Import libffi master

.appveyor.yml

@@ -8,9 +8,12 @@ shallow_clone: true
 #  32- and 64-bit clang/mingw
 #  and perhaps more.
 
-image: Visual Studio 2013
+image: Visual Studio 2017
 platform:
   - x64
+  - x86
+  - arm
+  - arm64
 
 environment:
   global:
@@ -18,30 +21,48 @@ environment:
     CYG_CACHE: C:/cygwin/var/cache/setup
     CYG_MIRROR: http://mirrors.kernel.org/sourceware/cygwin/
   matrix:
-    - VSVER: 12
+    - VSVER: 15
 
 install:
   - ps: >-
       If ($env:Platform -Match "x86") {
           $env:VCVARS_PLATFORM="x86"
-          $env:BUILD="x86-pc-cygwin"
-          $env:HOST="x86-pc-windows"
+          $env:BUILD="i686-pc-cygwin"
+          $env:HOST="i686-pc-cygwin"
+          $env:MSVCC="/cygdrive/c/projects/libffi/msvcc.sh"
+          $env:SRC_ARCHITECTURE="x86"
+        } ElseIf ($env:Platform -Match "arm64") {
+          $env:VCVARS_PLATFORM="x86_arm64"
+          $env:BUILD="i686-pc-cygwin"
+          $env:HOST="aarch64-w64-cygwin"
+          $env:MSVCC="/cygdrive/c/projects/libffi/msvcc.sh -marm64"
+          $env:SRC_ARCHITECTURE="aarch64"
+        } ElseIf ($env:Platform -Match "arm") {
+          $env:VCVARS_PLATFORM="x86_arm"
+          $env:BUILD="i686-pc-cygwin"
+          $env:HOST="arm-w32-cygwin"
+          $env:MSVCC="/cygdrive/c/projects/libffi/msvcc.sh -marm"
+          $env:SRC_ARCHITECTURE="arm"
         } Else {
           $env:VCVARS_PLATFORM="amd64"
           $env:BUILD="x86_64-w64-cygwin"
           $env:HOST="x86_64-w64-cygwin"
+          $env:MSVCC="/cygdrive/c/projects/libffi/msvcc.sh -m64"
+          $env:SRC_ARCHITECTURE="x86"
       }
   - 'appveyor DownloadFile https://cygwin.com/setup-x86.exe -FileName setup.exe'
   - 'setup.exe -qnNdO -R "%CYG_ROOT%" -s "%CYG_MIRROR%" -l "%CYG_CACHE%" -P dejagnu >NUL'
   - '%CYG_ROOT%/bin/bash -lc "cygcheck -dc cygwin"'
+  - echo call VsDevCmd to set VS150COMNTOOLS
+  - call "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\Tools\VsDevCmd.bat"
   - ps: $env:VSCOMNTOOLS=(Get-Content ("env:VS" + "$env:VSVER" + "0COMNTOOLS"))
   - echo "Using Visual Studio %VSVER%.0 at %VSCOMNTOOLS%"
-  - call "%VSCOMNTOOLS%\..\..\VC\vcvarsall.bat" %VCVARS_PLATFORM%
+  - call "%VSCOMNTOOLS%..\..\vc\Auxiliary\Build\vcvarsall.bat" %VCVARS_PLATFORM%
 
 build_script:
   - c:\cygwin\bin\sh -lc "(cd $OLDPWD; ./autogen.sh;)"
-  - c:\cygwin\bin\sh -lc "(cd $OLDPWD; ./configure CC='/cygdrive/c/projects/libffi/msvcc.sh -m64' CXX='/cygdrive/c/projects/libffi/msvcc.sh -m64' LD='link' CPP='cl -nologo -EP' CXXCPP='cl -nologo -EP' CPPFLAGS='-DFFI_BUILDING_DLL' AR='/cygdrive/c/projects/libffi/.travis/ar-lib lib' NM='dumpbin -symbols' STRIP=':' --build=$BUILD --host=$HOST;)"
-  - c:\cygwin\bin\sh -lc "(cd $OLDPWD; cp src/x86/ffitarget.h include; make; find .;)"
+  - c:\cygwin\bin\sh -lc "(cd $OLDPWD; ./configure CC='%MSVCC%' CXX='%MSVCC%' LD='link' CPP='cl -nologo -EP' CXXCPP='cl -nologo -EP' CPPFLAGS='-DFFI_BUILDING_DLL' AR='/cygdrive/c/projects/libffi/.travis/ar-lib lib' NM='dumpbin -symbols' STRIP=':' --build=$BUILD --host=$HOST;)"
+  - c:\cygwin\bin\sh -lc "(cd $OLDPWD; cp src/%SRC_ARCHITECTURE%/ffitarget.h include; make; find .;)"
   - c:\cygwin\bin\sh -lc "(cd $OLDPWD; cp `find . -name 'libffi-?.dll'` $HOST/testsuite/; make check; cat `find ./ -name libffi.log`)"
 
 # FIXME: "make check" currently fails.  It just looks like msvcc needs

.gitattributes

@@ -0,0 +1,4 @@
+*	text=auto
+
+*.sln	text eol=crlf
+*.vcxproj*	text eol=crlf

.gitignore

@@ -36,3 +36,7 @@ build_*/
 darwin_*/
 src/arm/trampoline.S
 **/texinfo.tex
+
+x86_64-w64-cygwin
+i686-pc-cygwin
+arm-w32-cygwin

.travis/build.sh

@@ -10,7 +10,16 @@ function build_linux()
     make
     make dist
     make check RUNTESTFLAGS="-a $RUNTESTFLAGS"
-    cat */testsuite/libffi.log
+
+    gzip -c -9 */testsuite/libffi.log > libffi.log.gz
+    echo ================================================================
+    echo The logs are too long for travis to handle, so we compress and
+    echo uuencode them.  Download, decode and uncompress if you need to
+    echo read them.
+    echo ================================================================
+    uuencode libffi.log.gz -
+    echo ================================================================
+    echo ================================================================
 }
 
 function build_ios()

.travis/install.sh

@@ -8,14 +8,17 @@ if [[ $TRAVIS_OS_NAME != 'linux' ]]; then
   brew install libtool dejagnu;
 else
   sudo apt-get update
-  sudo apt-get install dejagnu texinfo
+  sudo apt-get install dejagnu texinfo sharutils
   case "$HOST" in
     i386-pc-linux-gnu)
 	sudo apt-get install gcc-multilib g++-multilib
 	;;
     moxie-elf)
-	echo 'deb http://repos.moxielogic.org:7114/MoxieLogic moxiedev main' | sudo tee -a /etc/apt/sources.list
-	sudo apt-get update -qq
+	echo 'deb https://repos.moxielogic.org:7114/MoxieLogic moxiedev main' | sudo tee -a /etc/apt/sources.list
+	sudo apt-get clean # clear the cache
+	sudo apt-get update ## -qq
+	# debug...
+	curl https://repos.moxielogic.org:7114/MoxieLogic/dists/moxiedev/main/binary-amd64/Packages
 	sudo apt-get install -y --allow-unauthenticated moxielogic-moxie-elf-gcc moxielogic-moxie-elf-gcc-c++ moxielogic-moxie-elf-gcc-libstdc++ moxielogic-moxie-elf-gdb-sim
 	;;
   esac

