platform/upstream/libffi.git
10 years agoRemove stray hunk that shouldn't have been included in this patch.
Ryan VanderMeulen [Mon, 10 Mar 2014 19:12:47 +0000 (15:12 -0400)]
Remove stray hunk that shouldn't have been included in this patch.

10 years agoEnsure the linker supports @unwind sections in libffi. From Mozilla bug 756740.
Ryan VanderMeulen [Mon, 10 Mar 2014 18:53:48 +0000 (14:53 -0400)]
Ensure the linker supports @unwind sections in libffi. From Mozilla bug 756740.
https://bugzilla.mozilla.org/show_bug.cgi?id=778414

Also tracked as issue #42.
https://github.com/atgreen/libffi/issues/42

10 years agoUpdate Makefile for new darwin scripts
Anthony Green [Sat, 1 Mar 2014 23:37:29 +0000 (18:37 -0500)]
Update Makefile for new darwin scripts

10 years agoAdd README note
Anthony Green [Sat, 1 Mar 2014 23:34:18 +0000 (18:34 -0500)]
Add README note

10 years agoAdd missing -DFFI_DEBUG flag
Anthony Green [Fri, 28 Feb 2014 22:20:59 +0000 (17:20 -0500)]
Add missing -DFFI_DEBUG flag

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Fri, 28 Feb 2014 06:10:17 +0000 (01:10 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

10 years agoMerge pull request #46 from makotokato/android-clang
Anthony Green [Fri, 28 Feb 2014 06:06:48 +0000 (01:06 -0500)]
Merge pull request #46 from makotokato/android-clang

Fix build failure when using clang for Android

10 years agoMerge pull request #46 from makotokato/android-clang
Anthony Green [Fri, 28 Feb 2014 06:06:48 +0000 (01:06 -0500)]
Merge pull request #46 from makotokato/android-clang

Fix build failure when using clang for Android

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Fri, 28 Feb 2014 06:01:29 +0000 (01:01 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

10 years agoMerge pull request #40 from wojdyr/master
Anthony Green [Fri, 28 Feb 2014 06:01:02 +0000 (01:01 -0500)]
Merge pull request #40 from wojdyr/master

Correct the -L flag in libffi.pc.in

10 years agoMerge pull request #66 from ppizarro/master
Anthony Green [Fri, 28 Feb 2014 05:56:27 +0000 (00:56 -0500)]
Merge pull request #66 from ppizarro/master

BlackFin fixes - Fatal error when calling a function defined in a shared library from within the function called by FFI

10 years agoMerge pull request #66 from ppizarro/master
Anthony Green [Fri, 28 Feb 2014 05:56:27 +0000 (00:56 -0500)]
Merge pull request #66 from ppizarro/master

BlackFin fixes - Fatal error when calling a function defined in a shared library from within the function called by FFI

10 years agoAdd ChangeLog entry for Josh's change
Anthony Green [Fri, 28 Feb 2014 05:52:56 +0000 (00:52 -0500)]
Add ChangeLog entry for Josh's change

10 years agoMerge pull request #75 from joshtriplett/longdouble
Anthony Green [Fri, 28 Feb 2014 05:50:25 +0000 (00:50 -0500)]
Merge pull request #75 from joshtriplett/longdouble

Fix build error on x86 without distinct long double

10 years agoRebuilt with new libtool
Anthony Green [Fri, 28 Feb 2014 05:38:41 +0000 (00:38 -0500)]
Rebuilt with new libtool

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Fri, 28 Feb 2014 05:26:57 +0000 (00:26 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

Conflicts:
ChangeLog

10 years agoFix spelling errors
Anthony Green [Fri, 28 Feb 2014 05:23:04 +0000 (00:23 -0500)]
Fix spelling errors

10 years agoFix spelling errors
Anthony Green [Fri, 28 Feb 2014 05:23:04 +0000 (00:23 -0500)]
Fix spelling errors

10 years agoWhen no VFP arguments are present the IP register is used
Anthony Green [Fri, 28 Feb 2014 05:20:17 +0000 (00:20 -0500)]
When no VFP arguments are present the IP register is used
uninitialized. Initialize it to the value of FP.

