           David Edelsohn  <edelsohn@gnu.org>
+
+       * src/types.c (double): Add AIX and Darwin to the right TYPEDEF.
+       * src/powerpc/aix_closure.S: Remove the pointer to the outgoing
+       parameter stack.
+       * src/powerpc/darwin_closure.S: Likewise.
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Handle structures
+       according to the Darwin/AIX ABI.
+       (ffi_prep_cif_machdep): Likewise.
+       (ffi_closure_helper_DARWIN): Likewise.
+       Remove the outgoing parameter stack logic. Simplify the evaluation
+       of the different CASE types.
+       (ffi_prep_clousure): Avoid the casts on lvalues. Change the branch
+       statement in the trampoline code.
+
+2003-09-18  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (ffi_prep_args): Take account into the alignement
+       for the register size.
+       (ffi_closure_helper_SYSV): Handle the structure return value
+       address correctly.
+       (ffi_closure_helper_SYSV): Return the appropriate type when
+       the registers are used for the structure return value.
+       * src/sh/sysv.S (ffi_closure_SYSV): Fix the stack layout for
+       the 64-bit return value.  Update copyright years.
+
+2003-09-17  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * testsuite/lib/libffi-dg.exp (libffi_target_compile): Search in
+       srcdir for ffi_mips.h.
+
+2003-09-12  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/prep_cif.c (initialize_aggregate): Include tail padding in
+       structure size.
+       * src/powerpc/linux64_closure.S (ffi_closure_LINUX64): Correct
+       placement of float result.
+       * testsuite/libffi.special/unwindtest.cc (closure_test_fn1): Correct
+       cast of "resp" for big-endian 64 bit machines.
+
+2003-09-11  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/types.c (double, longdouble): Merge identical SH and ARM
+       typedefs, and add POWERPC64.
+       * src/powerpc/ffi.c (ffi_prep_args64): Correct next_arg calc for
+       struct split over gpr and rest.
+       (ffi_prep_cif_machdep): Correct intarg_count for structures.
+       * src/powerpc/linux64.S (ffi_call_LINUX64): Fix gpr offsets.
+
+2003-09-09  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/ffi.c (ffi_closure_helper_SYSV) Handle struct
+       passing correctly.
+
+2003-09-09  Alan Modra  <amodra@bigpond.net.au>
+
+       * configure: Regenerate.
+
+2003-09-04  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * Makefile.am: Remove build rules for ffitest.
+       * Makefile.in: Rebuilt.
+
+2003-09-04  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/java_raw_api.c: Include <stdlib.h> to fix compiler warning
+       about implicit declaration of abort().
+
+2003-09-04  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * Makefile.am: Add dejagnu test framework. Fixes PR other/11411.
+       * Makefile.in: Rebuilt.
+       * configure.in: Add dejagnu test framework.
+       * configure: Rebuilt.
+
+       * testsuite/Makefile.am: New file.
+       * testsuite/Makefile.in: Built
+       * testsuite/lib/libffi-dg.exp: New file.
+       * testsuite/config/default.exp: Likewise.
+       * testsuite/libffi.call/call.exp: Likewise.
+       * testsuite/libffi.call/ffitest.h: Likewise.
+       * testsuite/libffi.call/closure_fn0.c: Likewise.
+       * testsuite/libffi.call/closure_fn1.c: Likewise.
+       * testsuite/libffi.call/closure_fn2.c: Likewise.
+       * testsuite/libffi.call/closure_fn3.c: Likewise.
+       * testsuite/libffi.call/cls_1_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_3_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_4_1byte.c: Likewise.
+       * testsuite/libffi.call/cls_2byte.c: Likewise.
+       * testsuite/libffi.call/cls_3byte1.c: Likewise.
+       * testsuite/libffi.call/cls_3byte2.c: Likewise.
+       * testsuite/libffi.call/cls_4byte.c: Likewise.
+       * testsuite/libffi.call/cls_5byte.c: Likewise.
+       * testsuite/libffi.call/cls_6byte.c: Likewise.
+       * testsuite/libffi.call/cls_7byte.c: Likewise.
+       * testsuite/libffi.call/cls_8byte.c: Likewise.
+       * testsuite/libffi.call/cls_12byte.c: Likewise.
+       * testsuite/libffi.call/cls_16byte.c: Likewise.
+       * testsuite/libffi.call/cls_20byte.c: Likewise.
+       * testsuite/libffi.call/cls_24byte.c: Likewise.
+       * testsuite/libffi.call/cls_double.c: Likewise.
+       * testsuite/libffi.call/cls_float.c: Likewise.
+       * testsuite/libffi.call/cls_uchar.c: Likewise.
+       * testsuite/libffi.call/cls_uint.c: Likewise.
+       * testsuite/libffi.call/cls_ulonglong.c: Likewise.
+       * testsuite/libffi.call/cls_ushort.c: Likewise.
+       * testsuite/libffi.call/float.c: Likewise.
+       * testsuite/libffi.call/float1.c: Likewise.
+       * testsuite/libffi.call/float2.c: Likewise.
+       * testsuite/libffi.call/many.c: Likewise.
+       * testsuite/libffi.call/many_win32.c: Likewise.
+       * testsuite/libffi.call/nested_struct.c: Likewise.
+       * testsuite/libffi.call/nested_struct1.c: Likewise.
+       * testsuite/libffi.call/pyobjc-tc.c: Likewise.
+       * testsuite/libffi.call/problem1.c: Likewise.
+       * testsuite/libffi.call/promotion.c: Likewise.
+       * testsuite/libffi.call/return_ll.c: Likewise.
+       * testsuite/libffi.call/return_sc.c: Likewise.
+       * testsuite/libffi.call/return_uc.c: Likewise.
+       * testsuite/libffi.call/strlen.c: Likewise.
+       * testsuite/libffi.call/strlen_win32.c: Likewise.
+       * testsuite/libffi.call/struct1.c: Likewise.
+       * testsuite/libffi.call/struct2.c: Likewise.
+       * testsuite/libffi.call/struct3.c: Likewise.
+       * testsuite/libffi.call/struct4.c: Likewise.
+       * testsuite/libffi.call/struct5.c: Likewise.
+       * testsuite/libffi.call/struct6.c: Likewise.
+       * testsuite/libffi.call/struct7.c: Likewise.
+       * testsuite/libffi.call/struct8.c: Likewise.
+       * testsuite/libffi.call/struct9.c: Likewise.
+       * testsuite/libffi.special/special.exp: New file.
+       * testsuite/libffi.special/ffitestcxx.h: Likewise.
+       * testsuite/libffi.special/unwindtest.cc: Likewise.
+
+
+2003-08-13  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/ffi.c (OFS_INT16): Set 0 for little endian case.  Update
+       copyright years.
+
+2003-08-02  Alan Modra  <amodra@bigpond.net.au>
+
+       * src/powerpc/ffi.c (ffi_prep_args64): Modify for changed gcc
+       structure passing.
+       (ffi_closure_helper_LINUX64): Likewise.
+       * src/powerpc/linux64.S: Remove code writing to parm save area.
+       * src/powerpc/linux64_closure.S (ffi_closure_LINUX64): Use return
+       address in lr from ffi_closure_helper_LINUX64 call to calculate
+       table address.  Optimize function tail.
+
+2003-07-28  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/sparc/ffi.c: Handle all floating point registers.
+       * src/sparc/v9.S: Likewise. Fixes second part of PR target/11410.
+
+2003-07-11  Gerald Pfeifer  <pfeifer@dbai.tuwien.ac.at>
+
+       * README: Note that libffi is not part of GCC.  Update the project
+       URL and status.
+
+2003-06-19  Franz Sirl  <Franz.Sirl-kernel@lauterbach.com>
+
+       * src/powerpc/ppc_closure.S: Include ffi.h.
+
+2003-06-13  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * src/x86/sysv.S: Avoid gas-only .uleb128/.sleb128 directives.
+       Use C style comments.
+
+2003-06-13  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
+
+       * Makefile.am: Add SHmedia support.  Fix a typo of SH support.
+       * Makefile.in: Regenerate.
+       * configure.in (sh64-*-linux*, sh5*-*-linux*): Add target.
+       * configure: Regenerate.
+       * include/ffi.h.in: Add SHmedia support.
+       * src/sh64/ffi.c: New file.
+       * src/sh64/sysv.S: New file.
+
+2003-05-16  Jakub Jelinek  <jakub@redhat.com>
+
+       * configure.in (HAVE_RO_EH_FRAME): Check whether .eh_frame section
+       should be read-only.
+       * configure: Rebuilt.
+       * fficonfig.h.in: Rebuilt.
+       * include/ffi.h.in (EH_FRAME_FLAGS): Define.
+       * src/alpha/osf.S: Use EH_FRAME_FLAGS.
+       * src/powerpc/linux64.S: Likewise.
+       * src/powerpc/linux64_closure.S: Likewise.  Include ffi.h.
+       * src/powerpc/sysv.S: Use EH_FRAME_FLAGS.  Use pcrel encoding
+       if -fpic/-fPIC/-mrelocatable.
+       * src/powerpc/powerpc_closure.S: Likewise.
+       * src/sparc/v8.S: If HAVE_RO_EH_FRAME is defined, don't include
+       #write in .eh_frame flags.
+       * src/sparc/v9.S: Likewise.
+       * src/x86/unix64.S: Use EH_FRAME_FLAGS.
+       * src/x86/sysv.S: Likewise.  Use pcrel encoding if -fpic/-fPIC.
+       * src/s390/sysv.S: Use EH_FRAME_FLAGS.  Include ffi.h.
+
+2003-05-07  Jeff Sturm  <jsturm@one-point.com>
+
+       Fixes PR bootstrap/10656
+       * configure.in (HAVE_AS_REGISTER_PSEUDO_OP): Test assembler
+       support for .register pseudo-op.
+       * src/sparc/v8.S: Use it.
+       * fficonfig.h.in: Rebuilt.
+       * configure: Rebuilt.
+
+2003-04-18  Jakub Jelinek  <jakub@redhat.com>
+
+       * include/ffi.h.in (POWERPC64): Define if 64-bit.
+       (enum ffi_abi): Add FFI_LINUX64 on POWERPC.
+       Make it the default on POWERPC64.
+       (FFI_TRAMPOLINE_SIZE): Define to 24 on POWERPC64.
+       * configure.in: Change powerpc-*-linux* into powerpc*-*-linux*.
+       * configure: Rebuilt.
+       * src/powerpc/ffi.c (hidden): Define.
+       (ffi_prep_args_SYSV): Renamed from
+       ffi_prep_args.  Cast pointers to unsigned long to shut up warnings.
+       (NUM_GPR_ARG_REGISTERS64, NUM_FPR_ARG_REGISTERS64,
+       ASM_NEEDS_REGISTERS64): New.
+       (ffi_prep_args64): New function.
+       (ffi_prep_cif_machdep): Handle FFI_LINUX64 ABI.
+       (ffi_call): Likewise.
+       (ffi_prep_closure): Likewise.
+       (flush_icache): Surround by #ifndef POWERPC64.
+       (ffi_dblfl): New union type.
+       (ffi_closure_helper_SYSV): Use it to avoid aliasing problems.
+       (ffi_closure_helper_LINUX64): New function.
+       * src/powerpc/ppc_closure.S: Surround whole file by #ifndef
+       __powerpc64__.
+       * src/powerpc/sysv.S: Likewise.
+       (ffi_call_SYSV): Rename ffi_prep_args to ffi_prep_args_SYSV.
+       * src/powerpc/linux64.S: New file.
+       * src/powerpc/linux64_closure.S: New file.
+       * Makefile.am (EXTRA_DIST): Add src/powerpc/linux64.S and
+       src/powerpc/linux64_closure.S.
+       (TARGET_SRC_POWERPC): Likewise.
+
+       * src/ffitest.c (closure_test_fn, closure_test_fn1, closure_test_fn2,
+       closure_test_fn3): Fix result printing on big-endian 64-bit
+       machines.
+       (main): Print tst2_arg instead of uninitialized tst2_result.
+
+       * src/ffitest.c (main): Hide what closure pointer really points to
+       from the compiler.
+
+2003-04-16  Richard Earnshaw  <rearnsha@arm.com>
+
+       * configure.in (arm-*-netbsdelf*): Add configuration.
+       (configure): Regenerated.
+
+2003-04-04  Loren J. Rittle  <ljrittle@acm.org>
+
+       * include/Makefile.in: Regenerate.
+
+2003-03-21  Zdenek Dvorak  <rakdver@atrey.karlin.mff.cuni.cz>
+
+       * libffi/include/ffi.h.in: Define X86 instead of X86_64 in 32
+       bit mode.
+       * libffi/src/x86/ffi.c (ffi_closure_SYSV, ffi_closure_raw_SYSV):
+       Receive closure pointer through parameter, read args using
+       __builtin_dwarf_cfa.
+       (FFI_INIT_TRAMPOLINE): Send closure reference through eax.
+
+2003-03-12  Andreas Schwab  <schwab@suse.de>
+
+       * configure.in: Avoid trailing /. in toolexeclibdir.
+       * configure: Rebuilt.
+
+2003-03-03  Andreas Tobler <a.tobler@schweiz.ch>
+
+       * src/powerpc/darwin_closure.S: Recode to fit dynamic libraries.
+
+2003-02-06  Andreas Tobler <a.tobler@schweiz.ch>
+
+       * libffi/src/powerpc/darwin_closure.S:
+       Fix alignement bug, allocate 8 bytes for the result.
+       * libffi/src/powerpc/aix_closure.S:
+       Likewise.
+       * libffi/src/powerpc/ffi_darwin.c:
+       Update stackframe description for aix/darwin_closure.S.
+
+2003-02-06  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/s390/ffi.c (ffi_closure_helper_SYSV): Add hidden visibility
+       attribute.
+
+2003-01-31  Christian Cornelssen  <ccorn@cs.tu-berlin.de>,
+           Andreas Schwab  <schwab@suse.de>
+
+       * configure.in: Adjust command to source config-ml.in to account
+       for changes to the libffi_basedir definition.
+       (libffi_basedir): Remove ${srcdir} from value and include trailing
+       slash if nonempty.
+
+       * configure: Regenerate.
+
+2003-01-29  Franz Sirl  <Franz.Sirl-kernel@lauterbach.com>
+
+       * src/powerpc/ppc_closure.S: Recode to fit shared libs.
+
+2003-01-28  Andrew Haley  <aph@redhat.com>
+
+       * include/ffi.h.in: Enable FFI_CLOSURES for x86_64.
+       * src/x86/ffi64.c (ffi_prep_closure): New.
+       (ffi_closure_UNIX64_inner): New.
+       * src/x86/unix64.S (ffi_closure_UNIX64): New.
+
+2003-01-27  Alexandre Oliva  <aoliva@redhat.com>
+
+       * configure.in (toolexecdir, toolexeclibdir): Set and AC_SUBST.
+       Remove USE_LIBDIR conditional.
+       * Makefile.am (toolexecdir, toolexeclibdir): Don't override.
+       * Makefile.in, configure: Rebuilt.
+
+2003-01027  David Edelsohn  <edelsohn@gnu.org>
+
+       * Makefile.am (TARGET_SRC_POWERPC_AIX): Fix typo.
+       * Makefile.in: Regenerate.
+
+2003-01-22  Andrew Haley  <aph@redhat.com>
+
+       * src/powerpc/darwin.S (_ffi_call_AIX): Add Augmentation size to
+       unwind info.