Makefile.am

@@ -79,9 +79,11 @@ noinst_HEADERS = \
 
 EXTRA_libffi_la_SOURCES = \
 	src/aarch64/ffi.c src/aarch64/sysv.S				\
+	src/aarch64/win64_armasm.S						\
 	src/alpha/ffi.c src/alpha/osf.S					\
 	src/arc/ffi.c src/arc/arcompact.S				\
 	src/arm/ffi.c src/arm/sysv.S					\
+	src/arm/ffi.c src/arm/sysv_msvc_arm32.S			\
 	src/avr32/ffi.c src/avr32/sysv.S				\
 	src/bfin/ffi.c src/bfin/sysv.S					\
 	src/cris/ffi.c src/cris/sysv.S					\

README.md

@@ -51,11 +51,13 @@ tested:
 | --------------- | ---------------- | ----------------------- |
 | AArch64 (ARM64) | iOS              | Clang                   |
 | AArch64         | Linux            | GCC                     |
+| AArch64         | Windows          | MSVC                    |
 | Alpha           | Linux            | GCC                     |
 | Alpha           | Tru64            | GCC                     |
 | ARC             | Linux            | GCC                     |
 | ARM             | Linux            | GCC                     |
 | ARM             | iOS              | GCC                     |
+| ARM             | Windows          | MSVC                    |
 | AVR32           | Linux            | GCC                     |
 | Blackfin        | uClinux          | GCC                     |
 | HPPA            | HPUX             | GCC                     |
@@ -168,6 +170,11 @@ remove the line in configure that sets 'fix_srcfile_path' to a 'cygpath'
 command.  ('cygpath' is not present in MingW, and is not required when
 using MingW-style paths.)
 
+To build static library for ARM64 with MSVC using visual studio solution, msvc_build folder have
+   aarch64/Ffi_staticLib.sln
+   required header files in aarch64/aarch64_include/
+
+
 SPARC Solaris builds require the use of the GNU assembler and linker.
 Point ``AS`` and ``LD`` environment variables at those tool prior to
 configuration.
@@ -194,9 +201,11 @@ See the git log for details at http://github.com/libffi/libffi.
         Add RISC-V support.
         New API in support of GO closures.
         Default to Microsoft's 64 bit long double ABI with Visual C++.
-          GNU compiler uses 80 bits (128 in memory) FFI_GNUW64 ABI.
-	Many new tests cases and bug fixes.
-    
+        GNU compiler uses 80 bits (128 in memory) FFI_GNUW64 ABI.
+        Many new tests cases and bug fixes.
+        Add windows on arm64 (WOA) support.
+        Add Windows 32-bit arm support.
+        
     3.2.1 Nov-12-14
         Build fix for non-iOS AArch64 targets.
     

configure.host

@@ -6,6 +6,11 @@
 # THIS TABLE IS SORTED.  KEEP IT THAT WAY.
 # Most of the time we can define all the variables all at once...
 case "${host}" in
+  aarch64*-*-cygwin* | aarch64*-*-mingw* | aarch64*-*-win* )
+	TARGET=ARM_WIN64; TARGETDIR=aarch64
+	MSVC=1
+	;;
+
   aarch64*-*-*)
 	TARGET=AARCH64; TARGETDIR=aarch64
 	SOURCES="ffi.c sysv.S"
@@ -23,6 +28,11 @@ case "${host}" in
 	SOURCES="ffi.c arcompact.S"
 	;;
 
+  arm*-*-cygwin* | arm*-*-mingw* | arm*-*-win* )
+	TARGET=ARM_WIN32; TARGETDIR=arm
+	MSVC=1
+	;;
+
   arm*-*-*)
 	TARGET=ARM; TARGETDIR=arm
 	SOURCES="ffi.c sysv.S"
@@ -74,12 +84,6 @@ case "${host}" in
 	fi
 	if test "${ax_cv_c_compiler_vendor}" = "microsoft"; then
 	  MSVC=1
-	  if test $ac_cv_sizeof_size_t = 4; then
-	    # libffi does not support microsoft tools for 32-bit windows
-	    # hosts.  Try porting src/x86/sysv.S to intel assembly
-	    # format.
-	    UNSUPPORTED=1
-          fi
 	fi
 	# All mingw/cygwin/win32 builds require -no-undefined for sharedlib.
 	# We must also check with_cross_host to decide if this is a native
@@ -248,6 +252,12 @@ esac
 
 # ... but some of the cases above share configury.
 case "${TARGET}" in
+  ARM_WIN32)
+	SOURCES="ffi.c sysv_msvc_arm32.S"
+	;;
+  ARM_WIN64)
+	SOURCES="ffi.c win64_armasm.S"
+	;;
   MIPS)
 	SOURCES="ffi.c o32.S n32.S"
 	;;
@@ -265,7 +275,11 @@ case "${TARGET}" in
 	SOURCES="ffi.c ffi_sysv.c sysv.S ppc_closure.S"
 	;;
   X86 | X86_DARWIN | X86_FREEBSD | X86_WIN32)
-	SOURCES="ffi.c sysv.S"
+	if test "$MSVC" = 1; then
+		SOURCES="ffi.c sysv_intel.S"
+	else
+		SOURCES="ffi.c sysv.S"
+	fi
 	;;
   X86_64)
 	if test x"$TARGET_X32" = xyes; then

include/ffi.h.in

@@ -283,11 +283,13 @@ FFI_API size_t ffi_raw_size (ffi_cif *cif);
    packing, even on 64-bit machines.  I.e. on 64-bit machines longs
    and doubles are followed by an empty 64-bit word.  */
 
+#if !FFI_NATIVE_RAW_API
 FFI_API
 void ffi_java_raw_call (ffi_cif *cif,
 			void (*fn)(void),
 			void *rvalue,
 			ffi_java_raw *avalue);
+#endif
 
 FFI_API
 void ffi_java_ptrarray_to_raw (ffi_cif *cif, void **args, ffi_java_raw *raw);
@@ -414,6 +416,7 @@ ffi_prep_raw_closure_loc (ffi_raw_closure*,
 			  void *user_data,
 			  void *codeloc);
 
+#if !FFI_NATIVE_RAW_API
 FFI_API ffi_status
 ffi_prep_java_raw_closure (ffi_java_raw_closure*,
 		           ffi_cif *cif,
@@ -426,6 +429,7 @@ ffi_prep_java_raw_closure_loc (ffi_java_raw_closure*,
 			       void (*fun)(ffi_cif*,void*,ffi_java_raw*,void*),
 			       void *user_data,
 			       void *codeloc);
+#endif
 
 #endif /* FFI_CLOSURES */
 

include/ffi_common.h

@@ -99,6 +99,10 @@ ffi_status ffi_prep_cif_core(ffi_cif *cif,
 			     ffi_type *rtype,
 			     ffi_type **atypes);
 