This fixes a number of testsuite failures when configured for
armv7l-unknown-linux-gnueabihf

10 years agoaarch64 fix
Anthony Green [Fri, 28 Feb 2014 05:17:16 +0000 (00:17 -0500)]
aarch64 fix

10 years agoFix ChangeLog merge
Anthony Green [Thu, 27 Feb 2014 20:42:41 +0000 (15:42 -0500)]
Fix ChangeLog merge

10 years agoFix build error on x86 without distinct long double
Josh Triplett [Mon, 17 Feb 2014 19:20:33 +0000 (11:20 -0800)]
Fix build error on x86 without distinct long double

src/x86/ffi64.c: In function 'classify_argument':
src/x86/ffi64.c:205:5: error: duplicate case value
     case FFI_TYPE_LONGDOUBLE:
     ^
src/x86/ffi64.c:202:5: error: previously used here
     case FFI_TYPE_DOUBLE:
     ^

10 years agoMerge pull request #72 from heiher/devel
Anthony Green [Sat, 15 Feb 2014 13:06:29 +0000 (08:06 -0500)]
Merge pull request #72 from heiher/devel

MIPS N32: Fix call floating point va function

10 years agoMerge pull request #68 from zeldin/master
Anthony Green [Sat, 15 Feb 2014 13:06:11 +0000 (08:06 -0500)]
Merge pull request #68 from zeldin/master

Linux/ppc64: Remove assumption on contents of r11 in closure

10 years agoFix call floating point va function
Heiher [Tue, 21 Jan 2014 15:18:27 +0000 (23:18 +0800)]
Fix call floating point va function

I'm not sure floating-point arguments in GPR or FPR before calling
variable number arguments function. so, load all arguments to GPR and
FPR.

10 years agoAArch64: Fix void fall-through case when assertions are enabled
Zachary Waldowski [Wed, 5 Feb 2014 19:22:52 +0000 (14:22 -0500)]
AArch64: Fix void fall-through case when assertions are enabled

10 years agoAArch64: Fix missing semicolons when assertions are enabled
Zachary Waldowski [Tue, 21 Jan 2014 21:38:31 +0000 (16:38 -0500)]
AArch64: Fix missing semicolons when assertions are enabled

10 years agoRemove build-ios from Makefile
Anthony Green [Sun, 1 Dec 2013 01:58:31 +0000 (20:58 -0500)]
Remove build-ios from Makefile

Conflicts:
ChangeLog

10 years agoMention Aarch64 on iOS
Anthony Green [Sun, 1 Dec 2013 02:06:51 +0000 (21:06 -0500)]
Mention Aarch64 on iOS

10 years agoUpdate ChangeLog
Zachary Waldowski [Mon, 30 Dec 2013 22:36:39 +0000 (17:36 -0500)]
Update ChangeLog

10 years agoAArch64: Remove duplicitous element_count call.
Zachary Waldowski [Thu, 9 Jan 2014 18:50:17 +0000 (13:50 -0500)]
AArch64: Remove duplicitous element_count call.

This inhibits an analyzer warning by Clang.

10 years agoDarwin/aarch64: Respect iOS ABI re: stack argument alignment
Zachary Waldowski [Thu, 9 Jan 2014 18:53:30 +0000 (13:53 -0500)]
Darwin/aarch64: Respect iOS ABI re: stack argument alignment

10 years agoDarwin/aarch64: Fix size_t assumptions
Zachary Waldowski [Thu, 9 Jan 2014 19:03:29 +0000 (14:03 -0500)]
Darwin/aarch64: Fix size_t assumptions

10 years agoDarwin/aarch64: Fix "shadows declaration" warnings
Zachary Waldowski [Mon, 30 Dec 2013 21:14:02 +0000 (16:14 -0500)]
Darwin/aarch64: Fix "shadows declaration" warnings