+
+2003-01-21  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/darwin.S: Add unwind info.
+       * src/powerpc/darwin_closure.S: Likewise.
+
+2003-01-14  Andrew Haley  <aph@redhat.com>
+
+       * src/x86/ffi64.c (ffi_prep_args): Check for void retval.
+       (ffi_prep_cif_machdep): Likewise.
+       * src/x86/unix64.S: Add unwind info.
+
+2003-01-14  Andreas Jaeger  <aj@suse.de>
+
+       * src/ffitest.c (main): Only use ffi_closures if those are
+       supported.
+
+2003-01-13 Andreas Tobler <a.tobler@schweiz.ch>
+
+       * libffi/src/ffitest.c
+        add closure testcases
+
+2003-01-13 Kevin B. Hendricks <khendricks@ivey.uwo.ca>
+
+       * libffi/src/powerpc/ffi.c
+        fix alignment bug for float (4 byte aligned iso 8 byte)
+
+2003-01-09  Geoffrey Keating  <geoffk@apple.com>
+
+       * src/powerpc/ffi_darwin.c: Remove RCS version string.
+       * src/powerpc/darwin.S: Remove RCS version string.
+
+2003-01-03  Jeff Sturm  <jsturm@one-point.com>
+
+       * include/ffi.h.in: Add closure defines for SPARC, SPARC64.
+       * src/ffitest.c (main): Use static storage for closure.
+       * src/sparc/ffi.c (ffi_prep_closure, ffi_closure_sparc_inner): New.
+       * src/sparc/v8.S (ffi_closure_v8): New.
+       * src/sparc/v9.S (ffi_closure_v9): New.
+
+2002-11-10  Ranjit Mathew <rmathew@hotmail.com>
+
+       * include/ffi.h.in: Added FFI_STDCALL ffi_type
+         enumeration for X86_WIN32.
+       * src/x86/win32.S: Added ffi_call_STDCALL function
+         definition.
+       * src/x86/ffi.c (ffi_call/ffi_raw_call): Added
+         switch cases for recognising FFI_STDCALL and
+         calling ffi_call_STDCALL if target is X86_WIN32.
+       * src/ffitest.c (my_stdcall_strlen/stdcall_many):
+         stdcall versions of the "my_strlen" and "many"
+         test functions (for X86_WIN32).
+         Added test cases to test stdcall invocation using
+         these functions.
+
+2002-12-02  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * src/sh/sysv.S: Add DWARF2 unwind info.
+
+2002-11-27  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * src/s390/sysv.S (.eh_frame section): Make section read-only.
+
+2002-11-26  Jim Wilson  <wilson@redhat.com>
+
+       * src/types.c (FFI_TYPE_POINTER): Has size 8 on IA64.
+
+2002-11-23  H.J. Lu <hjl@gnu.org>
+
+       * acinclude.m4: Add dummy AM_PROG_LIBTOOL.
+       Include ../config/accross.m4.
+       * aclocal.m4; Rebuild.
+       * configure: Likewise.
+
+2002-11-15  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * src/s390/sysv.S (.eh_frame section): Adapt to pcrel FDE encoding.
+
+2002-11-11  DJ Delorie  <dj@redhat.com>
+
+       * configure.in: Look for common files in the right place.
+
+2002-10-08  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * src/java_raw_api.c (ffi_java_raw_to_ptrarray): Interpret
+       raw data as _Jv_word values, not ffi_raw.
+       (ffi_java_ptrarray_to_raw): Likewise.
+       (ffi_java_rvalue_to_raw): New function.
+       (ffi_java_raw_call): Call it.
+       (ffi_java_raw_to_rvalue): New function.
+       (ffi_java_translate_args): Call it.
+       * src/ffitest.c (closure_test_fn): Interpret return value
+       as ffi_arg, not int.
+       * src/s390/ffi.c (ffi_prep_cif_machdep): Add missing
+       FFI_TYPE_POINTER case.
+       (ffi_closure_helper_SYSV): Likewise.  Also, assume return
+       values extended to word size.
+
+2002-10-02  Andreas Jaeger  <aj@suse.de>
+
+       * src/x86/ffi64.c (ffi_prep_cif_machdep): Remove debug output.
+
+2002-10-01  Bo Thorsen  <bo@smetana.suse.de>
+
+       * include/ffi.h.in: Fix i386 win32 compilation.
+
+2002-09-30  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * configure.in: Add s390x-*-linux-* target.
+       * configure: Regenerate.
+       * include/ffi.h.in: Define S390X for s390x targets.
+       (FFI_CLOSURES): Define for s390/s390x.
+       (FFI_TRAMPOLINE_SIZE): Likewise.
+       (FFI_NATIVE_RAW_API): Likewise.
+       * src/prep_cif.c (ffi_prep_cif): Do not compute stack space for s390.
+       * src/types.c (FFI_TYPE_POINTER): Use 8-byte pointers on s390x.
+       * src/s390/ffi.c: Major rework of existing code.  Add support for
+       s390x targets.  Add closure support.
+       * src/s390/sysv.S: Likewise.
+
+2002-09-29  Richard Earnshaw  <rearnsha@arm.com>
+
+       * src/arm/sysv.S: Fix typo.
+
+2002-09-28  Richard Earnshaw  <rearnsha@arm.com>
+
+       * src/arm/sysv.S: If we don't have machine/asm.h and the pre-processor
+       has defined __USER_LABEL_PREFIX__, then use it in CNAME.
+       (ffi_call_SYSV): Handle soft-float.
+
+2002-09-27  Bo Thorsen  <bo@suse.de>
+
+       * include/ffi.h.in: Fix multilib x86-64 support.
+
+2002-09-22  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
+
+       * Makefile.am (all-multi): Fix multilib parallel build.
+
+2002-07-19  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * configure.in (sh[34]*-*-linux*): Add brackets.
+       * configure: Regenerate.
+
+2002-07-18  Kaz Kojima  <kkojima@gcc.gnu.org>
+
+       * Makefile.am: Add SH support.
+       * Makefile.in: Regenerate.
+       * configure.in (sh-*-linux*, sh[34]*-*-linux*): Add target.
+       * configure: Regenerate.
+       * include/ffi.h.in: Add SH support.
+       * src/sh/ffi.c: New file.
+       * src/sh/sysv.S: New file.
+       * src/types.c: Add SH support.
+
+2002-07-16  Bo Thorsen  <bo@suse.de>
+
+       * src/x86/ffi64.c: New file that adds x86-64 support.
+       * src/x86/unix64.S: New file that handles argument setup for
+       x86-64.
+       * src/x86/sysv.S: Don't use this on x86-64.
+       * src/x86/ffi.c: Don't use this on x86-64.
+       Remove unused vars.
+       * src/prep_cif.c (ffi_prep_cif): Don't do stack size calculation
+       for x86-64.
+       * src/ffitest.c (struct6): New test that tests a special case in
+       the x86-64 ABI.
+       (struct7): Likewise.
+       (struct8): Likewise.
+       (struct9): Likewise.
+       (closure_test_fn): Silence warning about this when it's not used.
+       (main): Add the new tests.
+       (main): Fix a couple of wrong casts and silence some compiler warnings.
+       * include/ffi.h.in: Add x86-64 ABI definition.
+       * fficonfig.h.in: Regenerate.
+       * Makefile.am: Add x86-64 support.
+       * configure.in: Likewise.
+       * Makefile.in: Regenerate.
+       * configure: Likewise.
+
+2002-06-24  Bo Thorsen  <bo@suse.de>
+
+       * src/types.c: Merge settings for similar architectures.
+       Add x86-64 sizes and alignments.
+
+2002-06-23  Bo Thorsen  <bo@suse.de>
+
+       * src/arm/ffi.c (ffi_prep_args): Remove unused vars.
+       * src/sparc/ffi.c (ffi_prep_args_v8): Likewise.
+       * src/mips/ffi.c (ffi_prep_args): Likewise.
+       * src/m68k/ffi.c (ffi_prep_args): Likewise.
+
+2002-07-18  H.J. Lu  (hjl@gnu.org)
+
+       * Makefile.am (TARGET_SRC_MIPS_LINUX): New.
+       (libffi_la_SOURCES): Support MIPS_LINUX.
+       (libffi_convenience_la_SOURCES): Likewise.
+       * Makefile.in: Regenerated.
+
+       * configure.in (mips64*-*): Skip.
+       (mips*-*-linux*): New.
+       * configure: Regenerated.
+
+       * src/mips/ffi.c: Include <sgidefs.h>.
+
+2002-06-06  Ulrich Weigand  <uweigand@de.ibm.com>
+
+       * src/s390/sysv.S: Save/restore %r6.  Add DWARF-2 unwind info.
+
+2002-05-27  Roger Sayle  <roger@eyesopen.com>
+
+       * src/x86/ffi.c (ffi_prep_args): Remove reference to avn.
+
+2002-05-27  Bo Thorsen  <bo@suse.de>
+
+       * src/x86/ffi.c (ffi_prep_args): Remove unused variable and
+       fix formatting.
+
+2002-05-13  Andreas Tobler  <a.tobler@schweiz.ch>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_closure): Declare fd at
+       beginning of function (for older apple cc).
+
+2002-05-08  Alexandre Oliva  <aoliva@redhat.com>
+
+       * configure.in (ORIGINAL_LD_FOR_MULTILIBS): Preserve LD at
+       script entry, and set LD to it when configuring multilibs.
+       * configure: Rebuilt.
+
+2002-05-05  Jason Thorpe  <thorpej@wasabisystems.com>
+
+       * configure.in (sparc64-*-netbsd*): Add target.
+       (sparc-*-netbsdelf*): Likewise.
+       * configure: Regenerate.
+
+2002-04-28  David S. Miller  <davem@redhat.com>
+
+       * configure.in, configure: Fix SPARC test in previous change.
+
+2002-04-29  Gerhard Tonn  <GerhardTonn@swol.de>
+
+       * Makefile.am: Add Linux for S/390 support.
+       * Makefile.in: Regenerate.
+       * configure.in: Add Linux for S/390 support.
+       * configure: Regenerate.
+       * include/ffi.h.in: Add Linux for S/390 support.
+       * src/s390/ffi.c: New file from libffi CVS tree.
+       * src/s390/sysv.S: New file from libffi CVS tree.
+
+2002-04-28  Jakub Jelinek  <jakub@redhat.com>
+
+       * configure.in (HAVE_AS_SPARC_UA_PCREL): Check for working
+       %r_disp32().
+       * src/sparc/v8.S: Use it.
+       * src/sparc/v9.S: Likewise.
+       * fficonfig.h.in: Rebuilt.
+       * configure: Rebuilt.
+
+2002-04-08  Hans Boehm  <Hans_Boehm@hp.com>
+
+       * src/java_raw_api.c (ffi_java_raw_size): Handle FFI_TYPE_DOUBLE
+       correctly.
+       * src/ia64/unix.S: Add unwind information. Fix comments.
+       Save sp in a way that's compatible with unwind info.
+       (ffi_call_unix): Correctly restore sp in all cases.
+       * src/ia64/ffi.c: Add, fix comments.
+
+2002-04-08  Jakub Jelinek  <jakub@redhat.com>
+
+       * src/sparc/v8.S: Make .eh_frame dependent on target word size.
+
+2002-04-06  Jason Thorpe  <thorpej@wasabisystems.com>
+
+       * configure.in (alpha*-*-netbsd*): Add target.
+       * configure: Regenerate.
+
+2002-04-04  Jeff Sturm  <jsturm@one-point.com>
+
+       * src/sparc/v8.S: Add unwind info.
+       * src/sparc/v9.S: Likewise.
+
+2002-03-30  Krister Walfridsson  <cato@df.lth.se>
+
+       * configure.in: Enable i*86-*-netbsdelf*.
+       * configure: Rebuilt.
+
+2002-03-29  David Billinghurst <David.Billinghurst@riotinto.com>
+
+       PR other/2620
+       * src/mips/n32.s: Delete
+       * src/mips/o32.s: Delete
+
+2002-03-21  Loren J. Rittle  <ljrittle@acm.org>
+
+       * configure.in: Enable alpha*-*-freebsd*.
+       * configure: Rebuilt.
+
+2002-03-17  Bryce McKinlay  <bryce@waitaki.otago.ac.nz>
+
+       * Makefile.am: libfficonvenience -> libffi_convenience.
+       * Makefile.in: Rebuilt.
+
+       * Makefile.am: Define ffitest_OBJECTS.
+       * Makefile.in: Rebuilt.
+
+2002-03-07  Andreas Tobler  <toa@pop.agri.ch>
+           David Edelsohn  <edelsohn@gnu.org>
+
+       * Makefile.am (EXTRA_DIST): Add Darwin and AIX closure files.
+       (TARGET_SRC_POWERPC_AIX): Add aix_closure.S.
+       (TARGET_SRC_POWERPC_DARWIN): Add darwin_closure.S.
+       * Makefile.in: Regenerate.
+       * include/ffi.h.in: Add AIX and Darwin closure definitions.
+       * src/powerpc/ffi_darwin.c (ffi_prep_closure): New function.
+       (flush_icache, flush_range): New functions.
+       (ffi_closure_helper_DARWIN): New function.
+       * src/powerpc/aix_closure.S: New file.
+       * src/powerpc/darwin_closure.S: New file.
+
+2002-02-24  Jeff Sturm  <jsturm@one-point.com>
+
+       * include/ffi.h.in: Add typedef for ffi_arg.
+       * src/ffitest.c (main): Declare rint with ffi_arg.
+
+2002-02-21  Andreas Tobler  <toa@pop.agri.ch>
+
+       * src/powerpc/ffi_darwin.c (ffi_prep_args): Skip appropriate
+       number of GPRs for floating-point arguments.
+
+2002-01-31  Anthony Green  <green@redhat.com>
+
+       * configure: Rebuilt.
+       * configure.in: Replace CHECK_SIZEOF and endian tests with
+       cross-compiler friendly macros.
+       * aclocal.m4 (AC_COMPILE_CHECK_SIZEOF, AC_C_BIGENDIAN_CROSS): New
+       macros.
+
+2002-01-18  David Edelsohn  <edelsohn@gnu.org>
+
+       * src/powerpc/darwin.S (_ffi_call_AIX): New.
+       * src/powerpc/aix.S (ffi_call_DARWIN): New.
+
+2002-01-17  David Edelsohn  <edelsohn@gnu.org>
+
+       * Makefile.am (EXTRA_DIST): Add Darwin and AIX files.
+       (TARGET_SRC_POWERPC_AIX): New.
+       (POWERPC_AIX): New stanza.
+       * Makefile.in: Regenerate.
+       * configure.in: Add AIX case.
+       * configure: Regenerate.
+       * include/ffi.h.in (ffi_abi): Add FFI_AIX.
+       * src/powerpc/ffi_darwin.c (ffi_status): Use "long" to scale frame
+       size.  Fix "long double" support.
+       (ffi_call): Add FFI_AIX case.
+       * src/powerpc/aix.S: New.
+
+2001-10-09  John Hornkvist  <john@toastedmarshmallow.com>
+
+       Implement Darwin PowerPC ABI.
+       * configure.in: Handle powerpc-*-darwin*.
+       * Makefile.am: Set source files for POWERPC_DARWIN.
+       * configure: Rebuilt.
+       * Makefile.in: Rebuilt.
+       * include/ffi.h.in: Define FFI_DARWIN and FFI_DEFAULT_ABI for
+       POWERPC_DARWIN.
+       * src/powerpc/darwin.S: New file.