+/* Translate a data pointer to a code pointer.  Needed for closures on
+   some targets.  */
+void *ffi_data_to_code_pointer (void *data) FFI_HIDDEN;
+
 /* Extended cif, used in callback from assembly routine */
 typedef struct
 {

m4/asmcfi.m4

@@ -2,7 +2,7 @@ AC_DEFUN([GCC_AS_CFI_PSEUDO_OP],
 [AC_CACHE_CHECK([assembler .cfi pseudo-op support],
     gcc_cv_as_cfi_pseudo_op, [
     gcc_cv_as_cfi_pseudo_op=unknown
-    AC_TRY_COMPILE([asm (".cfi_startproc\n\t.cfi_endproc");],,
+    AC_TRY_COMPILE([asm (".cfi_sections\n\t.cfi_startproc\n\t.cfi_endproc");],,
 		   [gcc_cv_as_cfi_pseudo_op=yes],
 		   [gcc_cv_as_cfi_pseudo_op=no])
  ])

msvc_build/aarch64/Ffi_staticLib.sln

@@ -0,0 +1,33 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio 15
+VisualStudioVersion = 15.0.28302.56
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "Ffi_staticLib_arm64", "Ffi_staticLib.vcxproj", "{115502C0-BE05-4767-BF19-5C87D805FAD6}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|ARM64 = Debug|ARM64
+		Debug|x64 = Debug|x64
+		Debug|x86 = Debug|x86
+		Release|ARM64 = Release|ARM64
+		Release|x64 = Release|x64
+		Release|x86 = Release|x86
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Debug|ARM64.ActiveCfg = Debug|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Debug|ARM64.Build.0 = Debug|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Debug|x64.ActiveCfg = Debug|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Debug|x86.ActiveCfg = Debug|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Release|ARM64.ActiveCfg = Release|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Release|ARM64.Build.0 = Release|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Release|x64.ActiveCfg = Release|ARM64
+		{115502C0-BE05-4767-BF19-5C87D805FAD6}.Release|x86.ActiveCfg = Release|ARM64
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+	GlobalSection(ExtensibilityGlobals) = postSolution
+		SolutionGuid = {241C54C7-20DD-4897-9376-E6B6D1B43BD5}
+	EndGlobalSection
+EndGlobal

msvc_build/aarch64/Ffi_staticLib.vcxproj

@@ -0,0 +1,130 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="15.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|ARM64">
+      <Configuration>Debug</Configuration>
+      <Platform>ARM64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|ARM64">
+      <Configuration>Release</Configuration>
+      <Platform>ARM64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <VCProjectVersion>15.0</VCProjectVersion>
+    <ProjectGuid>{115502C0-BE05-4767-BF19-5C87D805FAD6}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <RootNamespace>FfistaticLib</RootNamespace>
+    <WindowsTargetPlatformVersion>10.0.17763.0</WindowsTargetPlatformVersion>
+    <ProjectName>Ffi_staticLib_arm64</ProjectName>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>true</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <CharacterSet>Unicode</CharacterSet>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'" Label="Configuration">
+    <ConfigurationType>StaticLibrary</ConfigurationType>
+    <UseDebugLibraries>false</UseDebugLibraries>
+    <PlatformToolset>v141</PlatformToolset>
+    <WholeProgramOptimization>true</WholeProgramOptimization>
+    <CharacterSet>Unicode</CharacterSet>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+  </ImportGroup>
+  <ImportGroup Label="Shared">
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <ImportGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'" Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">
+    <LinkIncremental>true</LinkIncremental>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">
+    <LinkIncremental>false</LinkIncremental>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>Disabled</Optimization>
+      <SDLCheck>true</SDLCheck>
+      <PreprocessorDefinitions>FFI_BUILDING_DLL;_DEBUG;_LIB;USE_DL_PREFIX;ARM64;_M_ARM64;NDEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <ConformanceMode>true</ConformanceMode>
+      <AdditionalIncludeDirectories>..\..\include;.\aarch64_include;..\..\src\aarch64;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <IgnoreStandardIncludePath>false</IgnoreStandardIncludePath>
+      <BrowseInformation>true</BrowseInformation>
+      <OmitFramePointers>
+      </OmitFramePointers>
+      <WholeProgramOptimization>false</WholeProgramOptimization>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">
+    <ClCompile>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <WarningLevel>Level3</WarningLevel>
+      <Optimization>MaxSpeed</Optimization>
+      <FunctionLevelLinking>true</FunctionLevelLinking>
+      <IntrinsicFunctions>true</IntrinsicFunctions>
+      <SDLCheck>true</SDLCheck>
+      <PreprocessorDefinitions>FFI_BUILDING_DLL;USE_DL_PREFIX;ARM64;NDEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
+      <ConformanceMode>true</ConformanceMode>
+      <AdditionalIncludeDirectories>..\..\include;.\aarch64_include;..\..\src\aarch64;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OmitFramePointers>true</OmitFramePointers>
+      <FavorSizeOrSpeed>Speed</FavorSizeOrSpeed>
+      <EnableFiberSafeOptimizations>true</EnableFiberSafeOptimizations>
+      <AdditionalUsingDirectories>..\..\src;..\..\src\aarch64;%(AdditionalUsingDirectories)</AdditionalUsingDirectories>
+    </ClCompile>
+    <Link>
+      <SubSystem>Windows</SubSystem>
+      <EnableCOMDATFolding>true</EnableCOMDATFolding>
+      <OptimizeReferences>true</OptimizeReferences>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+    </Link>
+    <ProjectReference>
+      <LinkLibraryDependencies>true</LinkLibraryDependencies>
+    </ProjectReference>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <ClInclude Include=".\aarch64_include\ffi.h" />
+    <ClInclude Include=".\aarch64_include\fficonfig.h" />
+    <ClInclude Include="..\..\src\aarch64\ffitarget.h" />
+    <ClInclude Include="..\include\ffi_cfi.h" />
+    <ClInclude Include="..\include\ffi_common.h" />
+    <ClInclude Include="..\..\src\aarch64\internal.h" />
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="..\..\src\closures.c" />
+    <ClCompile Include="..\..\src\dlmalloc.c" />
+    <ClCompile Include="..\..\src\aarch64\ffi.c" />
+    <ClCompile Include="..\..\src\prep_cif.c" />
+    <ClCompile Include="..\..\src\types.c" />
+  </ItemGroup>
+  <!--ItemGroup>
+    <Object Include="..\..\..\..\Downloads\libffi-master-win64\src\aarch64\win64_armasm.obj" />
+  </ItemGroup-->
+  <ItemGroup>
+    <CustomBuild Include="..\..\src\aarch64\win64_armasm.S">
+      <!--ExcludedFromBuild Condition="'$(Platform)'!='ARM64'">true</ExcludedFromBuild -->
+      <Command>
+        cl /FA /EP /nologo /I"..\..\include" /I".\aarch64_include" /I"..\..\src\aarch64" "%(FullPath)" &gt; $(IntDir)win64_armasm.i
+        armasm64 $(IntDir)win64_armasm.i /I"src\" /I"..\..\include" /I"..\..\src\aarch64" -o "$(IntDir)win64_armasm.obj"
+      </Command>
+      <Outputs>win64_armasm.obj;%(Outputs)</Outputs>
+    </CustomBuild>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+  </ImportGroup>
+</Project>

msvc_build/aarch64/Ffi_staticLib.vcxproj.filters

@@ -0,0 +1,57 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <Filter Include="Source Files">
+      <UniqueIdentifier>{4FC737F1-C7A5-4376-A066-2A32D752A2FF}</UniqueIdentifier>
+      <Extensions>cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx</Extensions>
+    </Filter>
+    <Filter Include="Header Files">
+      <UniqueIdentifier>{93995380-89BD-4b04-88EB-625FBE52EBFB}</UniqueIdentifier>
+      <Extensions>h;hh;hpp;hxx;hm;inl;inc;ipp;xsd</Extensions>
+    </Filter>
+    <Filter Include="Resource Files">
+      <UniqueIdentifier>{67DA6AB6-F800-4c08-8B7A-83BB121AAD01}</UniqueIdentifier>
+      <Extensions>rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav;mfcribbon-ms</Extensions>
+    </Filter>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="..\include\ffi.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\include\ffi_cfi.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\include\ffi_common.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\include\fficonfig.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\include\ffitarget.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="..\include\internal.h">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="src\closures.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="src\dlmalloc.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="src\ffi.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="src\prep_cif.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="src\types.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+  </ItemGroup>
+  <ItemGroup>
+    <CustomBuild Include="src\win64_armasm.S" />
+  </ItemGroup>
+</Project>