10 years agoDarwin/aarch64: Use Clang cache invalidation builtin
Zachary Waldowski [Thu, 9 Jan 2014 18:55:21 +0000 (13:55 -0500)]
Darwin/aarch64: Use Clang cache invalidation builtin

10 years agoDarwin/aarch64: Account for long double being equal to double
Zachary Waldowski [Mon, 30 Dec 2013 20:45:51 +0000 (15:45 -0500)]
Darwin/aarch64: Account for long double being equal to double

10 years agoDarwin/aarch64: Use CNAME, restrict .size like ARM
Zachary Waldowski [Mon, 30 Dec 2013 21:33:47 +0000 (16:33 -0500)]
Darwin/aarch64: Use CNAME, restrict .size like ARM

10 years agoDarwin/aarch64: Fix invalid reference in assembly
Zachary Waldowski [Mon, 30 Dec 2013 22:48:22 +0000 (17:48 -0500)]
Darwin/aarch64: Fix invalid reference in assembly

10 years agoDarwin/x86_64: Fix 64-bit type shortening warnings
Zachary Waldowski [Mon, 30 Dec 2013 21:23:21 +0000 (16:23 -0500)]
Darwin/x86_64: Fix 64-bit type shortening warnings

10 years agoDarwin: Merge build scripts, redo project, incl. arm64
Zachary Waldowski [Thu, 9 Jan 2014 18:15:06 +0000 (13:15 -0500)]
Darwin: Merge build scripts, redo project, incl. arm64

10 years agoDarwin/iOS: Improve unified syntax use for LLVM
Zachary Waldowski [Mon, 30 Dec 2013 22:48:10 +0000 (17:48 -0500)]
Darwin/iOS: Improve unified syntax use for LLVM

10 years agoFix dlmalloc warnings due to set_segment_flags, sizeof(size_t)
Zachary Waldowski [Mon, 30 Dec 2013 20:27:44 +0000 (15:27 -0500)]
Fix dlmalloc warnings due to set_segment_flags, sizeof(size_t)

10 years agoDarwin/iOS: Fix mis-typing of vfp_reg_free
Zachary Waldowski [Mon, 30 Dec 2013 20:27:14 +0000 (15:27 -0500)]
Darwin/iOS: Fix mis-typing of vfp_reg_free

10 years agoDarwin/ARM: Assert on NULL dereference
Zachary Waldowski [Mon, 30 Dec 2013 20:26:20 +0000 (15:26 -0500)]
Darwin/ARM: Assert on NULL dereference

This inhibits an analyzer warning by Clang on all platforms.

10 years agoDarwin/i386: Inhibit Clang previous prototype warnings
Zachary Waldowski [Thu, 9 Jan 2014 18:42:08 +0000 (13:42 -0500)]
Darwin/i386: Inhibit Clang previous prototype warnings

10 years agoDarwin/ARM: Inhibit Clang previous prototype warnings
Zachary Waldowski [Thu, 9 Jan 2014 18:41:45 +0000 (13:41 -0500)]
Darwin/ARM: Inhibit Clang previous prototype warnings

10 years agoDarwin/AArch64: Inhibit Clang previous prototype warnings
Zachary Waldowski [Thu, 9 Jan 2014 18:41:27 +0000 (13:41 -0500)]
Darwin/AArch64: Inhibit Clang previous prototype warnings

10 years agoLinux/ppc64: Remove assumption on contents of r11 in closure
Marcus Comstedt [Sat, 4 Jan 2014 18:00:08 +0000 (19:00 +0100)]
Linux/ppc64: Remove assumption on contents of r11 in closure

10 years agoWhen the function called by the ffi called a function defined in a shared library...
Paulo Pizarro [Thu, 2 Jan 2014 18:17:59 +0000 (16:17 -0200)]
When the function called by the ffi called a function defined in a shared library generate a fatal error
The correction was to take into consideration the GOT.