+       * src/powerpc/ffi_darwin.c: New file.
+
+2001-10-07  Joseph S. Myers  <jsm28@cam.ac.uk>
+
+       * src/x86/ffi.c: Fix spelling error of "separate" as "seperate".
+
+2001-07-16  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * src/x86/sysv.S: Avoid gas-only .balign directive.
+       Use C style comments.
+
+2001-07-16  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * src/alpha/ffi.c (ffi_prep_closure): Avoid gas-only mnemonic.
+       Fixes PR bootstrap/3563.
+
+2001-06-26  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * src/alpha/osf.S (ffi_closure_osf): Use .rdata for ECOFF.
+
+2001-06-25  Rainer Orth  <ro@TechFak.Uni-Bielefeld.DE>
+
+       * configure.in: Recognize sparc*-sun-* host.
+       * configure: Regenerate.
+
+2001-06-06  Andrew Haley  <aph@redhat.com>
+
+       * src/alpha/osf.S (__FRAME_BEGIN__): Conditionalize for ELF.
+
+2001-06-03  Andrew Haley  <aph@redhat.com>
+
+       * src/alpha/osf.S: Add unwind info.
+       * src/powerpc/sysv.S: Add unwind info.
+       * src/powerpc/ppc_closure.S: Likewise.
+
+2000-05-31  Jeff Sturm  <jsturm@one-point.com>
+
+       * configure.in: Fix AC_ARG_ENABLE usage.
+       * configure: Rebuilt.
+
+2001-05-06  Bryce McKinlay  <bryce@waitaki.otago.ac.nz>
+
+       * configure.in: Remove warning about beta code.
+       * configure: Rebuilt.
+
+2001-04-25  Hans Boehm <Hans_Boehm@hp.com>
+
+       * src/ia64/unix.S: Restore stack pointer when returning from
+       ffi_closure_UNIX.
+       * src/ia64/ffi.c: Fix typo in comment.
+
+2001-04-18  Jim Wilson  <wilson@redhat.com>
+
+       * src/ia64/unix.S: Delete unnecessary increment and decrement of loc2
+       to eliminate RAW DV.
+
+2001-04-12  Bryce McKinlay  <bryce@albatross.co.nz>
+
+       * Makefile.am: Make a libtool convenience library.
+       * Makefile.in: Rebuilt.
+
+2001-03-29  Bryce McKinlay  <bryce@albatross.co.nz>
+
+       * configure.in: Use different syntax for subdirectory creation.
+       * configure: Rebuilt.
+
+2001-03-27  Jon Beniston  <jon@beniston.com>
+
+       * configure.in: Added X86_WIN32 target (Win32, CygWin, MingW).
+       * configure: Rebuilt.
+       * Makefile.am: Added X86_WIN32 target support.
+       * Makefile.in: Rebuilt.
+
+       * include/ffi.h.in: Added X86_WIN32 target support.
+
+       * src/ffitest.c: Doesn't run structure tests for X86_WIN32 targets.
+       * src/types.c: Added X86_WIN32 target support.
+
+       * src/x86/win32.S: New file. Based on sysv.S, but with EH
+       stuff removed and made to work with CygWin's gas.
+
+2001-03-26  Bryce McKinlay  <bryce@albatross.co.nz>
+
+       * configure.in: Make target subdirectory in build dir.
+       * Makefile.am: Override suffix based rules to specify correct output
+       subdirectory.
+       * Makefile.in: Rebuilt.
+       * configure: Rebuilt.
+
+2001-03-23  Kevin B Hendricks  <khendricks@ivey.uwo.ca>
+
+       * src/powerpc/ppc_closure.S: New file.
+       * src/powerpc/ffi.c (ffi_prep_args): Fixed ABI compatibility bug
+       involving long long and register pairs.
+       (ffi_prep_closure): New function.
+       (flush_icache): Likewise.
+       (ffi_closure_helper_SYSV): Likewise.
+       * include/ffi.h.in (FFI_CLOSURES): Define on PPC.
+       (FFI_TRAMPOLINE_SIZE): Likewise.
+       (FFI_NATIVE_RAW_API): Likewise.
+       * Makefile.in: Rebuilt.
+       * Makefile.am (EXTRA_DIST): Added src/powerpc/ppc_closure.S.
+       (TARGET_SRC_POWERPC): Likewise.
+
+2001-03-19  Tom Tromey  <tromey@redhat.com>
+
+       * Makefile.in: Rebuilt.
+       * Makefile.am (ffitest_LDFLAGS): New macro.
+
+2001-03-02  Nick Clifton  <nickc@redhat.com>
+
+       * include/ffi.h.in: Remove RCS ident string.
+       * include/ffi_mips.h: Remove RCS ident string.
+       * src/debug.c: Remove RCS ident string.
+       * src/ffitest.c: Remove RCS ident string.
+       * src/prep_cif.c: Remove RCS ident string.
+       * src/types.c: Remove RCS ident string.
+       * src/alpha/ffi.c: Remove RCS ident string.
+       * src/alpha/osf.S: Remove RCS ident string.
+       * src/arm/ffi.c: Remove RCS ident string.
+       * src/arm/sysv.S: Remove RCS ident string.
+       * src/mips/ffi.c: Remove RCS ident string.
+       * src/mips/n32.S: Remove RCS ident string.
+       * src/mips/o32.S: Remove RCS ident string.
+       * src/sparc/ffi.c: Remove RCS ident string.
+       * src/sparc/v8.S: Remove RCS ident string.
+       * src/sparc/v9.S: Remove RCS ident string.
+       * src/x86/ffi.c: Remove RCS ident string.
+       * src/x86/sysv.S: Remove RCS ident string.
+
+2001-02-08  Joseph S. Myers  <jsm28@cam.ac.uk>
+
+       * include/ffi.h.in: Change sourceware.cygnus.com references to
+       gcc.gnu.org.
+
+2000-12-09  Richard Henderson  <rth@redhat.com>
+
+       * src/alpha/ffi.c (ffi_call): Simplify struct return test.
+       (ffi_closure_osf_inner): Index rather than increment avalue
+       and arg_types.  Give ffi_closure_osf the raw return value type.
+       * src/alpha/osf.S (ffi_closure_osf): Handle return value type
+       promotion.
+
+2000-12-07  Richard Henderson  <rth@redhat.com>
+
+       * src/raw_api.c (ffi_translate_args): Fix typo.
+       (ffi_prep_closure): Likewise.
+
+       * include/ffi.h.in [ALPHA]: Define FFI_CLOSURES and
+       FFI_TRAMPOLINE_SIZE.
+       * src/alpha/ffi.c (ffi_prep_cif_machdep): Adjust minimal
+       cif->bytes for new ffi_call_osf implementation.
+       (ffi_prep_args): Absorb into ...
+       (ffi_call): ... here.  Do all stack allocation here and
+       avoid a callback function.
+       (ffi_prep_closure, ffi_closure_osf_inner): New.
+       * src/alpha/osf.S (ffi_call_osf): Reimplement with no callback.
+       (ffi_closure_osf): New.
+
+2000-09-10  Alexandre Oliva  <aoliva@redhat.com>
+
+       * config.guess, config.sub, install-sh: Removed.
+       * ltconfig, ltmain.sh, missing, mkinstalldirs: Likewise.
+       * Makefile.in: Rebuilt.
+
+       * acinclude.m4: Include libtool macros from the top level.
+       * aclocal.m4, configure: Rebuilt.
+
+2000-08-22  Alexandre Oliva  <aoliva@redhat.com>
+
+       * configure.in [i*86-*-freebsd*] (TARGET, TARGETDIR): Set.
+       * configure: Rebuilt.
+
+2000-05-11  Scott Bambrough  <scottb@netwinder.org>
+
+       * libffi/src/arm/sysv.S (ffi_call_SYSV): Doubles are not saved to
+       memory correctly.  Use conditional instructions, not branches where
+       possible.
+
+2000-05-04  Tom Tromey  <tromey@cygnus.com>
+
+       * configure: Rebuilt.
+       * configure.in: Match `arm*-*-linux-*'.
+       From Chris Dornan <cdornan@arm.com>.
+
+2000-04-28  Jakub Jelinek  <jakub@redhat.com>
+
+       * Makefile.am (SUBDIRS): Define.
+       (AM_MAKEFLAGS): Likewise.
+       (Multilib support.): Add section.
+       * Makefile.in: Rebuilt.
+       * ltconfig (extra_compiler_flags, extra_compiler_flags_value):
+       New variables. Set for gcc using -print-multi-lib. Export them
+       to libtool.
+       (sparc64-*-linux-gnu*): Use libsuff 64 for search paths.
+       * ltmain.sh (B|b|V): Don't throw away gcc's -B, -b and -V options
+       for -shared links.
+       (extra_compiler_flags_value, extra_compiler_flags): Check these
+       for extra compiler options which need to be passed down in
+       compiler_flags.
+
+2000-04-16  Anthony Green  <green@redhat.com>
+
+       * configure: Rebuilt.
+       * configure.in: Change i*86-pc-linux* to i*86-*-linux*.
+
+2000-04-14  Jakub Jelinek  <jakub@redhat.com>
+
+       * include/ffi.h.in (SPARC64): Define for 64bit SPARC builds.
+       Set SPARC FFI_DEFAULT_ABI based on SPARC64 define.
+       * src/sparc/ffi.c (ffi_prep_args_v8): Renamed from ffi_prep_args.
+       Replace all void * sizeofs with sizeof(int).
+       Only compare type with FFI_TYPE_LONGDOUBLE if LONGDOUBLE is
+       different than DOUBLE.
+       Remove FFI_TYPE_SINT32 and FFI_TYPE_UINT32 cases (handled elsewhere).
+       (ffi_prep_args_v9): New function.
+       (ffi_prep_cif_machdep): Handle V9 ABI and long long on V8.
+       (ffi_V9_return_struct): New function.
+       (ffi_call): Handle FFI_V9 ABI from 64bit code and FFI_V8 ABI from
+       32bit code (not yet cross-arch calls).
+       * src/sparc/v8.S: Add struct return delay nop.
+       Handle long long.
+       * src/sparc/v9.S: New file.
+       * src/prep_cif.c (ffi_prep_cif): Return structure pointer
+       is used on sparc64 only for structures larger than 32 bytes.
+       Pass by reference for structures is done for structure arguments
+       larger than 16 bytes.
+       * src/ffitest.c (main): Use 64bit rint on sparc64.
+       Run long long tests on sparc.
+       * src/types.c (FFI_TYPE_POINTER): Pointer is 64bit on alpha and
+       sparc64.
+       (FFI_TYPE_LONGDOUBLE): long double is 128 bit aligned to 128 bits
+       on sparc64.
+       * configure.in (sparc-*-linux*): New supported target.
+       (sparc64-*-linux*): Likewise.
+       * configure: Rebuilt.
+       * Makefile.am: Add v9.S to SPARC files.
+       * Makefile.in: Likewise.
+       (LINK): Surround $(CCLD) into double quotes, so that multilib
+       compiles work correctly.
+
+2000-04-04  Alexandre Petit-Bianco  <apbianco@cygnus.com>
+
+       * configure: Rebuilt.
+       * configure.in: (i*86-*-solaris*): New libffi target. Patch
+       proposed by Bryce McKinlay.
+
+2000-03-20  Tom Tromey  <tromey@cygnus.com>
+
+       * Makefile.in: Hand edit for java_raw_api.lo.
+
+2000-03-08  Bryce McKinlay  <bryce@albatross.co.nz>
+
+       * config.guess, config.sub: Update from the gcc tree.
+       Fix for PR libgcj/168.
+
+2000-03-03  Tom Tromey  <tromey@cygnus.com>
+
+       * Makefile.in: Fixed ia64 by hand.
+
+       * configure: Rebuilt.
+       * configure.in (--enable-multilib): New option.
+       (libffi_basedir): New subst.
+       (AC_OUTPUT): Added multilib code.
+
+2000-03-02  Tom Tromey  <tromey@cygnus.com>
+
+       * Makefile.in: Rebuilt.
+       * Makefile.am (TARGET_SRC_IA64): Use `ia64', not `alpha', as
+       directory name.
+
+2000-02-25  Hans Boehm <boehm@acm.org>
+
+       * src/ia64/ffi.c, src/ia64/ia64_flags.h, src/ia64/unix.S: New
+       files.
+       * src/raw_api.c (ffi_translate_args): Fixed typo in argument
+       list.
+       (ffi_prep_raw_closure): Use ffi_translate_args, not
+       ffi_closure_translate.
+       * src/java_raw_api.c: New file.
+       * src/ffitest.c (closure_test_fn): New function.
+       (main): Define `rint' as long long on IA64.  Added new test when
+       FFI_CLOSURES is defined.
+       * include/ffi.h.in (ALIGN): Use size_t, not unsigned.
+       (ffi_abi): Recognize IA64.
+       (ffi_raw): Added `flt' field.
+       Added "Java raw API" code.
+       * configure.in: Recognize ia64.
+       * Makefile.am (TARGET_SRC_IA64): New macro.
+       (libffi_la_common_SOURCES): Added java_raw_api.c.
+       (libffi_la_SOURCES): Define in IA64 case.
+
+2000-01-04  Tom Tromey  <tromey@cygnus.com>
+
+       * Makefile.in: Rebuilt with newer automake.
+
+1999-12-31  Tom Tromey  <tromey@cygnus.com>
+
+       * Makefile.am (INCLUDES): Added -I$(top_srcdir)/src.
+
+1999-09-01  Tom Tromey  <tromey@cygnus.com>
+
+       * include/ffi.h.in: Removed PACKAGE and VERSION defines and
+       undefs.
+       * fficonfig.h.in: Rebuilt.
+       * configure: Rebuilt.
+       * configure.in: Pass 3rd argument to AM_INIT_AUTOMAKE.
+       Use AM_PROG_LIBTOOL (automake 1.4 compatibility).
+       * acconfig.h: Don't #undef PACKAGE or VERSION.
+
+1999-08-09  Anthony Green  <green@cygnus.com>
+
+       * include/ffi.h.in: Try to work around messy header problem
+       with PACKAGE and VERSION.
+
+       * configure: Rebuilt.
+       * configure.in: Change version to 2.00-beta.
+
+       * fficonfig.h.in: Rebuilt.
+       * acconfig.h (FFI_NO_STRUCTS, FFI_NO_RAW_API): Define.
+
+       * src/x86/ffi.c (ffi_raw_call): Rename.
+
+1999-08-02  Kresten Krab Thorup  <krab@dominiq.is.s.u-tokyo.ac.jp>
+
+       * src/x86/ffi.c (ffi_closure_SYSV): New function.
+       (ffi_prep_incoming_args_SYSV): Ditto.
+       (ffi_prep_closure): Ditto.
+       (ffi_closure_raw_SYSV): Ditto.
+       (ffi_prep_raw_closure): More ditto.
+       (ffi_call_raw): Final ditto.
+
+       * include/ffi.h.in: Add definitions for closure and raw API.
+
+       * src/x86/ffi.c (ffi_prep_cif_machdep): Added case for
+       FFI_TYPE_UINT64.
+
+       * Makefile.am (libffi_la_common_SOURCES): Added raw_api.c
+
+       * src/raw_api.c: New file.
+
+       * include/ffi.h.in (ffi_raw): New type.
+       (UINT_ARG, SINT_ARG): New defines.
+       (ffi_closure, ffi_raw_closure): New types.
+       (ffi_prep_closure, ffi_prep_raw_closure): New declarations.