msvc_build/aarch64/Ffi_staticLib.vcxproj.user

@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="15.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <PropertyGroup />
+</Project>

msvc_build/aarch64/aarch64_include/ffi.h

@@ -49,26 +49,9 @@ extern "C" {
 #endif
 
 /* Specify which architecture libffi is configured for. */
-/* --- Begin CPython patch ---
-ZCW: This is patched by us to avoid having to keep two mostly-identical files.
-#ifndef X86_WIN32
-#define X86_WIN32
+#ifndef AARCH64
+#define AARCH64
 #endif
-*/
-
-#ifdef _MSC_VER
-#ifdef _WIN64
-# ifndef X86_WIN64
-#  define X86_WIN64
-# endif
-#else
-# ifndef X86_WIN32
-#  define X86_WIN32
-# endif
-#endif
-#endif /* _MSC_VER */
-
-/* --- End CPython patch --- */
 
 /* ---- System configuration information --------------------------------- */
 
@@ -134,10 +117,8 @@ typedef struct _ffi_type
 #if defined _MSC_VER
 # if defined FFI_BUILDING_DLL /* Building libffi.DLL with msvcc.sh */
 #  define FFI_API __declspec(dllexport)
-#error FFI_BUILDING_DLL should not be defined
 # elif !defined FFI_BUILDING  /* Importing libffi.DLL */
 #  define FFI_API __declspec(dllimport)
-#error FFI_BUILDING should be defined
 # else                        /* Building/linking static library */
 #  define FFI_API
 # endif
@@ -216,7 +197,7 @@ FFI_EXTERN ffi_type ffi_type_float;
 FFI_EXTERN ffi_type ffi_type_double;
 FFI_EXTERN ffi_type ffi_type_pointer;
 
-#if 1
+#ifndef _M_ARM64
 FFI_EXTERN ffi_type ffi_type_longdouble;
 #else
 #define ffi_type_longdouble ffi_type_double
@@ -503,7 +484,7 @@ ffi_status ffi_get_struct_offsets (ffi_abi abi, ffi_type *struct_type,
 #define FFI_TYPE_INT        1
 #define FFI_TYPE_FLOAT      2    
 #define FFI_TYPE_DOUBLE     3
-#if 1
+#ifndef _M_ARM64
 #define FFI_TYPE_LONGDOUBLE 4
 #else
 #define FFI_TYPE_LONGDOUBLE FFI_TYPE_DOUBLE
@@ -519,9 +500,9 @@ ffi_status ffi_get_struct_offsets (ffi_abi abi, ffi_type *struct_type,
 #define FFI_TYPE_STRUCT     13
 #define FFI_TYPE_POINTER    14
 #define FFI_TYPE_COMPLEX    15
-
 /* This should always refer to the last type code (for sanity checks).  */
-#define FFI_TYPE_LAST       FFI_TYPE_COMPLEX
+#define FFI_TYPE_LAST   FFI_TYPE_COMPLEX
+
 
 #ifdef __cplusplus
 }

msvc_build/aarch64/aarch64_include/fficonfig.h

@@ -1,14 +1,6 @@
-/* fficonfig.h.  Originally created by configure, now hand_maintained for MSVC. */
-
 /* fficonfig.h.  Generated from fficonfig.h.in by configure.  */
 /* fficonfig.h.in.  Generated from configure.ac by autoheader.  */
 
-/* Define this for MSVC, but not for mingw32! */
-#ifdef _MSC_VER
-#define __attribute__(x) /* */
-#endif
-#define alloca _alloca
-
 /* Define if building universal (internal helper macro) */
 /* #undef AC_APPLE_UNIVERSAL_BUILD */
 
@@ -21,10 +13,7 @@
 /* #undef C_ALLOCA */
 
 /* Define to the flags needed for the .section .eh_frame directive. */
-/* #undef EH_FRAME_FLAGS */
-
-/* using static library */
-#define FFI_BUILDING 1
+#define EH_FRAME_FLAGS "a"
 
 /* Define this if you want extra debugging. */
 /* #undef FFI_DEBUG */
@@ -49,10 +38,10 @@
 
 /* Define to 1 if you have <alloca.h> and it should be used (not on Ultrix).
    */
-/* #undef HAVE_ALLOCA_H */
+/*#define HAVE_ALLOCA_H 1 */
 
 /* Define if your assembler supports .cfi_* directives. */
-/* #undef HAVE_AS_CFI_PSEUDO_OP */
+#define HAVE_AS_CFI_PSEUDO_OP 1
 
 /* Define if your assembler supports .register. */
 /* #undef HAVE_AS_REGISTER_PSEUDO_OP */
@@ -68,26 +57,23 @@
 /* #undef HAVE_AS_X86_64_UNWIND_SECTION_TYPE */
 
 /* Define if your assembler supports PC relative relocs. */
-#define HAVE_AS_X86_PCREL 1
+/* #undef HAVE_AS_X86_PCREL */
 
 /* Define to 1 if you have the <dlfcn.h> header file. */
-/* #undef HAVE_DLFCN_H */
+#define HAVE_DLFCN_H 1
 
 /* Define if __attribute__((visibility("hidden"))) is supported. */
-/* #undef HAVE_HIDDEN_VISIBILITY_ATTRIBUTE */
+#define HAVE_HIDDEN_VISIBILITY_ATTRIBUTE 1
 
 /* Define to 1 if you have the <inttypes.h> header file. */
 #define HAVE_INTTYPES_H 1
 
 /* Define if you have the long double type and it is bigger than a double */
-/* #undef HAVE_LONG_DOUBLE */
+#define HAVE_LONG_DOUBLE 1
 
 /* Define if you support more than one size of the long double type */
 /* #undef HAVE_LONG_DOUBLE_VARIANT */
 
-/* Define if you have the ANSI C header files.  */
-#define STDC_HEADERS 1
-
 /* Define to 1 if you have the `memcpy' function. */
 #define HAVE_MEMCPY 1
 
@@ -95,37 +81,37 @@
 #define HAVE_MEMORY_H 1
 
 /* Define to 1 if you have the `mkostemp' function. */
-/* #undef HAVE_MKOSTEMP */
+#define HAVE_MKOSTEMP 1
 
 /* Define to 1 if you have the `mmap' function. */
-/* #undef HAVE_MMAP */
+#define HAVE_MMAP 1
 
 /* Define if mmap with MAP_ANON(YMOUS) works. */
-/* #undef HAVE_MMAP_ANON */
+#define HAVE_MMAP_ANON 1
 
 /* Define if mmap of /dev/zero works. */
-/* #undef HAVE_MMAP_DEV_ZERO */
+#define HAVE_MMAP_DEV_ZERO 1
 
 /* Define if read-only mmap of a plain file works. */
-/* #undef HAVE_MMAP_FILE */
+#define HAVE_MMAP_FILE 1
 
 /* Define if .eh_frame sections should be read-only. */
-/* #undef HAVE_RO_EH_FRAME */
+#define HAVE_RO_EH_FRAME 1
 