10 years agoUndo iOS ARM64 changes.
Anthony Green [Thu, 5 Dec 2013 21:23:25 +0000 (16:23 -0500)]
Undo iOS ARM64 changes.

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Sun, 1 Dec 2013 03:38:13 +0000 (22:38 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

10 years agoMention Aarch64 on iOS
Anthony Green [Sun, 1 Dec 2013 02:06:51 +0000 (21:06 -0500)]
Mention Aarch64 on iOS

10 years agoMention Aarch64 on iOS
Anthony Green [Sun, 1 Dec 2013 02:06:51 +0000 (21:06 -0500)]
Mention Aarch64 on iOS

10 years agoRemove build-ios from Makefile
Anthony Green [Sun, 1 Dec 2013 01:58:31 +0000 (20:58 -0500)]
Remove build-ios from Makefile

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Sun, 1 Dec 2013 01:54:54 +0000 (20:54 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi
Add ChangeLog entry.

10 years agoMerge pull request #60 from zwaldowski/ios-redo
Anthony Green [Sun, 1 Dec 2013 01:54:39 +0000 (17:54 -0800)]
Merge pull request #60 from zwaldowski/ios-redo

Mac/iOS support, including aarch64 port

10 years agoDarwin: Properly export headers from Xcode project
Zachary Waldowski [Sat, 30 Nov 2013 18:25:27 +0000 (13:25 -0500)]
Darwin: Properly export headers from Xcode project

10 years agoDarwin: Freshen gen scripts, remove old build-ios.sh
Zachary Waldowski [Sat, 30 Nov 2013 17:21:38 +0000 (12:21 -0500)]
Darwin: Freshen gen scripts, remove old build-ios.sh

10 years agoDarwin/iOS: Include x86_64+aarch64 pieces in library
Zachary Waldowski [Sat, 30 Nov 2013 17:03:51 +0000 (12:03 -0500)]
Darwin/iOS: Include x86_64+aarch64 pieces in library

10 years agoDarwin/aarch64: size_t assumptions
Zachary Waldowski [Sat, 30 Nov 2013 08:03:37 +0000 (03:03 -0500)]
Darwin/aarch64: size_t assumptions

10 years agoDarwin/aarch64: Fix “shadows declaration” warnings
Zachary Waldowski [Sat, 30 Nov 2013 07:39:34 +0000 (02:39 -0500)]
Darwin/aarch64: Fix “shadows declaration” warnings

10 years agoDarwin/aarch64: Use Clang cache invalidation builtin
Zachary Waldowski [Sat, 30 Nov 2013 07:08:14 +0000 (02:08 -0500)]
Darwin/aarch64: Use Clang cache invalidation builtin

10 years agoDarwin/aarch64: Inhibit Xcode warning
Zachary Waldowski [Sat, 30 Nov 2013 07:07:48 +0000 (02:07 -0500)]
Darwin/aarch64: Inhibit Xcode warning

10 years agoDarwin/aarch64: double == long double
Zachary Waldowski [Sat, 30 Nov 2013 07:07:34 +0000 (02:07 -0500)]
Darwin/aarch64: double == long double

10 years agoDarwin/iOS prep script: try and compile for arm64
Zachary Waldowski [Sat, 30 Nov 2013 07:06:00 +0000 (02:06 -0500)]
Darwin/iOS prep script: try and compile for arm64

10 years agoDarwin/aarch64: Restrict .size to ELF like arm32.
Zachary Waldowski [Sat, 30 Nov 2013 07:05:22 +0000 (02:05 -0500)]
Darwin/aarch64: Restrict .size to ELF like arm32.