+
+       * configure.in: Add check for endianness and sizeof void*.
+
+       * src/x86/sysv.S (ffi_call_SYSV): Call fixup routine via argument,
+       instead of directly.
+
+       * configure: Rebuilt.
+
+Thu Jul  8 14:28:42 1999  Anthony Green  <green@cygnus.com>
+
+       * configure.in: Add x86 and powerpc BeOS configurations.
+       From Makoto Kato <m_kato@ga2.so-net.ne.jp>.
+
+1999-05-09  Anthony Green  <green@cygnus.com>
+
+       * configure.in: Add warning about this being beta code.
+       Remove src/Makefile.am from the picture.
+       * configure: Rebuilt.
+
+       * Makefile.am: Move logic from src/Makefile.am.  Add changes
+       to support libffi as a target library.
+       * Makefile.in: Rebuilt.
+
+       * aclocal.m4, config.guess, config.sub, ltconfig, ltmain.sh:
+       Upgraded to new autoconf, automake, libtool.
+
+       * README: Tweaks.
+
+       * LICENSE: Update copyright date.
+
+       * src/Makefile.am, src/Makefile.in: Removed.
+
+1998-11-29  Anthony Green  <green@cygnus.com>
+
+       * include/ChangeLog: Removed.
+       * src/ChangeLog: Removed.
+       * src/mips/ChangeLog: Removed.
+       * src/sparc/ChangeLog: Remboved.
+       * src/x86/ChangeLog: Removed.
+
+       * ChangeLog.v1: Created.
diff --git a/.pc/apple-fixes/include/ffi.h.in b/.pc/apple-fixes/include/ffi.h.in
new file mode 100644 (file)
index 0000000..4f3b3de
--- /dev/null
@@ -0,0 +1,457 @@
+/* -----------------------------------------------------------------*-C-*-
+   libffi @VERSION@ - Copyright (c) 2011 Anthony Green
+                    - Copyright (c) 1996-2003, 2007, 2008 Red Hat, Inc.
+
+   Permission is hereby granted, free of charge, to any person
+   obtaining a copy of this software and associated documentation
+   files (the ``Software''), to deal in the Software without
+   restriction, including without limitation the rights to use, copy,
+   modify, merge, publish, distribute, sublicense, and/or sell copies
+   of the Software, and to permit persons to whom the Software is
+   furnished to do so, subject to the following conditions:
+
+   The above copyright notice and this permission notice shall be
+   included in all copies or substantial portions of the Software.
+
+   THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND,
+   EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+   NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+   HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+   WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+   OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+   DEALINGS IN THE SOFTWARE.
+
+   ----------------------------------------------------------------------- */
+
+/* -------------------------------------------------------------------
+   The basic API is described in the README file.
+
+   The raw API is designed to bypass some of the argument packing
+   and unpacking on architectures for which it can be avoided.
+
+   The closure API allows interpreted functions to be packaged up
+   inside a C function pointer, so that they can be called as C functions,
+   with no understanding on the client side that they are interpreted.
+   It can also be used in other cases in which it is necessary to package
+   up a user specified parameter and a function pointer as a single
+   function pointer.
+
+   The closure API must be implemented in order to get its functionality,
+   e.g. for use by gij.  Routines are provided to emulate the raw API
+   if the underlying platform doesn't allow faster implementation.
+
+   More details on the raw and cloure API can be found in:
+
+   http://gcc.gnu.org/ml/java/1999-q3/msg00138.html
+
+   and
+
+   http://gcc.gnu.org/ml/java/1999-q3/msg00174.html
+   -------------------------------------------------------------------- */
+
+#ifndef LIBFFI_H
+#define LIBFFI_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Specify which architecture libffi is configured for. */
+#ifndef @TARGET@
+#define @TARGET@
+#endif
+
+/* ---- System configuration information --------------------------------- */
+
+#include <ffitarget.h>
+
+#ifndef LIBFFI_ASM
+
+#ifdef _MSC_VER
+#define __attribute__(X)
+#endif
+
+#include <stddef.h>
+#include <limits.h>
+
+/* LONG_LONG_MAX is not always defined (not if STRICT_ANSI, for example).
+   But we can find it either under the correct ANSI name, or under GNU
+   C's internal name.  */
+
+#define FFI_64_BIT_MAX 9223372036854775807
+
+#ifdef LONG_LONG_MAX
+# define FFI_LONG_LONG_MAX LONG_LONG_MAX
+#else
+# ifdef LLONG_MAX
+#  define FFI_LONG_LONG_MAX LLONG_MAX
+#  ifdef _AIX52 /* or newer has C99 LLONG_MAX */
+#   undef FFI_64_BIT_MAX
+#   define FFI_64_BIT_MAX 9223372036854775807LL
+#  endif /* _AIX52 or newer */
+# else
+#  ifdef __GNUC__
+#   define FFI_LONG_LONG_MAX __LONG_LONG_MAX__
+#  endif
+#  ifdef _AIX /* AIX 5.1 and earlier have LONGLONG_MAX */
+#   ifndef __PPC64__
+#    if defined (__IBMC__) || defined (__IBMCPP__)
+#     define FFI_LONG_LONG_MAX LONGLONG_MAX
+#    endif
+#   endif /* __PPC64__ */
+#   undef  FFI_64_BIT_MAX
+#   define FFI_64_BIT_MAX 9223372036854775807LL
+#  endif
+# endif
+#endif
+
+/* The closure code assumes that this works on pointers, i.e. a size_t */
+/* can hold a pointer.                                                 */
+
+typedef struct _ffi_type
+{
+  size_t size;
+  unsigned short alignment;
+  unsigned short type;
+  struct _ffi_type **elements;
+} ffi_type;
+
+#ifndef LIBFFI_HIDE_BASIC_TYPES
+#if SCHAR_MAX == 127
+# define ffi_type_uchar                ffi_type_uint8
+# define ffi_type_schar                ffi_type_sint8
+#else
+ #error "char size not supported"
+#endif
+
+#if SHRT_MAX == 32767
+# define ffi_type_ushort       ffi_type_uint16
+# define ffi_type_sshort       ffi_type_sint16
+#elif SHRT_MAX == 2147483647
+# define ffi_type_ushort       ffi_type_uint32
+# define ffi_type_sshort       ffi_type_sint32
+#else
+ #error "short size not supported"
+#endif
+
+#if INT_MAX == 32767
+# define ffi_type_uint         ffi_type_uint16
+# define ffi_type_sint         ffi_type_sint16
+#elif INT_MAX == 2147483647
+# define ffi_type_uint         ffi_type_uint32
+# define ffi_type_sint         ffi_type_sint32
+#elif INT_MAX == 9223372036854775807
+# define ffi_type_uint         ffi_type_uint64
+# define ffi_type_sint         ffi_type_sint64
+#else
+ #error "int size not supported"
+#endif
+
+#if LONG_MAX == 2147483647
+# if FFI_LONG_LONG_MAX != FFI_64_BIT_MAX
+ #error "no 64-bit data type supported"
+# endif
+#elif LONG_MAX != FFI_64_BIT_MAX
+ #error "long size not supported"
+#endif
+
+#if LONG_MAX == 2147483647
+# define ffi_type_ulong        ffi_type_uint32
+# define ffi_type_slong        ffi_type_sint32
+#elif LONG_MAX == FFI_64_BIT_MAX
+# define ffi_type_ulong        ffi_type_uint64
+# define ffi_type_slong        ffi_type_sint64
+#else
+ #error "long size not supported"
+#endif
+
+/* Need minimal decorations for DLLs to works on Windows. */
+/* GCC has autoimport and autoexport.  Rely on Libtool to */
+/* help MSVC export from a DLL, but always declare data   */
+/* to be imported for MSVC clients.  This costs an extra  */
+/* indirection for MSVC clients using the static version  */
+/* of the library, but don't worry about that.  Besides,  */
+/* as a workaround, they can define FFI_BUILDING if they  */
+/* *know* they are going to link with the static library. */
+#if defined _MSC_VER && !defined FFI_BUILDING
+#define FFI_EXTERN extern __declspec(dllimport)
+#else
+#define FFI_EXTERN extern
+#endif
+
+/* These are defined in types.c */
+FFI_EXTERN ffi_type ffi_type_void;
+FFI_EXTERN ffi_type ffi_type_uint8;
+FFI_EXTERN ffi_type ffi_type_sint8;
+FFI_EXTERN ffi_type ffi_type_uint16;
+FFI_EXTERN ffi_type ffi_type_sint16;
+FFI_EXTERN ffi_type ffi_type_uint32;
+FFI_EXTERN ffi_type ffi_type_sint32;
+FFI_EXTERN ffi_type ffi_type_uint64;
+FFI_EXTERN ffi_type ffi_type_sint64;
+FFI_EXTERN ffi_type ffi_type_float;
+FFI_EXTERN ffi_type ffi_type_double;
+FFI_EXTERN ffi_type ffi_type_pointer;
+
+#if @HAVE_LONG_DOUBLE@
+FFI_EXTERN ffi_type ffi_type_longdouble;
+#else
+#define ffi_type_longdouble ffi_type_double
+#endif
+#endif /* LIBFFI_HIDE_BASIC_TYPES */
+
+typedef enum {
+  FFI_OK = 0,
+  FFI_BAD_TYPEDEF,
+  FFI_BAD_ABI
+} ffi_status;
+
+typedef unsigned FFI_TYPE;
+
+typedef struct {
+  ffi_abi abi;
+  unsigned nargs;
+  ffi_type **arg_types;
+  ffi_type *rtype;
+  unsigned bytes;
+  unsigned flags;
+#ifdef FFI_EXTRA_CIF_FIELDS
+  FFI_EXTRA_CIF_FIELDS;
+#endif
+} ffi_cif;
+
+/* Used internally, but overridden by some architectures */
+ffi_status ffi_prep_cif_core(ffi_cif *cif,
+                            ffi_abi abi,
+                            unsigned int isvariadic,
+                            unsigned int nfixedargs,
+                            unsigned int ntotalargs,
+                            ffi_type *rtype,
+                            ffi_type **atypes);
+
+/* ---- Definitions for the raw API -------------------------------------- */
+
+#ifndef FFI_SIZEOF_ARG
+# if LONG_MAX == 2147483647
+#  define FFI_SIZEOF_ARG        4
+# elif LONG_MAX == FFI_64_BIT_MAX
+#  define FFI_SIZEOF_ARG        8
+# endif
+#endif
+
+#ifndef FFI_SIZEOF_JAVA_RAW
+#  define FFI_SIZEOF_JAVA_RAW FFI_SIZEOF_ARG
+#endif
+
+typedef union {
+  ffi_sarg  sint;
+  ffi_arg   uint;
+  float            flt;
+  char      data[FFI_SIZEOF_ARG];
+  void*     ptr;
+} ffi_raw;
+
+#if FFI_SIZEOF_JAVA_RAW == 4 && FFI_SIZEOF_ARG == 8
+/* This is a special case for mips64/n32 ABI (and perhaps others) where
+   sizeof(void *) is 4 and FFI_SIZEOF_ARG is 8.  */
+typedef union {
+  signed int   sint;
+  unsigned int uint;
+  float                flt;
+  char         data[FFI_SIZEOF_JAVA_RAW];
+  void*                ptr;
+} ffi_java_raw;
+#else
+typedef ffi_raw ffi_java_raw;
+#endif
+
+
+void ffi_raw_call (ffi_cif *cif,
+                  void (*fn)(void),
+                  void *rvalue,
+                  ffi_raw *avalue);
+
+void ffi_ptrarray_to_raw (ffi_cif *cif, void **args, ffi_raw *raw);
+void ffi_raw_to_ptrarray (ffi_cif *cif, ffi_raw *raw, void **args);
+size_t ffi_raw_size (ffi_cif *cif);
+
+/* This is analogous to the raw API, except it uses Java parameter     */
+/* packing, even on 64-bit machines.  I.e. on 64-bit machines          */
+/* longs and doubles are followed by an empty 64-bit word.             */
+
+void ffi_java_raw_call (ffi_cif *cif,
+                       void (*fn)(void),
+                       void *rvalue,
+                       ffi_java_raw *avalue);
+
+void ffi_java_ptrarray_to_raw (ffi_cif *cif, void **args, ffi_java_raw *raw);
+void ffi_java_raw_to_ptrarray (ffi_cif *cif, ffi_java_raw *raw, void **args);
+size_t ffi_java_raw_size (ffi_cif *cif);
+
+/* ---- Definitions for closures ----------------------------------------- */
+
+#if FFI_CLOSURES
+
+#ifdef _MSC_VER
+__declspec(align(8))
+#endif
+typedef struct {
+  char tramp[FFI_TRAMPOLINE_SIZE];
+  ffi_cif   *cif;
+  void     (*fun)(ffi_cif*,void*,void**,void*);
+  void      *user_data;
+#ifdef __GNUC__
+} ffi_closure __attribute__((aligned (8)));
+#else
+} ffi_closure;
+# ifdef __sgi
+#  pragma pack 0
+# endif
+#endif
+
+void *ffi_closure_alloc (size_t size, void **code);
+void ffi_closure_free (void *);
+
+ffi_status
+ffi_prep_closure (ffi_closure*,
+                 ffi_cif *,
+                 void (*fun)(ffi_cif*,void*,void**,void*),
+                 void *user_data);
+
+ffi_status
+ffi_prep_closure_loc (ffi_closure*,
+                     ffi_cif *,
+                     void (*fun)(ffi_cif*,void*,void**,void*),
+                     void *user_data,
+                     void*codeloc);
+
+#ifdef __sgi
+# pragma pack 8
+#endif
+typedef struct {
+  char tramp[FFI_TRAMPOLINE_SIZE];
+
+  ffi_cif   *cif;
+
+#if !FFI_NATIVE_RAW_API
+
+  /* if this is enabled, then a raw closure has the same layout 
+     as a regular closure.  We use this to install an intermediate 
+     handler to do the transaltion, void** -> ffi_raw*. */
+
+  void     (*translate_args)(ffi_cif*,void*,void**,void*);
+  void      *this_closure;
+
+#endif
+
+  void     (*fun)(ffi_cif*,void*,ffi_raw*,void*);
+  void      *user_data;
+
+} ffi_raw_closure;
+
+typedef struct {
+  char tramp[FFI_TRAMPOLINE_SIZE];
+
+  ffi_cif   *cif;
+
+#if !FFI_NATIVE_RAW_API
+
+  /* if this is enabled, then a raw closure has the same layout 
+     as a regular closure.  We use this to install an intermediate 
+     handler to do the transaltion, void** -> ffi_raw*. */
+
+  void     (*translate_args)(ffi_cif*,void*,void**,void*);
+  void      *this_closure;
+
+#endif
+
+  void     (*fun)(ffi_cif*,void*,ffi_java_raw*,void*);
+  void      *user_data;
+
+} ffi_java_raw_closure;
+
+ffi_status
+ffi_prep_raw_closure (ffi_raw_closure*,
+                     ffi_cif *cif,
+                     void (*fun)(ffi_cif*,void*,ffi_raw*,void*),
+                     void *user_data);
+
+ffi_status
+ffi_prep_raw_closure_loc (ffi_raw_closure*,
+                         ffi_cif *cif,
+                         void (*fun)(ffi_cif*,void*,ffi_raw*,void*),
+                         void *user_data,
+                         void *codeloc);
+
+ffi_status
+ffi_prep_java_raw_closure (ffi_java_raw_closure*,
+                          ffi_cif *cif,
+                          void (*fun)(ffi_cif*,void*,ffi_java_raw*,void*),
+                          void *user_data);
+
+ffi_status
+ffi_prep_java_raw_closure_loc (ffi_java_raw_closure*,
+                              ffi_cif *cif,
+                              void (*fun)(ffi_cif*,void*,ffi_java_raw*,void*),
+                              void *user_data,
+                              void *codeloc);
+
+#endif /* FFI_CLOSURES */
+
+/* ---- Public interface definition -------------------------------------- */
+
+ffi_status ffi_prep_cif(ffi_cif *cif,
+                       ffi_abi abi,
+                       unsigned int nargs,
+                       ffi_type *rtype,
+                       ffi_type **atypes);
+
+ffi_status ffi_prep_cif_var(ffi_cif *cif,
+                           ffi_abi abi,
+                           unsigned int nfixedargs,
+                           unsigned int ntotalargs,
+                           ffi_type *rtype,
+                           ffi_type **atypes);
+
+void ffi_call(ffi_cif *cif,
+             void (*fn)(void),
+             void *rvalue,
+             void **avalue);
+
+/* Useful for eliminating compiler warnings */
+#define FFI_FN(f) ((void (*)(void))f)
+
+/* ---- Definitions shared with assembly code ---------------------------- */
+
+#endif
+
+/* If these change, update src/mips/ffitarget.h. */
+#define FFI_TYPE_VOID       0    
+#define FFI_TYPE_INT        1
+#define FFI_TYPE_FLOAT      2    
+#define FFI_TYPE_DOUBLE     3
+#if @HAVE_LONG_DOUBLE@
+#define FFI_TYPE_LONGDOUBLE 4
+#else
+#define FFI_TYPE_LONGDOUBLE FFI_TYPE_DOUBLE
+#endif
+#define FFI_TYPE_UINT8      5   
+#define FFI_TYPE_SINT8      6
+#define FFI_TYPE_UINT16     7 
+#define FFI_TYPE_SINT16     8
+#define FFI_TYPE_UINT32     9
+#define FFI_TYPE_SINT32     10
+#define FFI_TYPE_UINT64     11
+#define FFI_TYPE_SINT64     12
+#define FFI_TYPE_STRUCT     13
+#define FFI_TYPE_POINTER    14
+
+/* This should always refer to the last type code (for sanity checks) */
+#define FFI_TYPE_LAST       FFI_TYPE_POINTER
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/.pc/apple-fixes/src/arm/sysv.S b/.pc/apple-fixes/src/arm/sysv.S
new file mode 100644 (file)
index 0000000..a6728cf
--- /dev/null
@@ -0,0 +1,505 @@
+/* -----------------------------------------------------------------------
+   sysv.S - Copyright (c) 1998, 2008, 2011 Red Hat, Inc.