 /* Define to 1 if you have the <stdint.h> header file. */
 #define HAVE_STDINT_H 1
 
 /* Define to 1 if you have the <stdlib.h> header file. */
-#define HAVE_STDLIB_H 1
-
+//#define HAVE_STDLIB_H 0
+#define LACKS_STDLIB_H 1
 /* Define to 1 if you have the <strings.h> header file. */
-/* #undef HAVE_STRINGS_H */
+#define HAVE_STRINGS_H 1
 
 /* Define to 1 if you have the <string.h> header file. */
 #define HAVE_STRING_H 1
 
 /* Define to 1 if you have the <sys/mman.h> header file. */
-/* #undef HAVE_SYS_MMAN_H */
+#define HAVE_SYS_MMAN_H 1
 
 /* Define to 1 if you have the <sys/stat.h> header file. */
 #define HAVE_SYS_STAT_H 1
@@ -134,14 +120,18 @@
 #define HAVE_SYS_TYPES_H 1
 
 /* Define to 1 if you have the <unistd.h> header file. */
-/* #undef HAVE_UNISTD_H */
+#define HAVE_UNISTD_H 1
 
 /* Define to 1 if GNU symbol versioning is used for libatomic. */
-/* #undef LIBFFI_GNU_SYMBOL_VERSIONING */
+#define LIBFFI_GNU_SYMBOL_VERSIONING 1
 
-/* Define to the sub-directory where libtool stores uninstalled libraries. */
+/* Define to the sub-directory in which libtool stores uninstalled libraries.
+   */
 #define LT_OBJDIR ".libs/"
 
+/* Define to 1 if your C compiler doesn't accept -c and -o together. */
+/* #undef NO_MINUS_C_MINUS_O */
+
 /* Name of package */
 #define PACKAGE "libffi"
 
@@ -167,10 +157,7 @@
 #define SIZEOF_DOUBLE 8
 
 /* The size of `long double', as computed by sizeof. */
-#define SIZEOF_LONG_DOUBLE 12
-
-/* Define if you have the long double type and it is bigger than a double */
-#define HAVE_LONG_DOUBLE 1
+#define SIZEOF_LONG_DOUBLE 8
 
 /* The size of `size_t', as computed by sizeof. */
 #define SIZEOF_SIZE_T 8

msvcc.sh

@@ -60,7 +60,7 @@ do
   case $1
   in
     --verbose)
-      $verbose=1
+      verbose=1
       shift 1
     ;;
     --version)
@@ -80,6 +80,16 @@ do
       safeseh=
       shift 1
     ;;
+    -marm)
+      ml='armasm'
+      safeseh=
+      shift 1
+    ;;
+    -marm64)
+      ml='armasm64'
+      safeseh=
+      shift 1
+    ;;
     -clang-cl)
       cl="clang-cl"
       shift 1
@@ -237,6 +247,7 @@ do
       else
         output="-Fe$2"
       fi
+      armasm_output="-o $2"
       if [ -n "$assembly" ]; then
         args="$args $output"
       else
@@ -289,13 +300,27 @@ if [ -n "$assembly" ]; then
     fi
     ppsrc="$outdir/$(basename $src|sed 's/.S$/.asm/g')"
 
+    if [ $ml = "armasm" ]; then
+      defines="$defines -D_M_ARM"
+    fi
+
+    if [ $ml = "armasm64" ]; then
+      defines="$defines -D_M_ARM64"
+    fi
+
     if test -n "$verbose"; then
       echo "$cl -nologo -EP $includes $defines $src > $ppsrc"
     fi
 
     "$cl" -nologo -EP $includes $defines $src > $ppsrc || exit $?
     output="$(echo $output | sed 's%/F[dpa][^ ]*%%g')"
-    args="-nologo $safeseh $single $output $ppsrc"
+    if [ $ml = "armasm" ]; then
+      args="-nologo -g -oldit $armasm_output $ppsrc -errorReport:prompt"
+    elif [ $ml = "armasm64" ]; then
+      args="-nologo -g $armasm_output $ppsrc -errorReport:prompt"
+    else
+      args="-nologo $safeseh $single $output $ppsrc"
+    fi
 
     if test -n "$verbose"; then
       echo "$ml $args"

src/aarch64/ffi.c

@@ -19,6 +19,7 @@ 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(__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
@@ -26,6 +27,9 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 #include <ffi.h>
 #include <ffi_common.h>
 #include "internal.h"
+#ifdef _M_ARM64
+#include <windows.h> /* FlushInstructionCache */
+#endif
 
 /* Force FFI_TYPE_LONGDOUBLE to be different than FFI_TYPE_DOUBLE;
    all further uses in this file will refer to the 128-bit type.  */
@@ -74,6 +78,8 @@ ffi_clear_cache (void *start, void *end)
   sys_icache_invalidate (start, (char *)end - (char *)start);
 #elif defined (__GNUC__)
   __builtin___clear_cache (start, end);
+#elif defined (_M_ARM64)
+  FlushInstructionCache(GetCurrentProcess(), start, (char*)end - (char*)start);
 #else
 #error "Missing builtin to flush instruction cache"
 #endif
@@ -315,6 +321,9 @@ extend_integer_type (void *source, int type)
     }
 }
 
+#if defined(_MSC_VER)
+void extend_hfa_type (void *dest, void *src, int h);
+#else
 static void
 extend_hfa_type (void *dest, void *src, int h)
 {
@@ -368,7 +377,11 @@ extend_hfa_type (void *dest, void *src, int h)
     : "r"(f * 12), "r"(dest), "r"(src)
     : "memory", "v16", "v17", "v18", "v19");
 }
+#endif
 
+#if defined(_MSC_VER)
+void* compress_hfa_type (void *dest, void *src, int h);
+#else
 static void *
 compress_hfa_type (void *dest, void *reg, int h)
 {
@@ -437,6 +450,7 @@ compress_hfa_type (void *dest, void *reg, int h)
     }
   return dest;
 }
+#endif
 
 /* Either allocate an appropriate register for the argument type, or if
    none are available, allocate a stack slot and return a pointer
@@ -454,7 +468,7 @@ allocate_int_to_reg_or_stack (struct call_context *context,
   return allocate_to_stack (state, stack, size, size);
 }
 
-ffi_status
+ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep (ffi_cif *cif)
 {
   ffi_type *rtype = cif->rtype;
@@ -539,9 +553,9 @@ ffi_prep_cif_machdep (ffi_cif *cif)
 