10 years agoDarwin/aarch64: Potentially(?) fix compile error
Zachary Waldowski [Sat, 30 Nov 2013 07:04:57 +0000 (02:04 -0500)]
Darwin/aarch64: Potentially(?) fix compile error

10 years agoDarwin/aarch64: Use CNAME refs
Zachary Waldowski [Sat, 30 Nov 2013 07:04:22 +0000 (02:04 -0500)]
Darwin/aarch64: Use CNAME refs

10 years agoDarwin/Mac: Fix 64/32 shortening warnings
Zachary Waldowski [Sat, 30 Nov 2013 09:13:42 +0000 (04:13 -0500)]
Darwin/Mac: Fix 64/32 shortening warnings

10 years agoDarwin: Misc size_t warnings
Zachary Waldowski [Sat, 30 Nov 2013 08:03:00 +0000 (03:03 -0500)]
Darwin: Misc size_t warnings

10 years agoDarwin: Fix dlmalloc warnings due to sizeof(size_t)
Zachary Waldowski [Sat, 30 Nov 2013 07:55:48 +0000 (02:55 -0500)]
Darwin: Fix dlmalloc warnings due to sizeof(size_t)

10 years agoDarwin: Rebuild Xcode project
Zachary Waldowski [Sat, 30 Nov 2013 09:09:30 +0000 (04:09 -0500)]
Darwin: Rebuild Xcode project

10 years agoDarwin/iOS: Fix LLVM 3.3 warning re: memcpy.
Zachary Waldowski [Sat, 30 Nov 2013 09:09:18 +0000 (04:09 -0500)]
Darwin/iOS: Fix LLVM 3.3 warning re: memcpy.

10 years agoDarwin: Clean up, modernize generator scripts
Zachary Waldowski [Sat, 30 Nov 2013 08:43:42 +0000 (03:43 -0500)]
Darwin: Clean up, modernize generator scripts

10 years agoDarwin/Mac: Also exclude OS X generated source
Zachary Waldowski [Sat, 30 Nov 2013 08:38:02 +0000 (03:38 -0500)]
Darwin/Mac: Also exclude OS X generated source

10 years agoDarwin/iOS: More unified syntax support w/ Clang.
Zachary Waldowski [Tue, 24 Apr 2012 15:16:20 +0000 (11:16 -0400)]
Darwin/iOS: More unified syntax support w/ Clang.

Signed-off-by: Zachary Waldowski <zwaldowski@gmail.com>
10 years agoDarwin/iOS: Simplify RETLDM arguments for LLVM 3.1
Zachary Waldowski [Tue, 24 Apr 2012 14:25:29 +0000 (10:25 -0400)]
Darwin/iOS: Simplify RETLDM arguments for LLVM 3.1

Signed-off-by: Zachary Waldowski <zwaldowski@gmail.com>
10 years agoDarwin: Silence Clang warnings.
Zachary Waldowski [Thu, 12 Apr 2012 03:26:04 +0000 (23:26 -0400)]
Darwin: Silence Clang warnings.

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Fri, 22 Nov 2013 02:25:44 +0000 (21:25 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

Conflicts:
ChangeLog

10 years agoThis separates the 32-bit sysv/linux/bsd code from the 64-bit linux
Alan Modra [Thu, 21 Nov 2013 11:12:35 +0000 (06:12 -0500)]
This separates the 32-bit sysv/linux/bsd code from the 64-bit linux
code, and makes it possible to link code compiled with different
options to those used to compile libffi.  For example, a
-mlong-double-128 libffi can be used with -mlong-double-64 code.

Using the return value area as a place to pass parameters wasn't such
a good idea, causing a failure of cls_ulonglong.c.  I didn't see this
when running the mainline gcc libffi testsuite because that version of
the test is inferior to the upstreamm libffi test.

Using NUM_FPR_ARG_REGISTERS rather than NUM_FPR_ARG_REGISTERS64 meant
that a parameter save area could be allocated before it was strictly
necessary.  Wrong but harmless.  Found when splitting apart ffi.c
into 32-bit and 64-bit support.