+           Copyright (c) 2011 Plausible Labs Cooperative, Inc.
+   
+   ARM Foreign Function Interface 
+
+   Permission is hereby granted, free of charge, to any person obtaining
+   a copy of this software and associated documentation files (the
+   ``Software''), to deal in the Software without restriction, including
+   without limitation the rights to use, copy, modify, merge, publish,
+   distribute, sublicense, and/or sell copies of the Software, and to
+   permit persons to whom the Software is furnished to do so, subject to
+   the following conditions:
+
+   The above copyright notice and this permission notice shall be included
+   in all copies or substantial portions of the Software.
+
+   THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND,
+   EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+   NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+   HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+   WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+   OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+   DEALINGS IN THE SOFTWARE.
+   ----------------------------------------------------------------------- */
+
+#define LIBFFI_ASM     
+#include <fficonfig.h>
+#include <ffi.h>
+#ifdef HAVE_MACHINE_ASM_H
+#include <machine/asm.h>
+#else
+#ifdef __USER_LABEL_PREFIX__
+#define CONCAT1(a, b) CONCAT2(a, b)
+#define CONCAT2(a, b) a ## b
+
+/* Use the right prefix for global labels.  */
+#define CNAME(x) CONCAT1 (__USER_LABEL_PREFIX__, x)
+#else
+#define CNAME(x) x
+#endif
+#ifdef __APPLE__
+#define ENTRY(x) .globl CNAME(x); CNAME(x):
+#else
+#define ENTRY(x) .globl CNAME(x); .type CNAME(x),%function; CNAME(x):
+#endif /* __APPLE__ */
+#endif
+
+#ifdef __ELF__
+#define LSYM(x) .x
+#else
+#define LSYM(x) x
+#endif
+
+/* Use the SOFTFP return value ABI on Mac OS X, as per the iOS ABI
+  Function Call Guide */
+#ifdef __APPLE__
+#define __SOFTFP__
+#endif
+
+/* We need a better way of testing for this, but for now, this is all 
+   we can do.  */
+@ This selects the minimum architecture level required.
+#define __ARM_ARCH__ 3
+
+#if defined(__ARM_ARCH_4__) || defined(__ARM_ARCH_4T__)
+# undef __ARM_ARCH__
+# define __ARM_ARCH__ 4
+#endif
+        
+#if defined(__ARM_ARCH_5__) || defined(__ARM_ARCH_5T__) \
+       || defined(__ARM_ARCH_5E__) || defined(__ARM_ARCH_5TE__) \
+       || defined(__ARM_ARCH_5TEJ__)
+# undef __ARM_ARCH__
+# define __ARM_ARCH__ 5
+#endif
+
+#if defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) \
+        || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) \
+        || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) \
+       || defined(__ARM_ARCH_6M__)
+# undef __ARM_ARCH__
+# define __ARM_ARCH__ 6
+#endif
+
+#if defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) \
+        || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) \
+       || defined(__ARM_ARCH_7EM__)
+# undef __ARM_ARCH__
+# define __ARM_ARCH__ 7
+#endif
+
+#if __ARM_ARCH__ >= 5
+# define call_reg(x)   blx     x
+#elif defined (__ARM_ARCH_4T__)
+# define call_reg(x)   mov     lr, pc ; bx     x
+# if defined(__thumb__) || defined(__THUMB_INTERWORK__)
+#  define __INTERWORKING__
+# endif
+#else
+# define call_reg(x)   mov     lr, pc ; mov    pc, x
+#endif
+
+/* Conditionally compile unwinder directives.  */
+#ifdef __ARM_EABI__
+#define UNWIND
+#else
+#define UNWIND @
+#endif 
+
+       
+#if defined(__thumb__) && !defined(__THUMB_INTERWORK__)
+.macro ARM_FUNC_START name
+       .text
+       .align 0
+       .thumb
+       .thumb_func
+#ifdef __APPLE__
+       ENTRY($0)
+#else
+       ENTRY(\name)
+#endif
+       bx      pc
+       nop
+       .arm
+       UNWIND .fnstart
+/* A hook to tell gdb that we've switched to ARM mode.  Also used to call
+   directly from other local arm routines.  */
+#ifdef __APPLE__
+_L__$0:
+#else
+_L__\name:
+#endif
+.endm
+#else
+.macro ARM_FUNC_START name
+       .text
+       .align 0
+       .arm
+#ifdef __APPLE__
+       ENTRY($0)
+#else
+       ENTRY(\name)
+#endif
+       UNWIND .fnstart
+.endm
+#endif
+
+.macro RETLDM  regs=, cond=, dirn=ia
+#if defined (__INTERWORKING__)
+       .ifc "\regs",""
+       ldr\cond        lr, [sp], #4
+       .else
+       ldm\cond\dirn   sp!, {\regs, lr}
+       .endif
+       bx\cond lr
+#else
+       .ifc "\regs",""
+       ldr\cond        pc, [sp], #4
+       .else
+       ldm\cond\dirn   sp!, {\regs, pc}
+       .endif
+#endif
+.endm
+
+       @ r0:   ffi_prep_args
+       @ r1:   &ecif
+       @ r2:   cif->bytes
+       @ r3:   fig->flags
+       @ sp+0: ecif.rvalue
+
+       @ This assumes we are using gas.
+ARM_FUNC_START ffi_call_SYSV
+       @ Save registers
+        stmfd  sp!, {r0-r3, fp, lr}
+       UNWIND .save    {r0-r3, fp, lr}
+       mov     fp, sp
+
+       UNWIND .setfp   fp, sp
+
+       @ Make room for all of the new args.
+       sub     sp, fp, r2
+
+       @ Place all of the ffi_prep_args in position
+       mov     r0, sp
+       @     r1 already set
+
+       @ Call ffi_prep_args(stack, &ecif)
+       bl      ffi_prep_args
+
+       @ move first 4 parameters in registers
+       ldmia   sp, {r0-r3}
+
+       @ and adjust stack
+       sub     lr, fp, sp      @ cif->bytes == fp - sp
+       ldr     ip, [fp]        @ load fn() in advance
+       cmp     lr, #16
+       movhs   lr, #16
+       add     sp, sp, lr
+
+       @ call (fn) (...)
+       call_reg(ip)
+       
+       @ Remove the space we pushed for the args
+       mov     sp, fp
+
+       @ Load r2 with the pointer to storage for the return value
+       ldr     r2, [sp, #24]
+
+       @ Load r3 with the return type code 
+       ldr     r3, [sp, #12]
+
+       @ If the return value pointer is NULL, assume no return value.
+       cmp     r2, #0
+       beq     LSYM(Lepilogue)
+
+@ return INT
+       cmp     r3, #FFI_TYPE_INT
+#if defined(__SOFTFP__) || defined(__ARM_EABI__)
+       cmpne   r3, #FFI_TYPE_FLOAT
+#endif
+       streq   r0, [r2]
+       beq     LSYM(Lepilogue)
+
+       @ return INT64
+       cmp     r3, #FFI_TYPE_SINT64
+#if defined(__SOFTFP__) || defined(__ARM_EABI__)
+       cmpne   r3, #FFI_TYPE_DOUBLE
+#endif
+       stmeqia r2, {r0, r1}
+
+#if !defined(__SOFTFP__) && !defined(__ARM_EABI__)
+       beq     LSYM(Lepilogue)
+
+@ return FLOAT
+       cmp     r3, #FFI_TYPE_FLOAT
+       stfeqs  f0, [r2]
+       beq     LSYM(Lepilogue)
+
+@ return DOUBLE or LONGDOUBLE
+       cmp     r3, #FFI_TYPE_DOUBLE
+       stfeqd  f0, [r2]
+#endif
+
+LSYM(Lepilogue):
+#if defined (__INTERWORKING__)
+       ldmia   sp!, {r0-r3,fp, lr}
+       bx      lr
+#else
+       ldmia   sp!, {r0-r3,fp, pc}
+#endif
+
+.ffi_call_SYSV_end:
+       UNWIND .fnend
+#ifdef __ELF__
+        .size    CNAME(ffi_call_SYSV),.ffi_call_SYSV_end-CNAME(ffi_call_SYSV)
+#endif
+
+
+/*
+       unsigned int FFI_HIDDEN
+       ffi_closure_SYSV_inner (closure, respp, args)
+            ffi_closure *closure;
+            void **respp;
+            void *args;
+*/
+
+ARM_FUNC_START ffi_closure_SYSV
+       UNWIND .pad #16
+       add     ip, sp, #16
+       stmfd   sp!, {ip, lr}
+       UNWIND .save    {r0, lr}
+       add     r2, sp, #8
+       UNWIND .pad #16
+       sub     sp, sp, #16
+       str     sp, [sp, #8]
+       add     r1, sp, #8
+       bl      CNAME(ffi_closure_SYSV_inner)
+       cmp     r0, #FFI_TYPE_INT
+       beq     .Lretint
+
+       cmp     r0, #FFI_TYPE_FLOAT
+#if defined(__SOFTFP__) || defined(__ARM_EABI__)
+       beq     .Lretint
+#else
+       beq     .Lretfloat
+#endif
+
+       cmp     r0, #FFI_TYPE_DOUBLE
+#if defined(__SOFTFP__) || defined(__ARM_EABI__)
+       beq     .Lretlonglong
+#else
+       beq     .Lretdouble
+#endif
+
+       cmp     r0, #FFI_TYPE_LONGDOUBLE
+#if defined(__SOFTFP__) || defined(__ARM_EABI__)
+       beq     .Lretlonglong
+#else
+       beq     .Lretlongdouble
+#endif
+
+       cmp     r0, #FFI_TYPE_SINT64
+       beq     .Lretlonglong
+.Lclosure_epilogue:
+       add     sp, sp, #16
+       ldmfd   sp, {sp, pc}
+.Lretint:
+       ldr     r0, [sp]
+       b       .Lclosure_epilogue
+.Lretlonglong:
+       ldr     r0, [sp]
+       ldr     r1, [sp, #4]
+       b       .Lclosure_epilogue
+
+#if !defined(__SOFTFP__) && !defined(__ARM_EABI__)
+.Lretfloat:
+       ldfs    f0, [sp]
+       b       .Lclosure_epilogue
+.Lretdouble:
+       ldfd    f0, [sp]
+       b       .Lclosure_epilogue
+.Lretlongdouble:
+       ldfd    f0, [sp]
+       b       .Lclosure_epilogue
+#endif
+
+.ffi_closure_SYSV_end:
+       UNWIND .fnend
+#ifdef __ELF__
+        .size    CNAME(ffi_closure_SYSV),.ffi_closure_SYSV_end-CNAME(ffi_closure_SYSV)
+#endif
+
+
+/* Below are VFP hard-float ABI call and closure implementations.
+   Add VFP FPU directive here. This is only compiled into the library
+   under EABI.  */
+#ifdef __ARM_EABI__
+       .fpu    vfp
+
+       @ r0:   fn
+       @ r1:   &ecif
+       @ r2:   cif->bytes
+       @ r3:   fig->flags
+       @ sp+0: ecif.rvalue
+
+ARM_FUNC_START ffi_call_VFP
+       @ Save registers
+        stmfd  sp!, {r0-r3, fp, lr}
+       UNWIND .save    {r0-r3, fp, lr}
+       mov     fp, sp
+       UNWIND .setfp   fp, sp
+
+       @ Make room for all of the new args.
+       sub     sp, sp, r2
+
+       @ Make room for loading VFP args
+       sub     sp, sp, #64
+
+       @ Place all of the ffi_prep_args in position
+       mov     r0, sp
+       @     r1 already set
+       sub     r2, fp, #64   @ VFP scratch space
+
+       @ Call ffi_prep_args(stack, &ecif, vfp_space)
+       bl      ffi_prep_args
+
+       @ Load VFP register args if needed
+       cmp     r0, #0
+       beq     LSYM(Lbase_args)
+
+       @ Load only d0 if possible
+       cmp     r0, #3
+       sub     ip, fp, #64
+       flddle  d0, [ip]
+       fldmiadgt       ip, {d0-d7}
+
+LSYM(Lbase_args):
+       @ move first 4 parameters in registers
+       ldmia   sp, {r0-r3}
+
+       @ and adjust stack
+       sub     lr, ip, sp      @ cif->bytes == (fp - 64) - sp
+       ldr     ip, [fp]        @ load fn() in advance
+        cmp    lr, #16
+       movhs   lr, #16
+        add    sp, sp, lr
+
+       @ call (fn) (...)
+       call_reg(ip)
+
+       @ Remove the space we pushed for the args
+       mov     sp, fp
+
+       @ Load r2 with the pointer to storage for
+       @ the return value
+       ldr     r2, [sp, #24]
+
+       @ Load r3 with the return type code 
+       ldr     r3, [sp, #12]
+
+       @ If the return value pointer is NULL,
+       @ assume no return value.