 #if defined (__APPLE__)
 /* Perform Apple-specific cif processing for variadic calls */
-ffi_status ffi_prep_cif_machdep_var(ffi_cif *cif,
-				    unsigned int nfixedargs,
-				    unsigned int ntotalargs)
+ffi_status FFI_HIDDEN
+ffi_prep_cif_machdep_var(ffi_cif *cif, unsigned int nfixedargs,
+			 unsigned int ntotalargs)
 {
   ffi_status status = ffi_prep_cif_machdep (cif);
   cif->aarch64_nfixedargs = nfixedargs;
@@ -588,8 +602,8 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
   /* Allocate consectutive stack for everything we'll need.  */
   context = alloca (sizeof(struct call_context) + stack_bytes + 32 + rsize);
   stack = context + 1;
-  frame = stack + stack_bytes;
-  rvalue = (rsize ? frame + 32 : orig_rvalue);
+  frame = (void*)((uintptr_t)stack + (uintptr_t)stack_bytes);
+  rvalue = (rsize ? (void*)((uintptr_t)frame + 32) : orig_rvalue);
 
   arg_init (&state);
   for (i = 0, nargs = cif->nargs; i < nargs; i++)
@@ -651,6 +665,22 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
 	    if (h)
 	      {
 		int elems = 4 - (h & 3);
+#ifdef _M_ARM64 /* for handling armasm calling convention */
+                if (cif->is_variadic)
+                  {
+                    if (state.ngrn + elems <= N_X_ARG_REG)
+                      {
+                        dest = &context->x[state.ngrn];
+                        state.ngrn += elems;
+                        extend_hfa_type(dest, a, h);
+                        break;
+                      }
+                    state.nsrn = N_X_ARG_REG;
+                    dest = allocate_to_stack(&state, stack, ty->alignment, s);
+                  }
+                else
+                  {
+#endif /* for handling armasm calling convention */
 	        if (state.nsrn + elems <= N_V_ARG_REG)
 		  {
 		    dest = &context->v[state.nsrn];
@@ -660,6 +690,9 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
 		  }
 		state.nsrn = N_V_ARG_REG;
 		dest = allocate_to_stack (&state, stack, ty->alignment, s);
+#ifdef _M_ARM64 /* for handling armasm calling convention */
+	      }
+#endif /* for handling armasm calling convention */
 	      }
 	    else if (s > 16)
 	      {
@@ -681,7 +714,7 @@ ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
 		       X registers, then the argument is copied into
 		       consecutive X registers.  */
 		    dest = &context->x[state.ngrn];
-		    state.ngrn += n;
+                    state.ngrn += (unsigned int)n;
 		  }
 		else
 		  {
@@ -773,6 +806,16 @@ ffi_prep_closure_loc (ffi_closure *closure,
   *(UINT64 *)(tramp + 16) = (uintptr_t)start;
 
   ffi_clear_cache(tramp, tramp + FFI_TRAMPOLINE_SIZE);
+
+  /* Also flush the cache for code mapping.  */
+#ifdef _M_ARM64
+  // Not using dlmalloc.c for Windows ARM64 builds
+  // so calling ffi_data_to_code_pointer() isn't necessary
+  unsigned char *tramp_code = tramp;
+  #else
+  unsigned char *tramp_code = ffi_data_to_code_pointer (tramp);
+  #endif
+  ffi_clear_cache (tramp_code, tramp_code + FFI_TRAMPOLINE_SIZE);
 #endif
 
   closure->cif = cif;
@@ -871,55 +914,78 @@ ffi_closure_SYSV_inner (ffi_cif *cif,
 	  if (h)
 	    {
 	      n = 4 - (h & 3);
-	      if (state.nsrn + n <= N_V_ARG_REG)
-		{
-		  void *reg = &context->v[state.nsrn];
-		  state.nsrn += n;
+#ifdef _M_ARM64  /* for handling armasm calling convention */
+              if (cif->is_variadic)
+                {
+                  if (state.ngrn + n <= N_X_ARG_REG)
+                    {
+                      void *reg = &context->x[state.ngrn];
+                      state.ngrn += (unsigned int)n;
+    
+                      /* Eeek! We need a pointer to the structure, however the
+                       homogeneous float elements are being passed in individual
+                       registers, therefore for float and double the structure
+                       is not represented as a contiguous sequence of bytes in
+                       our saved register context.  We don't need the original
+                       contents of the register storage, so we reformat the
+                       structure into the same memory.  */
+                      avalue[i] = compress_hfa_type(reg, reg, h);
+                    }
+                  else
+                    {
+                      state.ngrn = N_X_ARG_REG;
+                      state.nsrn = N_V_ARG_REG;
+                      avalue[i] = allocate_to_stack(&state, stack,
+                             ty->alignment, s);
+                    }
+                }
+              else
+                {
+#endif  /* for handling armasm calling convention */
+                  if (state.nsrn + n <= N_V_ARG_REG)
+                    {
+                      void *reg = &context->v[state.nsrn];
+                      state.nsrn += (unsigned int)n;
+                      avalue[i] = compress_hfa_type(reg, reg, h);
+                    }
+                  else
+                    {
+                      state.nsrn = N_V_ARG_REG;
+                      avalue[i] = allocate_to_stack(&state, stack,
+                                                   ty->alignment, s);
+                    }
+#ifdef _M_ARM64  /* for handling armasm calling convention */
+                }
+#endif  /* for handling armasm calling convention */
+            }
+          else if (s > 16)
+            {
+              /* Replace Composite type of size greater than 16 with a
+                  pointer.  */
+              avalue[i] = *(void **)
+              allocate_int_to_reg_or_stack (context, &state, stack,
+                                         sizeof (void *));
+            }
+          else
+            {
+              n = (s + 7) / 8;
+              if (state.ngrn + n <= N_X_ARG_REG)
+                {
+                  avalue[i] = &context->x[state.ngrn];
+                  state.ngrn += (unsigned int)n;
+                }
+              else
+                {
+                  state.ngrn = N_X_ARG_REG;
+                  avalue[i] = allocate_to_stack(&state, stack,
+                                           ty->alignment, s);
+                }
+            }
+          break;
 
-		  /* Eeek! We need a pointer to the structure, however the
-		     homogeneous float elements are being passed in individual
-		     registers, therefore for float and double the structure
-		     is not represented as a contiguous sequence of bytes in
-		     our saved register context.  We don't need the original
-		     contents of the register storage, so we reformat the
-		     structure into the same memory.  */
-		  avalue[i] = compress_hfa_type (reg, reg, h);
-		}
-	      else
-		{
-		  state.nsrn = N_V_ARG_REG;
-		  avalue[i] = allocate_to_stack (&state, stack,
-						 ty->alignment, s);
-		}
-	    }
-	  else if (s > 16)
-	    {
-	      /* Replace Composite type of size greater than 16 with a
-		 pointer.  */
-	      avalue[i] = *(void **)
-		allocate_int_to_reg_or_stack (context, &state, stack,
-					      sizeof (void *));
-	    }
-	  else
-	    {
-	      n = (s + 7) / 8;
-	      if (state.ngrn + n <= N_X_ARG_REG)
-		{
-		  avalue[i] = &context->x[state.ngrn];
-		  state.ngrn += n;
-		}
-	      else
-		{
-		  state.ngrn = N_X_ARG_REG;
-		  avalue[i] = allocate_to_stack (&state, stack,
-						 ty->alignment, s);
-		}
-	    }
-	  break;
-
-	default:
-	  abort();
-	}
+        default:
+          abort();
+      }
 
 #if defined (__APPLE__)
       if (i + 1 == cif->aarch64_nfixedargs)
@@ -939,3 +1005,5 @@ ffi_closure_SYSV_inner (ffi_cif *cif,
 
   return flags;
 }
+
+#endif /* (__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)*/

src/aarch64/ffitarget.h

@@ -32,6 +32,10 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.  */
 #define FFI_SIZEOF_JAVA_RAW  4
 typedef unsigned long long ffi_arg;
 typedef signed long long ffi_sarg;
+#elif defined(_M_ARM64)
+#define FFI_SIZEOF_ARG 8
+typedef unsigned long long ffi_arg;
+typedef signed long long ffi_sarg;
 #else
 typedef unsigned long ffi_arg;
 typedef signed long ffi_sarg;
@@ -65,17 +69,24 @@ typedef enum ffi_abi
 #define FFI_TRAMPOLINE_CLOSURE_OFFSET FFI_TRAMPOLINE_SIZE
 #endif
 