10 years agoThis separates the 32-bit sysv/linux/bsd code from the 64-bit linux
Alan Modra [Thu, 21 Nov 2013 11:12:35 +0000 (06:12 -0500)]
This separates the 32-bit sysv/linux/bsd code from the 64-bit linux
code, and makes it possible to link code compiled with different
options to those used to compile libffi.  For example, a
-mlong-double-128 libffi can be used with -mlong-double-64 code.

Using the return value area as a place to pass parameters wasn't such
a good idea, causing a failure of cls_ulonglong.c.  I didn't see this
when running the mainline gcc libffi testsuite because that version of
the test is inferior to the upstreamm libffi test.

Using NUM_FPR_ARG_REGISTERS rather than NUM_FPR_ARG_REGISTERS64 meant
that a parameter save area could be allocated before it was strictly
necessary.  Wrong but harmless.  Found when splitting apart ffi.c
into 32-bit and 64-bit support.

10 years agoMerge pull request #59 from iains/powerpc-darwin-unwind-fix
Anthony Green [Mon, 18 Nov 2013 14:34:04 +0000 (06:34 -0800)]
Merge pull request #59 from iains/powerpc-darwin-unwind-fix

Fix PowerPC Darwin FDE encodings to use pcrel correctly.  Modernise the picbase labels.

10 years agoFix PowerPC Darwin FDE encodings to use pcrel correctly. Modernise the picbase labels.
Iain Sandoe [Mon, 18 Nov 2013 13:11:56 +0000 (13:11 +0000)]
Fix PowerPC Darwin FDE encodings to use pcrel correctly.  Modernise the picbase labels.

10 years agoClean up code to appease modern GCC compiler.
Anthony Green [Mon, 18 Nov 2013 11:52:29 +0000 (06:52 -0500)]
Clean up code to appease modern GCC compiler.

10 years agoAn #endif in the wrong place would cause compile failure on powerpcle.
Alan Modra [Mon, 18 Nov 2013 11:36:03 +0000 (06:36 -0500)]
An #endif in the wrong place would cause compile failure on powerpcle.
Using bl instead of b doesn't cause runtime failures as you might think,
but does mess the processor branch prediction.

10 years agoMerge branch 'master' of github.com:/atgreen/libffi
Anthony Green [Sat, 16 Nov 2013 11:57:54 +0000 (06:57 -0500)]
Merge branch 'master' of github.com:/atgreen/libffi

Conflicts:
ChangeLog
src/powerpc/ffi.c

10 years agoFinally, this adds _CALL_ELF == 2 support. ELFv1 objects can't be
Alan Modra [Sat, 16 Nov 2013 11:53:50 +0000 (06:53 -0500)]
Finally, this adds _CALL_ELF == 2 support.  ELFv1 objects can't be
linked with ELFv2 objects, so this is one case where preprocessor
tests in ffi.c are fine.  Also, there is no need to define a new
FFI_ELFv2 or somesuch value in enum ffi_abi.  FFI_LINUX64 will happily
serve both ABIs.

10 years agoFinally, this adds _CALL_ELF == 2 support. ELFv1 objects can't be
Alan Modra [Sat, 16 Nov 2013 11:53:50 +0000 (06:53 -0500)]
Finally, this adds _CALL_ELF == 2 support.  ELFv1 objects can't be
linked with ELFv2 objects, so this is one case where preprocessor
tests in ffi.c are fine.  Also, there is no need to define a new
FFI_ELFv2 or somesuch value in enum ffi_abi.  FFI_LINUX64 will happily
serve both ABIs.

10 years agoAndreas' 2013-02-08 change reverted some breakage for struct return
Alan Modra [Sat, 16 Nov 2013 11:52:43 +0000 (06:52 -0500)]
Andreas' 2013-02-08 change reverted some breakage for struct return
values from 2011-11-12, but in so doing reintroduced string
instructions to sysv.S that are not supported on all powerpc variants.
This patch properly copies the bounce buffer to destination in C code
rather than in asm.