+       cmp     r2, #0
+       beq     LSYM(Lepilogue_vfp)
+
+       cmp     r3, #FFI_TYPE_INT
+       streq   r0, [r2]
+       beq     LSYM(Lepilogue_vfp)
+
+       cmp     r3, #FFI_TYPE_SINT64
+       stmeqia r2, {r0, r1}
+       beq     LSYM(Lepilogue_vfp)
+
+       cmp     r3, #FFI_TYPE_FLOAT
+       fstseq  s0, [r2]
+       beq     LSYM(Lepilogue_vfp)
+       
+       cmp     r3, #FFI_TYPE_DOUBLE
+       fstdeq  d0, [r2]
+       beq     LSYM(Lepilogue_vfp)
+
+       cmp     r3, #FFI_TYPE_STRUCT_VFP_FLOAT
+       cmpne   r3, #FFI_TYPE_STRUCT_VFP_DOUBLE
+       fstmiadeq       r2, {d0-d3}
+
+LSYM(Lepilogue_vfp):
+       RETLDM  "r0-r3,fp"
+
+.ffi_call_VFP_end:
+       UNWIND .fnend
+        .size    CNAME(ffi_call_VFP),.ffi_call_VFP_end-CNAME(ffi_call_VFP)
+
+
+ARM_FUNC_START ffi_closure_VFP
+       fstmfdd sp!, {d0-d7}
+       @ r0-r3, then d0-d7
+       UNWIND .pad #80
+       add     ip, sp, #80
+       stmfd   sp!, {ip, lr}
+       UNWIND .save    {r0, lr}
+       add     r2, sp, #72
+       add     r3, sp, #8
+       UNWIND .pad #72
+       sub     sp, sp, #72
+       str     sp, [sp, #64]
+       add     r1, sp, #64
+       bl      ffi_closure_SYSV_inner
+
+       cmp     r0, #FFI_TYPE_INT
+       beq     .Lretint_vfp
+
+       cmp     r0, #FFI_TYPE_FLOAT
+       beq     .Lretfloat_vfp
+
+       cmp     r0, #FFI_TYPE_DOUBLE
+       cmpne   r0, #FFI_TYPE_LONGDOUBLE
+       beq     .Lretdouble_vfp
+
+       cmp     r0, #FFI_TYPE_SINT64
+       beq     .Lretlonglong_vfp
+
+       cmp     r0, #FFI_TYPE_STRUCT_VFP_FLOAT
+       beq     .Lretfloat_struct_vfp
+
+       cmp     r0, #FFI_TYPE_STRUCT_VFP_DOUBLE
+       beq     .Lretdouble_struct_vfp
+       
+.Lclosure_epilogue_vfp:
+       add     sp, sp, #72
+       ldmfd   sp, {sp, pc}
+
+.Lretfloat_vfp:
+       flds    s0, [sp]
+       b       .Lclosure_epilogue_vfp
+.Lretdouble_vfp:
+       fldd    d0, [sp]
+       b       .Lclosure_epilogue_vfp
+.Lretint_vfp:
+       ldr     r0, [sp]
+       b       .Lclosure_epilogue_vfp
+.Lretlonglong_vfp:
+       ldmia   sp, {r0, r1}
+       b       .Lclosure_epilogue_vfp
+.Lretfloat_struct_vfp:
+       fldmiad sp, {d0-d1}
+       b       .Lclosure_epilogue_vfp
+.Lretdouble_struct_vfp:
+       fldmiad sp, {d0-d3}
+       b       .Lclosure_epilogue_vfp
+
+.ffi_closure_VFP_end:
+       UNWIND .fnend
+        .size    CNAME(ffi_closure_VFP),.ffi_closure_VFP_end-CNAME(ffi_closure_VFP)
+#endif
+
+ENTRY(ffi_arm_trampoline)
+       stmfd sp!, {r0-r3}
+       ldr r0, [pc]
+       ldr pc, [pc]
+
+#if defined __ELF__ && defined __linux__
+       .section        .note.GNU-stack,"",%progbits
+#endif
diff --git a/.pc/apple-fixes/src/x86/ffi.c b/.pc/apple-fixes/src/x86/ffi.c
new file mode 100644 (file)
index 0000000..22e2997
--- /dev/null
@@ -0,0 +1,841 @@
+/* -----------------------------------------------------------------------
+   ffi.c - Copyright (c) 1996, 1998, 1999, 2001, 2007, 2008  Red Hat, Inc.
+           Copyright (c) 2002  Ranjit Mathew
+           Copyright (c) 2002  Bo Thorsen
+           Copyright (c) 2002  Roger Sayle
+           Copyright (C) 2008, 2010  Free Software Foundation, Inc.
+
+   x86 Foreign Function Interface
+
+   Permission is hereby granted, free of charge, to any person obtaining
+   a copy of this software and associated documentation files (the
+   ``Software''), to deal in the Software without restriction, including
+   without limitation the rights to use, copy, modify, merge, publish,
+   distribute, sublicense, and/or sell copies of the Software, and to
+   permit persons to whom the Software is furnished to do so, subject to
+   the following conditions:
+
+   The above copyright notice and this permission notice shall be included
+   in all copies or substantial portions of the Software.
+
+   THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND,
+   EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+   NONINFRINGEMENT.  IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+   HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+   WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+   OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+   DEALINGS IN THE SOFTWARE.
+   ----------------------------------------------------------------------- */
+
+#if !defined(__x86_64__) || defined(_WIN64)
+
+#ifdef _WIN64
+#include <windows.h>
+#endif
+
+#include <ffi.h>
+#include <ffi_common.h>
+
+#include <stdlib.h>
+
+/* ffi_prep_args is called by the assembly routine once stack space
+   has been allocated for the function's arguments */
+
+void ffi_prep_args(char *stack, extended_cif *ecif)
+{
+  register unsigned int i;
+  register void **p_argv;
+  register char *argp;
+  register ffi_type **p_arg;
+#ifdef X86_WIN32
+  size_t p_stack_args[2];
+  void *p_stack_data[2];
+  char *argp2 = stack;
+  int stack_args_count = 0;
+  int cabi = ecif->cif->abi;
+#endif
+
+  argp = stack;
+
+  if ((ecif->cif->flags == FFI_TYPE_STRUCT
+       || ecif->cif->flags == FFI_TYPE_MS_STRUCT)
+#ifdef X86_WIN64
+      && (ecif->cif->rtype->size != 1 && ecif->cif->rtype->size != 2
+          && ecif->cif->rtype->size != 4 && ecif->cif->rtype->size != 8)
+#endif
+      )
+    {
+      *(void **) argp = ecif->rvalue;
+#ifdef X86_WIN32
+      /* For fastcall/thiscall this is first register-passed
+         argument.  */
+      if (cabi == FFI_THISCALL || cabi == FFI_FASTCALL)
+       {
+         p_stack_args[stack_args_count] = sizeof (void*);
+         p_stack_data[stack_args_count] = argp;
+         ++stack_args_count;
+       }
+#endif
+      argp += sizeof(void*);
+    }
+
+  p_argv = ecif->avalue;
+
+  for (i = ecif->cif->nargs, p_arg = ecif->cif->arg_types;
+       i != 0;
+       i--, p_arg++)
+    {
+      size_t z;
+
+      /* Align if necessary */
+      if ((sizeof(void*) - 1) & (size_t) argp)
+        argp = (char *) ALIGN(argp, sizeof(void*));
+
+      z = (*p_arg)->size;
+#ifdef X86_WIN64
+      if (z > sizeof(ffi_arg)
+          || ((*p_arg)->type == FFI_TYPE_STRUCT
+              && (z != 1 && z != 2 && z != 4 && z != 8))
+#if FFI_TYPE_DOUBLE != FFI_TYPE_LONGDOUBLE
+          || ((*p_arg)->type == FFI_TYPE_LONGDOUBLE)
+#endif
+          )
+        {
+          z = sizeof(ffi_arg);
+          *(void **)argp = *p_argv;
+        }
+      else if ((*p_arg)->type == FFI_TYPE_FLOAT)
+        {
+          memcpy(argp, *p_argv, z);
+        }
+      else
+#endif
+      if (z < sizeof(ffi_arg))
+        {
+          z = sizeof(ffi_arg);
+          switch ((*p_arg)->type)
+            {
+            case FFI_TYPE_SINT8:
+              *(ffi_sarg *) argp = (ffi_sarg)*(SINT8 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_UINT8:
+              *(ffi_arg *) argp = (ffi_arg)*(UINT8 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_SINT16:
+              *(ffi_sarg *) argp = (ffi_sarg)*(SINT16 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_UINT16:
+              *(ffi_arg *) argp = (ffi_arg)*(UINT16 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_SINT32:
+              *(ffi_sarg *) argp = (ffi_sarg)*(SINT32 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_UINT32:
+              *(ffi_arg *) argp = (ffi_arg)*(UINT32 *)(* p_argv);
+              break;
+
+            case FFI_TYPE_STRUCT:
+              *(ffi_arg *) argp = *(ffi_arg *)(* p_argv);
+              break;
+
+            default:
+              FFI_ASSERT(0);
+            }
+        }
+      else
+        {
+          memcpy(argp, *p_argv, z);
+        }
+
+#ifdef X86_WIN32
+    /* For thiscall/fastcall convention register-passed arguments
+       are the first two none-floating-point arguments with a size
+       smaller or equal to sizeof (void*).  */
+    if ((cabi == FFI_THISCALL && stack_args_count < 1)
+        || (cabi == FFI_FASTCALL && stack_args_count < 2))
+      {
+       if (z <= 4
+           && ((*p_arg)->type != FFI_TYPE_FLOAT
+               && (*p_arg)->type != FFI_TYPE_STRUCT))
+         {
+           p_stack_args[stack_args_count] = z;
+           p_stack_data[stack_args_count] = argp;
+           ++stack_args_count;
+         }
+      }
+#endif
+      p_argv++;
+#ifdef X86_WIN64
+      argp += (z + sizeof(void*) - 1) & ~(sizeof(void*) - 1);
+#else
+      argp += z;
+#endif
+    }
+
+#ifdef X86_WIN32
+  /* We need to move the register-passed arguments for thiscall/fastcall
+     on top of stack, so that those can be moved to registers ecx/edx by
+     call-handler.  */
+  if (stack_args_count > 0)
+    {
+      size_t zz = (p_stack_args[0] + 3) & ~3;
+      char *h;
+
+      /* Move first argument to top-stack position.  */
+      if (p_stack_data[0] != argp2)
+       {
+         h = alloca (zz + 1);
+         memcpy (h, p_stack_data[0], zz);
+         memmove (argp2 + zz, argp2,
+                  (size_t) ((char *) p_stack_data[0] - (char*)argp2));
+         memcpy (argp2, h, zz);
+       }
+
+      argp2 += zz;
+      --stack_args_count;
+      if (zz > 4)
+       stack_args_count = 0;
+
+      /* If we have a second argument, then move it on top
+         after the first one.  */
+      if (stack_args_count > 0 && p_stack_data[1] != argp2)
+       {
+         zz = p_stack_args[1];
+         zz = (zz + 3) & ~3;
+         h = alloca (zz + 1);
+         h = alloca (zz + 1);
+         memcpy (h, p_stack_data[1], zz);
+         memmove (argp2 + zz, argp2, (size_t) ((char*) p_stack_data[1] - (char*)argp2));
+         memcpy (argp2, h, zz);
+       }
+    }
+#endif
+  return;
+}
+
+/* Perform machine dependent cif processing */
+ffi_status ffi_prep_cif_machdep(ffi_cif *cif)
+{
+  unsigned int i;
+  ffi_type **ptr;
+
+  /* Set the return type flag */
+  switch (cif->rtype->type)
+    {
+    case FFI_TYPE_VOID:
+    case FFI_TYPE_UINT8:
+    case FFI_TYPE_UINT16:
+    case FFI_TYPE_SINT8:
+    case FFI_TYPE_SINT16:
+#ifdef X86_WIN64
+    case FFI_TYPE_UINT32:
+    case FFI_TYPE_SINT32:
+#endif
+    case FFI_TYPE_SINT64:
+    case FFI_TYPE_FLOAT:
+    case FFI_TYPE_DOUBLE:
+#ifndef X86_WIN64
+#if FFI_TYPE_DOUBLE != FFI_TYPE_LONGDOUBLE
+    case FFI_TYPE_LONGDOUBLE:
+#endif
+#endif
+      cif->flags = (unsigned) cif->rtype->type;
+      break;
+
+    case FFI_TYPE_UINT64:
+#ifdef X86_WIN64
+    case FFI_TYPE_POINTER:
+#endif
+      cif->flags = FFI_TYPE_SINT64;
+      break;
+
+    case FFI_TYPE_STRUCT:
+#ifndef X86
+      if (cif->rtype->size == 1)
+        {
+          cif->flags = FFI_TYPE_SMALL_STRUCT_1B; /* same as char size */
+        }
+      else if (cif->rtype->size == 2)
+        {
+          cif->flags = FFI_TYPE_SMALL_STRUCT_2B; /* same as short size */
+        }
+      else if (cif->rtype->size == 4)
+        {
+#ifdef X86_WIN64
+          cif->flags = FFI_TYPE_SMALL_STRUCT_4B;
+#else
+          cif->flags = FFI_TYPE_INT; /* same as int type */
+#endif
+        }
+      else if (cif->rtype->size == 8)
+        {
+          cif->flags = FFI_TYPE_SINT64; /* same as int64 type */
+        }
+      else
+#endif
+        {
+          if (cif->abi == FFI_MS_CDECL)
+            cif->flags = FFI_TYPE_MS_STRUCT;
+          else
+            cif->flags = FFI_TYPE_STRUCT;
+          /* allocate space for return value pointer */
+          cif->bytes += ALIGN(sizeof(void*), FFI_SIZEOF_ARG);
+        }
+      break;
+
+    default:
+#ifdef X86_WIN64
+      cif->flags = FFI_TYPE_SINT64;
+      break;
+    case FFI_TYPE_INT:
+      cif->flags = FFI_TYPE_SINT32;
+#else
+      cif->flags = FFI_TYPE_INT;
+#endif
+      break;
+    }
+
+  for (ptr = cif->arg_types, i = cif->nargs; i > 0; i--, ptr++)
+    {
+      if (((*ptr)->alignment - 1) & cif->bytes)
+        cif->bytes = ALIGN(cif->bytes, (*ptr)->alignment);
+      cif->bytes += ALIGN((*ptr)->size, FFI_SIZEOF_ARG);
+    }
+
+#ifdef X86_WIN64
+  /* ensure space for storing four registers */
+  cif->bytes += 4 * sizeof(ffi_arg);
+#endif
+
+#ifdef X86_DARWIN
+  cif->bytes = (cif->bytes + 15) & ~0xF;
+#endif
+
+  return FFI_OK;
+}
+
+#ifdef X86_WIN64
+extern int
+ffi_call_win64(void (*)(char *, extended_cif *), extended_cif *,
+               unsigned, unsigned, unsigned *, void (*fn)(void));
+#elif defined(X86_WIN32)
+extern void
+ffi_call_win32(void (*)(char *, extended_cif *), extended_cif *,
+               unsigned, unsigned, unsigned, unsigned *, void (*fn)(void));
+#else
+extern void ffi_call_SYSV(void (*)(char *, extended_cif *), extended_cif *,
+                          unsigned, unsigned, unsigned *, void (*fn)(void));
+#endif
+
+void ffi_call(ffi_cif *cif, void (*fn)(void), void *rvalue, void **avalue)
+{
+  extended_cif ecif;
+
+  ecif.cif = cif;
+  ecif.avalue = avalue;
+  
+  /* If the return value is a struct and we don't have a return */
+  /* value address then we need to make one                     */
+
+#ifdef X86_WIN64
+  if (rvalue == NULL
+      && cif->flags == FFI_TYPE_STRUCT
+      && cif->rtype->size != 1 && cif->rtype->size != 2
+      && cif->rtype->size != 4 && cif->rtype->size != 8)
+    {
+      ecif.rvalue = alloca((cif->rtype->size + 0xF) & ~0xF);
+    }
+#else
+  if (rvalue == NULL
+      && (cif->flags == FFI_TYPE_STRUCT
+          || cif->flags == FFI_TYPE_MS_STRUCT))
+    {
+      ecif.rvalue = alloca(cif->rtype->size);
+    }
+#endif
+  else
+    ecif.rvalue = rvalue;
+    
+  
+  switch (cif->abi) 
+    {
+#ifdef X86_WIN64