+#ifdef _M_ARM64
+#define FFI_EXTRA_CIF_FIELDS unsigned is_variadic
+#endif
+
 /* ---- Internal ---- */
 
 #if defined (__APPLE__)
 #define FFI_TARGET_SPECIFIC_VARIADIC
 #define FFI_EXTRA_CIF_FIELDS unsigned aarch64_nfixedargs
-#else
-/* iOS reserves x18 for the system.  Disable Go closures until
+#elif !defined(_M_ARM64)
+/* iOS and Windows reserve x18 for the system.  Disable Go closures until
    a new static chain is chosen.  */
 #define FFI_GO_CLOSURES 1
 #endif
 
+#ifndef _M_ARM64
+/* No complex type on Windows */
 #define FFI_TARGET_HAS_COMPLEX_TYPE
+#endif
 
 #endif

src/aarch64/sysv.S

@@ -19,6 +19,7 @@ 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(__aarch64__) || defined(__arm64__)
 #define LIBFFI_ASM
 #include <fficonfig.h>
 #include <ffi.h>
@@ -198,9 +199,9 @@ CNAME(ffi_call_SYSV):
 	cfi_endproc
 
 	.globl	CNAME(ffi_call_SYSV)
+	FFI_HIDDEN(CNAME(ffi_call_SYSV))
 #ifdef __ELF__
 	.type	CNAME(ffi_call_SYSV), #function
-	.hidden	CNAME(ffi_call_SYSV)
 	.size CNAME(ffi_call_SYSV), .-CNAME(ffi_call_SYSV)
 #endif
 
@@ -237,9 +238,9 @@ CNAME(ffi_closure_SYSV_V):
 	cfi_endproc
 
 	.globl	CNAME(ffi_closure_SYSV_V)
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV_V))
 #ifdef __ELF__
 	.type	CNAME(ffi_closure_SYSV_V), #function
-	.hidden	CNAME(ffi_closure_SYSV_V)
 	.size	CNAME(ffi_closure_SYSV_V), . - CNAME(ffi_closure_SYSV_V)
 #endif
 
@@ -349,9 +350,9 @@ CNAME(ffi_closure_SYSV):
 	cfi_endproc
 
 	.globl	CNAME(ffi_closure_SYSV)
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV))
 #ifdef __ELF__
 	.type	CNAME(ffi_closure_SYSV), #function
-	.hidden	CNAME(ffi_closure_SYSV)
 	.size	CNAME(ffi_closure_SYSV), . - CNAME(ffi_closure_SYSV)
 #endif
 
@@ -369,9 +370,9 @@ CNAME(ffi_closure_trampoline_table_page):
     .endr
 
     .globl CNAME(ffi_closure_trampoline_table_page)
+    FFI_HIDDEN(CNAME(ffi_closure_trampoline_table_page))
     #ifdef __ELF__
     	.type	CNAME(ffi_closure_trampoline_table_page), #function
-    	.hidden	CNAME(ffi_closure_trampoline_table_page)
     	.size	CNAME(ffi_closure_trampoline_table_page), . - CNAME(ffi_closure_trampoline_table_page)
     #endif
 #endif
@@ -396,9 +397,9 @@ CNAME(ffi_go_closure_SYSV_V):
 	cfi_endproc
 
 	.globl	CNAME(ffi_go_closure_SYSV_V)
+	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV_V))
 #ifdef __ELF__
 	.type	CNAME(ffi_go_closure_SYSV_V), #function
-	.hidden	CNAME(ffi_go_closure_SYSV_V)
 	.size	CNAME(ffi_go_closure_SYSV_V), . - CNAME(ffi_go_closure_SYSV_V)
 #endif
 
@@ -425,12 +426,13 @@ CNAME(ffi_go_closure_SYSV):
 	cfi_endproc
 
 	.globl	CNAME(ffi_go_closure_SYSV)
+	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV))
 #ifdef __ELF__
 	.type	CNAME(ffi_go_closure_SYSV), #function
-	.hidden	CNAME(ffi_go_closure_SYSV)
 	.size	CNAME(ffi_go_closure_SYSV), . - CNAME(ffi_go_closure_SYSV)
 #endif
 #endif /* FFI_GO_CLOSURES */
+#endif /* __arm64__ */
 