I have tested this on powerpc64-linux, powerpc-linux and
powerpc-freebsd.  Well, the last on powerpc-linux by lying to
configure with

CC="gcc -m32 -msvr4-struct-return -mlong-double-64" \
CXX="g++ -m32 -msvr4-struct-return -mlong-double-64" \
/src/libffi-current/configure --build=powerpc-freebsd

and then

make && make CC="gcc -m32" CXX="g++ -m32" \
RUNTESTFLAGS=--target_board=unix/-m32/-msvr4-struct-return/-mlong-double-64\
 check

10 years agoThe powerpc64 ABIs align structs passed by value, a fact ignored by
Alan Modra [Sat, 16 Nov 2013 11:41:36 +0000 (06:41 -0500)]
The powerpc64 ABIs align structs passed by value, a fact ignored by
gcc for quite some time.  Since gcc now does the correct alignment,
libffi needs to follow suit.  This ought to be made selectable via
a new abi value, and the #ifdefs removed from ffi.c along with many
other #ifdefs present there and in assembly.  I'll do that with a
followup patch sometime.

This is a revised version of
https://sourceware.org/ml/libffi-discuss/2013/msg00162.html

10 years agoThis patch prepares for ELFv2, where sizes of these areas change. It
Alan Modra [Sat, 16 Nov 2013 11:40:13 +0000 (06:40 -0500)]
This patch prepares for ELFv2, where sizes of these areas change.  It
also makes some minor changes to improve code efficiency.

10 years agoThe powerpc64 support opted to pass floating point values both in the
Alan Modra [Sat, 16 Nov 2013 11:38:55 +0000 (06:38 -0500)]
The powerpc64 support opted to pass floating point values both in the
fpr area and the parameter save area, necessary when the backend
doesn't know if a function argument corresponds to the ellipsis
arguments of a variadic function.  This patch adds powerpc support for
variadic functions, and changes the code to only pass fp in the ABI
mandated area.  ELFv2 needs this change since the parameter save area
may not exist there.

This also fixes two faulty tests that used a non-variadic function
cast to call a variadic function, and spuriously reasoned that this is
somehow necessary for static functions..

10 years agoFix sample closure code
Andrew Haley [Sat, 16 Nov 2013 11:35:51 +0000 (06:35 -0500)]
Fix sample closure code

10 years agoFix broken test cases
Andrew Haley [Sat, 16 Nov 2013 11:29:25 +0000 (06:29 -0500)]
Fix broken test cases

10 years agoMerge branch 'master' of https://github.com/bivab/libffi
Anthony Green [Thu, 14 Nov 2013 15:56:29 +0000 (10:56 -0500)]
Merge branch 'master' of https://github.com/bivab/libffi

Conflicts:
ChangeLog

10 years agoupdate Changelog
David Schneider [Thu, 14 Nov 2013 12:02:16 +0000 (13:02 +0100)]
update Changelog

10 years agoThis enshrines the current testsuite practice of using ffi_arg for
Alan Modra [Wed, 13 Nov 2013 21:55:36 +0000 (16:55 -0500)]
This enshrines the current testsuite practice of using ffi_arg for
returned values.  It would be reasonable and logical to use the actual
return argument type as passed to ffi_prep_cif, but this would mean
changing a large number of tests that use ffi_arg and all backends
that write results to an ffi_arg.

10 years agoRespect HAVE_ALLOCA_H
Anthony Green [Wed, 13 Nov 2013 21:40:28 +0000 (16:40 -0500)]
Respect HAVE_ALLOCA_H

10 years agoadd a testcase for the double/float issue on ARMHF
David Schneider [Wed, 13 Nov 2013 14:50:21 +0000 (15:50 +0100)]
add a testcase for the double/float issue on ARMHF

10 years agostop trying to assing vfp regs once we are done with the registers
David Schneider [Wed, 13 Nov 2013 13:26:57 +0000 (14:26 +0100)]
stop trying to assing vfp regs once we are done with the registers