+    case FFI_WIN64:
+      ffi_call_win64(ffi_prep_args, &ecif, cif->bytes,
+                     cif->flags, ecif.rvalue, fn);
+      break;
+#elif defined(X86_WIN32)
+    case FFI_SYSV:
+    case FFI_STDCALL:
+    case FFI_MS_CDECL:
+      ffi_call_win32(ffi_prep_args, &ecif, cif->abi, cif->bytes, cif->flags,
+                    ecif.rvalue, fn);
+      break;
+    case FFI_THISCALL:
+    case FFI_FASTCALL:
+      {
+       unsigned int abi = cif->abi;
+       unsigned int i, passed_regs = 0;
+
+       if (cif->flags == FFI_TYPE_STRUCT)
+         ++passed_regs;
+
+       for (i=0; i < cif->nargs && passed_regs < 2;i++)
+         {
+           size_t sz;
+
+           if (cif->arg_types[i]->type == FFI_TYPE_FLOAT
+               || cif->arg_types[i]->type == FFI_TYPE_STRUCT)
+             continue;
+           sz = (cif->arg_types[i]->size + 3) & ~3;
+           if (sz == 0 || sz > 4)
+             continue;
+           ++passed_regs;
+         }
+       if (passed_regs < 2 && abi == FFI_FASTCALL)
+         abi = FFI_THISCALL;
+       if (passed_regs < 1 && abi == FFI_THISCALL)
+         abi = FFI_STDCALL;
+        ffi_call_win32(ffi_prep_args, &ecif, abi, cif->bytes, cif->flags,
+                       ecif.rvalue, fn);
+      }
+      break;
+#else
+    case FFI_SYSV:
+      ffi_call_SYSV(ffi_prep_args, &ecif, cif->bytes, cif->flags, ecif.rvalue,
+                    fn);
+      break;
+#endif
+    default:
+      FFI_ASSERT(0);
+      break;
+    }
+}
+
+
+/** private members **/
+
+/* The following __attribute__((regparm(1))) decorations will have no effect
+   on MSVC - standard cdecl convention applies. */
+static void ffi_prep_incoming_args_SYSV (char *stack, void **ret,
+                                         void** args, ffi_cif* cif);
+void FFI_HIDDEN ffi_closure_SYSV (ffi_closure *)
+     __attribute__ ((regparm(1)));
+unsigned int FFI_HIDDEN ffi_closure_SYSV_inner (ffi_closure *, void **, void *)
+     __attribute__ ((regparm(1)));
+void FFI_HIDDEN ffi_closure_raw_SYSV (ffi_raw_closure *)
+     __attribute__ ((regparm(1)));
+#ifdef X86_WIN32
+void FFI_HIDDEN ffi_closure_raw_THISCALL (ffi_raw_closure *)
+     __attribute__ ((regparm(1)));
+void FFI_HIDDEN ffi_closure_STDCALL (ffi_closure *)
+     __attribute__ ((regparm(1)));
+void FFI_HIDDEN ffi_closure_THISCALL (ffi_closure *)
+     __attribute__ ((regparm(1)));
+#endif
+#ifdef X86_WIN64
+void FFI_HIDDEN ffi_closure_win64 (ffi_closure *);
+#endif
+
+/* This function is jumped to by the trampoline */
+
+#ifdef X86_WIN64
+void * FFI_HIDDEN
+ffi_closure_win64_inner (ffi_closure *closure, void *args) {
+  ffi_cif       *cif;
+  void         **arg_area;
+  void          *result;
+  void          *resp = &result;
+
+  cif         = closure->cif;
+  arg_area    = (void**) alloca (cif->nargs * sizeof (void*));  
+
+  /* this call will initialize ARG_AREA, such that each
+   * element in that array points to the corresponding 
+   * value on the stack; and if the function returns
+   * a structure, it will change RESP to point to the
+   * structure return address.  */
+
+  ffi_prep_incoming_args_SYSV(args, &resp, arg_area, cif);
+  
+  (closure->fun) (cif, resp, arg_area, closure->user_data);
+
+  /* The result is returned in rax.  This does the right thing for
+     result types except for floats; we have to 'mov xmm0, rax' in the
+     caller to correct this.
+     TODO: structure sizes of 3 5 6 7 are returned by reference, too!!!
+  */
+  return cif->rtype->size > sizeof(void *) ? resp : *(void **)resp;
+}
+
+#else
+unsigned int FFI_HIDDEN __attribute__ ((regparm(1)))
+ffi_closure_SYSV_inner (ffi_closure *closure, void **respp, void *args)
+{
+  /* our various things...  */
+  ffi_cif       *cif;
+  void         **arg_area;
+
+  cif         = closure->cif;
+  arg_area    = (void**) alloca (cif->nargs * sizeof (void*));  
+
+  /* this call will initialize ARG_AREA, such that each
+   * element in that array points to the corresponding 
+   * value on the stack; and if the function returns
+   * a structure, it will change RESP to point to the
+   * structure return address.  */
+
+  ffi_prep_incoming_args_SYSV(args, respp, arg_area, cif);
+
+  (closure->fun) (cif, *respp, arg_area, closure->user_data);
+
+  return cif->flags;
+}
+#endif /* !X86_WIN64 */
+
+static void
+ffi_prep_incoming_args_SYSV(char *stack, void **rvalue, void **avalue,
+                            ffi_cif *cif)
+{
+  register unsigned int i;
+  register void **p_argv;
+  register char *argp;
+  register ffi_type **p_arg;
+
+  argp = stack;
+
+#ifdef X86_WIN64
+  if (cif->rtype->size > sizeof(ffi_arg)
+      || (cif->flags == FFI_TYPE_STRUCT
+          && (cif->rtype->size != 1 && cif->rtype->size != 2
+              && cif->rtype->size != 4 && cif->rtype->size != 8))) {
+    *rvalue = *(void **) argp;
+    argp += sizeof(void *);
+  }
+#else
+  if ( cif->flags == FFI_TYPE_STRUCT
+       || cif->flags == FFI_TYPE_MS_STRUCT ) {
+    *rvalue = *(void **) argp;
+    argp += sizeof(void *);
+  }
+#endif
+
+  p_argv = avalue;
+
+  for (i = cif->nargs, p_arg = cif->arg_types; (i != 0); i--, p_arg++)
+    {
+      size_t z;
+
+      /* Align if necessary */
+      if ((sizeof(void*) - 1) & (size_t) argp) {
+        argp = (char *) ALIGN(argp, sizeof(void*));
+      }
+
+#ifdef X86_WIN64
+      if ((*p_arg)->size > sizeof(ffi_arg)
+          || ((*p_arg)->type == FFI_TYPE_STRUCT
+              && ((*p_arg)->size != 1 && (*p_arg)->size != 2
+                  && (*p_arg)->size != 4 && (*p_arg)->size != 8)))
+        {
+          z = sizeof(void *);
+          *p_argv = *(void **)argp;
+        }
+      else
+#endif
+        {
+          z = (*p_arg)->size;
+          
+          /* because we're little endian, this is what it turns into.   */
+          
+          *p_argv = (void*) argp;
+        }
+          
+      p_argv++;
+#ifdef X86_WIN64
+      argp += (z + sizeof(void*) - 1) & ~(sizeof(void*) - 1);
+#else
+      argp += z;
+#endif
+    }
+  
+  return;
+}
+
+#define FFI_INIT_TRAMPOLINE_WIN64(TRAMP,FUN,CTX,MASK) \
+{ unsigned char *__tramp = (unsigned char*)(TRAMP); \
+   void*  __fun = (void*)(FUN); \
+   void*  __ctx = (void*)(CTX); \
+   *(unsigned char*) &__tramp[0] = 0x41; \
+   *(unsigned char*) &__tramp[1] = 0xbb; \
+   *(unsigned int*) &__tramp[2] = MASK; /* mov $mask, %r11 */ \
+   *(unsigned char*) &__tramp[6] = 0x48; \
+   *(unsigned char*) &__tramp[7] = 0xb8; \
+   *(void**) &__tramp[8] = __ctx; /* mov __ctx, %rax */ \
+   *(unsigned char *)  &__tramp[16] = 0x49; \
+   *(unsigned char *)  &__tramp[17] = 0xba; \
+   *(void**) &__tramp[18] = __fun; /* mov __fun, %r10 */ \
+   *(unsigned char *)  &__tramp[26] = 0x41; \
+   *(unsigned char *)  &__tramp[27] = 0xff; \
+   *(unsigned char *)  &__tramp[28] = 0xe2; /* jmp %r10 */ \
+ }
+
+/* How to make a trampoline.  Derived from gcc/config/i386/i386.c. */
+
+#define FFI_INIT_TRAMPOLINE(TRAMP,FUN,CTX) \
+{ unsigned char *__tramp = (unsigned char*)(TRAMP); \
+   unsigned int  __fun = (unsigned int)(FUN); \
+   unsigned int  __ctx = (unsigned int)(CTX); \
+   unsigned int  __dis = __fun - (__ctx + 10);  \
+   *(unsigned char*) &__tramp[0] = 0xb8; \
+   *(unsigned int*)  &__tramp[1] = __ctx; /* movl __ctx, %eax */ \
+   *(unsigned char *)  &__tramp[5] = 0xe9; \
+   *(unsigned int*)  &__tramp[6] = __dis; /* jmp __fun  */ \
+ }
+
+#define FFI_INIT_TRAMPOLINE_THISCALL(TRAMP,FUN,CTX,SIZE) \
+{ unsigned char *__tramp = (unsigned char*)(TRAMP); \
+   unsigned int  __fun = (unsigned int)(FUN); \
+   unsigned int  __ctx = (unsigned int)(CTX); \
+   unsigned int  __dis = __fun - (__ctx + 49);  \
+   unsigned short __size = (unsigned short)(SIZE); \
+   *(unsigned int *) &__tramp[0] = 0x8324048b; /* mov (%esp), %eax */ \
+   *(unsigned int *) &__tramp[4] = 0x4c890cec; /* sub $12, %esp */ \
+   *(unsigned int *) &__tramp[8] = 0x04890424; /* mov %ecx, 4(%esp) */ \
+   *(unsigned char*) &__tramp[12] = 0x24;      /* mov %eax, (%esp) */ \
+   *(unsigned char*) &__tramp[13] = 0xb8; \
+   *(unsigned int *) &__tramp[14] = __size;    /* mov __size, %eax */ \
+   *(unsigned int *) &__tramp[18] = 0x08244c8d;        /* lea 8(%esp), %ecx */ \
+   *(unsigned int *) &__tramp[22] = 0x4802e8c1; /* shr $2, %eax ; dec %eax */ \
+   *(unsigned short*) &__tramp[26] = 0x0b74;   /* jz 1f */ \
+   *(unsigned int *) &__tramp[28] = 0x8908518b;        /* 2b: mov 8(%ecx), %edx */ \
+   *(unsigned int *) &__tramp[32] = 0x04c18311; /* mov %edx, (%ecx) ; add $4, %ecx */ \
+   *(unsigned char*) &__tramp[36] = 0x48;      /* dec %eax */ \
+   *(unsigned short*) &__tramp[37] = 0xf575;   /* jnz 2b ; 1f: */ \
+   *(unsigned char*) &__tramp[39] = 0xb8; \
+   *(unsigned int*)  &__tramp[40] = __ctx; /* movl __ctx, %eax */ \
+   *(unsigned char *)  &__tramp[44] = 0xe8; \
+   *(unsigned int*)  &__tramp[45] = __dis; /* call __fun  */ \
+   *(unsigned char*)  &__tramp[49] = 0xc2; /* ret  */ \
+   *(unsigned short*)  &__tramp[50] = (__size + 8); /* ret (__size + 8)  */ \
+ }
+
+#define FFI_INIT_TRAMPOLINE_STDCALL(TRAMP,FUN,CTX,SIZE)  \
+{ unsigned char *__tramp = (unsigned char*)(TRAMP); \
+   unsigned int  __fun = (unsigned int)(FUN); \
+   unsigned int  __ctx = (unsigned int)(CTX); \
+   unsigned int  __dis = __fun - (__ctx + 10); \
+   unsigned short __size = (unsigned short)(SIZE); \
+   *(unsigned char*) &__tramp[0] = 0xb8; \
+   *(unsigned int*)  &__tramp[1] = __ctx; /* movl __ctx, %eax */ \
+   *(unsigned char *)  &__tramp[5] = 0xe8; \
+   *(unsigned int*)  &__tramp[6] = __dis; /* call __fun  */ \
+   *(unsigned char *)  &__tramp[10] = 0xc2; \
+   *(unsigned short*)  &__tramp[11] = __size; /* ret __size  */ \
+ }
+
+/* the cif must already be prep'ed */
+
+ffi_status
+ffi_prep_closure_loc (ffi_closure* closure,
+                      ffi_cif* cif,
+                      void (*fun)(ffi_cif*,void*,void**,void*),
+                      void *user_data,
+                      void *codeloc)
+{
+#ifdef X86_WIN64
+#define ISFLOAT(IDX) (cif->arg_types[IDX]->type == FFI_TYPE_FLOAT || cif->arg_types[IDX]->type == FFI_TYPE_DOUBLE)
+#define FLAG(IDX) (cif->nargs>(IDX)&&ISFLOAT(IDX)?(1<<(IDX)):0)
+  if (cif->abi == FFI_WIN64) 
+    {
+      int mask = FLAG(0)|FLAG(1)|FLAG(2)|FLAG(3);
+      FFI_INIT_TRAMPOLINE_WIN64 (&closure->tramp[0],
+                                 &ffi_closure_win64,
+                                 codeloc, mask);
+      /* make sure we can execute here */
+    }
+#else
+  if (cif->abi == FFI_SYSV)
+    {
+      FFI_INIT_TRAMPOLINE (&closure->tramp[0],
+                           &ffi_closure_SYSV,
+                           (void*)codeloc);
+    }
+#ifdef X86_WIN32
+  else if (cif->abi == FFI_THISCALL)
+    {
+      FFI_INIT_TRAMPOLINE_THISCALL (&closure->tramp[0],
+                                   &ffi_closure_THISCALL,
+                                   (void*)codeloc,
+                                   cif->bytes);
+    }
+  else if (cif->abi == FFI_STDCALL)
+    {
+      FFI_INIT_TRAMPOLINE_STDCALL (&closure->tramp[0],
+                                   &ffi_closure_STDCALL,
+                                   (void*)codeloc, cif->bytes);
+    }
+  else if (cif->abi == FFI_MS_CDECL)
+    {
+      FFI_INIT_TRAMPOLINE (&closure->tramp[0],
+                           &ffi_closure_SYSV,
+                           (void*)codeloc);
+    }
+#endif /* X86_WIN32 */
+#endif /* !X86_WIN64 */
+  else
+    {
+      return FFI_BAD_ABI;
+    }
+    
+  closure->cif  = cif;
+  closure->user_data = user_data;
+  closure->fun  = fun;
+
+  return FFI_OK;
+}
+
+/* ------- Native raw API support -------------------------------- */
+
+#if !FFI_NO_RAW_API
+
+ffi_status
+ffi_prep_raw_closure_loc (ffi_raw_closure* closure,
+                          ffi_cif* cif,
+                          void (*fun)(ffi_cif*,void*,ffi_raw*,void*),
+                          void *user_data,
+                          void *codeloc)
+{
+  int i;
+
+  if (cif->abi != FFI_SYSV) {
+#ifdef X86_WIN32
+    if (cif->abi != FFI_THISCALL)
+#endif
+    return FFI_BAD_ABI;
+  }
+
+  /* we currently don't support certain kinds of arguments for raw
+     closures.  This should be implemented by a separate assembly
+     language routine, since it would require argument processing,
+     something we don't do now for performance.  */
+
+  for (i = cif->nargs-1; i >= 0; i--)
+    {
+      FFI_ASSERT (cif->arg_types[i]->type != FFI_TYPE_STRUCT);
+      FFI_ASSERT (cif->arg_types[i]->type != FFI_TYPE_LONGDOUBLE);
+    }
+  
+#ifdef X86_WIN32
+  if (cif->abi == FFI_SYSV)
+    {
+#endif
+  FFI_INIT_TRAMPOLINE (&closure->tramp[0], &ffi_closure_raw_SYSV,
+                       codeloc);
+#ifdef X86_WIN32
+    }
+  else if (cif->abi == FFI_THISCALL)
+    {
+      FFI_INIT_TRAMPOLINE_THISCALL (&closure->tramp[0], &ffi_closure_raw_THISCALL,
+                                   codeloc, cif->bytes);
+    }
+#endif
+  closure->cif  = cif;
+  closure->user_data = user_data;
+  closure->fun  = fun;
+
+  return FFI_OK;
+}
+
+static void 
+ffi_prep_args_raw(char *stack, extended_cif *ecif)
+{
+  memcpy (stack, ecif->avalue, ecif->cif->bytes);
+}
+
+/* we borrow this routine from libffi (it must be changed, though, to
+ * actually call the function passed in the first argument.  as of
+ * libffi-1.20, this is not the case.)
+ */
+
+void
+ffi_raw_call(ffi_cif *cif, void (*fn)(void), void *rvalue, ffi_raw *fake_avalue)
+{
+  extended_cif ecif;
+  void **avalue = (void **)fake_avalue;
+
+  ecif.cif = cif;
+  ecif.avalue = avalue;
+  
+  /* If the return value is a struct and we don't have a return */
+  /* value address then we need to make one                     */
+
+  if (rvalue == NULL
+      && (cif->flags == FFI_TYPE_STRUCT
+          || cif->flags == FFI_TYPE_MS_STRUCT))
+    {
+      ecif.rvalue = alloca(cif->rtype->size);
+    }
+  else
+    ecif.rvalue = rvalue;
+    
+  
+  switch (cif->abi) 
+    {
+#ifdef X86_WIN32
+    case FFI_SYSV:
+    case FFI_STDCALL:
+    case FFI_MS_CDECL:
+      ffi_call_win32(ffi_prep_args_raw, &ecif, cif->abi, cif->bytes, cif->flags,
+                    ecif.rvalue, fn);
+      break;
+    case FFI_THISCALL:
+    case FFI_FASTCALL:
+      {
+       unsigned int abi = cif->abi;
+       unsigned int i, passed_regs = 0;
+
+       if (cif->flags == FFI_TYPE_STRUCT)
+         ++passed_regs;
+
+       for (i=0; i < cif->nargs && passed_regs < 2;i++)
+         {
+           size_t sz;
+
+           if (cif->arg_types[i]->type == FFI_TYPE_FLOAT
+               || cif->arg_types[i]->type == FFI_TYPE_STRUCT)
+             continue;
+           sz = (cif->arg_types[i]->size + 3) & ~3;
+           if (sz == 0 || sz > 4)
+             continue;
+           ++passed_regs;
+         }
+       if (passed_regs < 2 && abi == FFI_FASTCALL)
+         cif->abi = abi = FFI_THISCALL;
+       if (passed_regs < 1 && abi == FFI_THISCALL)
+         cif->abi = abi = FFI_STDCALL;
+        ffi_call_win32(ffi_prep_args_raw, &ecif, abi, cif->bytes, cif->flags,
+                       ecif.rvalue, fn);
+      }
+      break;
+#else
+    case FFI_SYSV:
+      ffi_call_SYSV(ffi_prep_args_raw, &ecif, cif->bytes, cif->flags,
+                    ecif.rvalue, fn);
+      break;
+#endif
+    default:
+      FFI_ASSERT(0);
+      break;
+    }
+}
+
+#endif
+
+#endif /* !__x86_64__  || X86_WIN64 */
+
index 0608e26..5ccbf37 100644 (file)
@@ -7,3 +7,4 @@ vararg_float_test_fix
 vfp-eabi
 msvc-changes
 win32-return-sign
+apple-fixes
index e1892aa..afd5ae1 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,10 @@
+2012-04-05  Zachary Waldowski  <zwaldowski@gmail.com>
+
+       * include/ffi.h.in: Add missing trampoline table fields.
+       * src/arm/sysv.S: Fix ENTRY definition, and wrap symbol references
+       in CNAME.
+       * src/x86/ffi.c: Wrap Windows specific code in ifdefs.
+
 2012-04-02  Peter Rosin  <peda@lysator.liu.se>
 
        * src/x86/win32.S (ffi_call_win32): Sign/zero extend the return
index 4f3b3de..a51583b 100644 (file)
@@ -297,7 +297,12 @@ size_t ffi_java_raw_size (ffi_cif *cif);
 __declspec(align(8))
 #endif
 typedef struct {
+#if @FFI_EXEC_TRAMPOLINE_TABLE@
+  void *trampoline_table;
+  void *trampoline_table_entry;
+#else
   char tramp[FFI_TRAMPOLINE_SIZE];
+#endif
   ffi_cif   *cif;
   void     (*fun)(ffi_cif*,void*,void**,void*);
   void      *user_data;
@@ -330,8 +335,12 @@ ffi_prep_closure_loc (ffi_closure*,
 # pragma pack 8
 #endif
 typedef struct {
+#if @FFI_EXEC_TRAMPOLINE_TABLE@
+  void *trampoline_table;
+  void *trampoline_table_entry;
+#else
   char tramp[FFI_TRAMPOLINE_SIZE];
-
+#endif
   ffi_cif   *cif;
 
 #if !FFI_NATIVE_RAW_API
@@ -351,7 +360,12 @@ typedef struct {
 } ffi_raw_closure;
 
 typedef struct {
+#if @FFI_EXEC_TRAMPOLINE_TABLE@
+  void *trampoline_table;
+  void *trampoline_table_entry;
+#else
   char tramp[FFI_TRAMPOLINE_SIZE];
+#endif
 
   ffi_cif   *cif;
 
diff --git a/patches/apple-fixes b/patches/apple-fixes
new file mode 100644 (file)
index 0000000..b05fcdf
--- /dev/null
@@ -0,0 +1,115 @@
+Index: libffi/ChangeLog
+===================================================================
+--- libffi.orig/ChangeLog
++++ libffi/ChangeLog
+@@ -1,3 +1,10 @@
++2012-04-05  Zachary Waldowski  <zwaldowski@gmail.com>
++
++      * include/ffi.h.in: Add missing trampoline table fields.
++      * src/arm/sysv.S: Fix ENTRY definition, and wrap symbol references
++      in CNAME.
++      * src/x86/ffi.c: Wrap Windows specific code in ifdefs.
++
+ 2012-04-02  Peter Rosin  <peda@lysator.liu.se>
+       * src/x86/win32.S (ffi_call_win32): Sign/zero extend the return
+Index: libffi/include/ffi.h.in
+===================================================================
+--- libffi.orig/include/ffi.h.in
++++ libffi/include/ffi.h.in
+@@ -297,7 +297,12 @@ size_t ffi_java_raw_size (ffi_cif *cif);
+ __declspec(align(8))
+ #endif
+ typedef struct {
++#if @FFI_EXEC_TRAMPOLINE_TABLE@
++  void *trampoline_table;
++  void *trampoline_table_entry;
++#else
+   char tramp[FFI_TRAMPOLINE_SIZE];
++#endif
+   ffi_cif   *cif;
+   void     (*fun)(ffi_cif*,void*,void**,void*);
+   void      *user_data;
+@@ -330,8 +335,12 @@ ffi_prep_closure_loc (ffi_closure*,
+ # pragma pack 8
+ #endif
+ typedef struct {
++#if @FFI_EXEC_TRAMPOLINE_TABLE@
++  void *trampoline_table;
++  void *trampoline_table_entry;
++#else
+   char tramp[FFI_TRAMPOLINE_SIZE];
+-
++#endif
+   ffi_cif   *cif;
+ #if !FFI_NATIVE_RAW_API
+@@ -351,7 +360,12 @@ typedef struct {
+ } ffi_raw_closure;
+ typedef struct {
++#if @FFI_EXEC_TRAMPOLINE_TABLE@
++  void *trampoline_table;
++  void *trampoline_table_entry;
++#else
+   char tramp[FFI_TRAMPOLINE_SIZE];
++#endif
+   ffi_cif   *cif;
+Index: libffi/src/arm/sysv.S
+===================================================================
+--- libffi.orig/src/arm/sysv.S
++++ libffi/src/arm/sysv.S
+@@ -41,7 +41,7 @@
+ #define CNAME(x) x
+ #endif
+ #ifdef __APPLE__
+-#define ENTRY(x) .globl CNAME(x); CNAME(x):
++#define ENTRY(x) .globl _##x; _##x:
+ #else
+ #define ENTRY(x) .globl CNAME(x); .type CNAME(x),%function; CNAME(x):
+ #endif /* __APPLE__ */
+@@ -187,7 +187,7 @@ ARM_FUNC_START ffi_call_SYSV
+       @     r1 already set
+       @ Call ffi_prep_args(stack, &ecif)
+-      bl      ffi_prep_args
++      bl      CNAME(ffi_prep_args)
+       @ move first 4 parameters in registers
+       ldmia   sp, {r0-r3}
+@@ -364,7 +364,7 @@ ARM_FUNC_START ffi_call_VFP
+       sub     r2, fp, #64   @ VFP scratch space
+       @ Call ffi_prep_args(stack, &ecif, vfp_space)
+-      bl      ffi_prep_args
++      bl      CNAME(ffi_prep_args)
+       @ Load VFP register args if needed
+       cmp     r0, #0
+@@ -446,7 +446,7 @@ ARM_FUNC_START ffi_closure_VFP
+       sub     sp, sp, #72
+       str     sp, [sp, #64]
+       add     r1, sp, #64
+-      bl      ffi_closure_SYSV_inner
++      bl      CNAME(ffi_closure_SYSV_inner)
+       cmp     r0, #FFI_TYPE_INT
+       beq     .Lretint_vfp
+Index: libffi/src/x86/ffi.c
+===================================================================
+--- libffi.orig/src/x86/ffi.c
++++ libffi/src/x86/ffi.c
+@@ -280,9 +280,11 @@ ffi_status ffi_prep_cif_machdep(ffi_cif
+       else
+ #endif
+         {
++#ifdef X86_WIN32
+           if (cif->abi == FFI_MS_CDECL)
+             cif->flags = FFI_TYPE_MS_STRUCT;
+           else
++#endif
+             cif->flags = FFI_TYPE_STRUCT;
+           /* allocate space for return value pointer */
+           cif->bytes += ALIGN(sizeof(void*), FFI_SIZEOF_ARG);
index f48071e..0a002b3 100644 (file)
@@ -8,3 +8,4 @@ vararg_float_test_fix
 vfp-eabi
 msvc-changes
 win32-return-sign
+apple-fixes
index a6728cf..fb38cd6 100644 (file)
@@ -41,7 +41,7 @@
 #define CNAME(x) x
 #endif
 #ifdef __APPLE__
-#define ENTRY(x) .globl CNAME(x); CNAME(x):
+#define ENTRY(x) .globl _##x; _##x:
 #else
 #define ENTRY(x) .globl CNAME(x); .type CNAME(x),%function; CNAME(x):
 #endif /* __APPLE__ */
@@ -187,7 +187,7 @@ ARM_FUNC_START ffi_call_SYSV
        @     r1 already set
 
        @ Call ffi_prep_args(stack, &ecif)
-       bl      ffi_prep_args
+       bl      CNAME(ffi_prep_args)
 
        @ move first 4 parameters in registers
        ldmia   sp, {r0-r3}
@@ -364,7 +364,7 @@ ARM_FUNC_START ffi_call_VFP
        sub     r2, fp, #64   @ VFP scratch space
 
        @ Call ffi_prep_args(stack, &ecif, vfp_space)
-       bl      ffi_prep_args
+       bl      CNAME(ffi_prep_args)
 
        @ Load VFP register args if needed
        cmp     r0, #0
@@ -446,7 +446,7 @@ ARM_FUNC_START ffi_closure_VFP
        sub     sp, sp, #72
        str     sp, [sp, #64]
        add     r1, sp, #64
-       bl      ffi_closure_SYSV_inner
+       bl      CNAME(ffi_closure_SYSV_inner)
 
        cmp     r0, #FFI_TYPE_INT
        beq     .Lretint_vfp
index 22e2997..611e221 100644 (file)
@@ -280,9 +280,11 @@ ffi_status ffi_prep_cif_machdep(ffi_cif *cif)
       else
 #endif
         {
+#ifdef X86_WIN32
           if (cif->abi == FFI_MS_CDECL)
             cif->flags = FFI_TYPE_MS_STRUCT;
           else
+#endif
             cif->flags = FFI_TYPE_STRUCT;
           /* allocate space for return value pointer */
           cif->bytes += ALIGN(sizeof(void*), FFI_SIZEOF_ARG);
diff --git a/testsuite/libffi.call/a.out b/testsuite/libffi.call/a.out
new file mode 100755 (executable)
index 0000000..93ee85b
Binary files /dev/null and b/testsuite/libffi.call/a.out differ
index 4456161..633519c 100644 (file)
@@ -9,6 +9,7 @@
 static int promotion(signed char sc, signed short ss,
                     unsigned char uc, unsigned short us)
 {
+  printf ("%d + %d + %d + %d\n", (int) sc, (int) ss, (int) uc, (int) us);
   int r = (int) sc + (int) ss + (int) uc + (int) us;
 
   return r;
@@ -50,6 +51,7 @@ int main (void)
        for (us = 0; us <= 60000; us += 10000)
          {
            ul++;
+  printf ("%d + %d + %d + %d\n", (int) sc, (int) ss, (int) uc, (int) us);
            ffi_call(&cif, FFI_FN(promotion), &rint, values);
            CHECK((int)rint == (signed char) sc + (signed short) ss +
                  (unsigned char) uc + (unsigned short) us);