 #if defined __ELF__ && defined __linux__
 	.section .note.GNU-stack,"",%progbits

src/aarch64/win64_armasm.S

@@ -0,0 +1,506 @@
+/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
+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>
+#include <ffi_cfi.h>
+#include "internal.h"
+
+	OPT	2 /*disable listing */
+/* For some macros to add unwind information */
+#include "ksarm64.h"
+	OPT	1 /*re-enable listing */
+
+#define BE(X)	0
+#define PTR_REG(n)      x##n
+#define PTR_SIZE	8
+
+	IMPORT ffi_closure_SYSV_inner
+	EXPORT	ffi_call_SYSV
+	EXPORT	ffi_closure_SYSV_V
+	EXPORT	ffi_closure_SYSV
+	EXPORT	extend_hfa_type
+	EXPORT	compress_hfa_type
+#ifdef FFI_GO_CLOSURES
+	EXPORT	ffi_go_closure_SYSV_V
+	EXPORT	ffi_go_closure_SYSV
+#endif
+
+	TEXTAREA, ALLIGN=8
+
+/* ffi_call_SYSV
+   extern void ffi_call_SYSV (void *stack, void *frame,
+			      void (*fn)(void), void *rvalue,
+			      int flags, void *closure);
+   Therefore on entry we have:
+   x0 stack
+   x1 frame
+   x2 fn
+   x3 rvalue
+   x4 flags
+   x5 closure
+*/
+
+	NESTED_ENTRY ffi_call_SYSV_fake
+
+	/* For unwind information, Windows has to store fp and lr  */
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-32!
+
+	ALTERNATE_ENTRY ffi_call_SYSV
+	/* Use a stack frame allocated by our caller. */
+	stp	x29, x30, [x1]
+	mov	x29, x1
+	mov	sp, x0
+
+	mov	x9, x2			/* save fn */
+	mov	x8, x3			/* install structure return */
+#ifdef FFI_GO_CLOSURES
+	/*mov	x18, x5			install static chain */
+#endif
+	stp	x3, x4, [x29, #16]	/* save rvalue and flags */
+	
+	/* Load the vector argument passing registers, if necessary.  */
+	tbz	x4, #AARCH64_FLAG_ARG_V_BIT, ffi_call_SYSV_L1
+	ldp	q0, q1, [sp, #0]
+	ldp	q2, q3, [sp, #32]
+	ldp	q4, q5, [sp, #64]
+	ldp	q6, q7, [sp, #96]
+
+ffi_call_SYSV_L1
+	/* Load the core argument passing registers, including
+	   the structure return pointer.  */
+	ldp     x0, x1, [sp, #16*N_V_ARG_REG + 0]
+	ldp     x2, x3, [sp, #16*N_V_ARG_REG + 16]
+	ldp     x4, x5, [sp, #16*N_V_ARG_REG + 32]
+	ldp     x6, x7, [sp, #16*N_V_ARG_REG + 48]
+
+	/* Deallocate the context, leaving the stacked arguments.  */
+	add	sp, sp, #CALL_CONTEXT_SIZE	
+
+	blr     x9			/* call fn */
+
+	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */
+
+	/* Partially deconstruct the stack frame. */
+	mov     sp, x29 
+	ldp     x29, x30, [x29]
+
+	/* Save the return value as directed.  */
+	adr	x5, ffi_call_SYSV_return
+	and	w4, w4, #AARCH64_RET_MASK
+	add	x5, x5, x4, lsl #3
+	br	x5
+	
+	/* Note that each table entry is 2 insns, and thus 8 bytes.
+	   For integer data, note that we're storing into ffi_arg
+	   and therefore we want to extend to 64 bits; these types
+	   have two consecutive entries allocated for them.  */
+	ALIGN 4
+ffi_call_SYSV_return
+	ret				/* VOID */
+	nop
+	str	x0, [x3]		/* INT64 */
+	ret
+	stp	x0, x1, [x3]		/* INT128 */
+	ret
+	brk	#1000			/* UNUSED */
+	ret
+	brk	#1000			/* UNUSED */
+	ret
+	brk	#1000			/* UNUSED */
+	ret
+	brk	#1000			/* UNUSED */
+	ret
+	brk	#1000			/* UNUSED */
+	ret
+	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */
+	ret
+	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */
+	ret
+	stp	s0, s1, [x3]		/* S2 */
+	ret
+	str	s0, [x3]		/* S1 */
+	ret
+	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */
+	ret
+	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */
+	ret
+	stp	d0, d1, [x3]		/* D2 */
+	ret
+	str	d0, [x3]		/* D1 */
+	ret
+	str	q3, [x3, #48]		/* Q4 */
+	nop
+	str	q2, [x3, #32]		/* Q3 */
+	nop
+	stp	q0, q1, [x3]		/* Q2 */
+	ret
+	str	q0, [x3]		/* Q1 */
+	ret
+	uxtb	w0, w0			/* UINT8 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	uxth	w0, w0			/* UINT16 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	mov	w0, w0			/* UINT32 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	sxtb	x0, w0			/* SINT8 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	sxth	x0, w0			/* SINT16 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	sxtw	x0, w0			/* SINT32 */
+	str	x0, [x3]
+	ret				/* reserved */
+	nop
+	
+	
+	NESTED_END ffi_call_SYSV_fake
+	
+
+/* ffi_closure_SYSV
+   Closure invocation glue. This is the low level code invoked directly by
+   the closure trampoline to setup and call a closure.
+   On entry x17 points to a struct ffi_closure, x16 has been clobbered
+   all other registers are preserved.
+   We allocate a call context and save the argument passing registers,
+   then invoked the generic C ffi_closure_SYSV_inner() function to do all
+   the real work, on return we load the result passing registers back from
+   the call context.
+*/
+
+#define ffi_closure_SYSV_FS (8*2 + CALL_CONTEXT_SIZE + 64)
+
+	NESTED_ENTRY	ffi_closure_SYSV_V
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
+
+	/* Save the argument passing vector registers.  */
+	stp	q0, q1, [sp, #16 + 0]
+	stp	q2, q3, [sp, #16 + 32]
+	stp	q4, q5, [sp, #16 + 64]
+	stp	q6, q7, [sp, #16 + 96]
+
+	b	ffi_closure_SYSV_save_argument
+	NESTED_END	ffi_closure_SYSV_V
+
+	NESTED_ENTRY	ffi_closure_SYSV
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
+
+ffi_closure_SYSV_save_argument
+	/* Save the argument passing core registers.  */
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
+
+	/* Load ffi_closure_inner arguments.  */
+	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */
+	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */
+
+do_closure
+	add	x3, sp, #16							/* load context */
+	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */
+	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */
+	mov	x6, x8					/* load struct_rval */
+
+	bl	ffi_closure_SYSV_inner
+
+	/* Load the return value as directed.  */
+	adr	x1, ffi_closure_SYSV_return_base
+	and	w0, w0, #AARCH64_RET_MASK
+	add	x1, x1, x0, lsl #3
+	add	x3, sp, #16+CALL_CONTEXT_SIZE
+	br	x1
+
+	/* Note that each table entry is 2 insns, and thus 8 bytes.  */
+	ALIGN	8
+ffi_closure_SYSV_return_base
+	b	ffi_closure_SYSV_epilog			/* VOID */
+	nop
+	ldr	x0, [x3]		/* INT64 */
+	b	ffi_closure_SYSV_epilog
+	ldp	x0, x1, [x3]		/* INT128 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* UNUSED */
+	nop
+	brk	#1000			/* UNUSED */
+	nop
+	brk	#1000			/* UNUSED */
+	nop
+	brk	#1000			/* UNUSED */
+	nop
+	brk	#1000			/* UNUSED */
+	nop
+	ldr	s3, [x3, #12]		/* S4 */
+	nop
+	ldr	s2, [x3, #8]		/* S3 */
+	nop
+	ldp	s0, s1, [x3]		/* S2 */
+	b	ffi_closure_SYSV_epilog
+	ldr	s0, [x3]		/* S1 */
+	b	ffi_closure_SYSV_epilog
+	ldr	d3, [x3, #24]		/* D4 */
+	nop
+	ldr	d2, [x3, #16]		/* D3 */
+	nop
+	ldp	d0, d1, [x3]		/* D2 */
+	b	ffi_closure_SYSV_epilog
+	ldr	d0, [x3]		/* D1 */
+	b	ffi_closure_SYSV_epilog
+	ldr	q3, [x3, #48]		/* Q4 */
+	nop
+	ldr	q2, [x3, #32]		/* Q3 */
+	nop
+	ldp	q0, q1, [x3]		/* Q2 */
+	b	ffi_closure_SYSV_epilog
+	ldr	q0, [x3]		/* Q1 */
+	b	ffi_closure_SYSV_epilog
+	ldrb	w0, [x3, #BE(7)]	/* UINT8 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* reserved */
+	nop
+	ldrh	w0, [x3, #BE(6)]	/* UINT16 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* reserved */
+	nop
+	ldr	w0, [x3, #BE(4)]	/* UINT32 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* reserved */
+	nop
+	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* reserved */
+	nop
+	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */
+	b	ffi_closure_SYSV_epilog
+	brk	#1000			/* reserved */
+	nop
+	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */
+	nop
+					/* reserved */
+
+ffi_closure_SYSV_epilog
+	EPILOG_RESTORE_REG_PAIR	x29, x30, #ffi_closure_SYSV_FS!
+	EPILOG_RETURN
+	NESTED_END	ffi_closure_SYSV
+
+
+#ifdef FFI_GO_CLOSURES
+	NESTED_ENTRY	ffi_go_closure_SYSV_V
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
+
+	/* Save the argument passing vector registers.  */
+	stp	q0, q1, [sp, #16 + 0]
+	stp	q2, q3, [sp, #16 + 32]
+	stp	q4, q5, [sp, #16 + 64]
+	stp	q6, q7, [sp, #16 + 96]
+	b	ffi_go_closure_SYSV_save_argument
+	NESTED_END	ffi_go_closure_SYSV_V
+
+	NESTED_ENTRY	ffi_go_closure_SYSV
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
+
+ffi_go_closure_SYSV_save_argument
+	/* Save the argument passing core registers.  */
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
+
+	/* Load ffi_closure_inner arguments.  */
+	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */
+	mov	x2, x18					/* load user_data */
+	b	do_closure
+	NESTED_END	ffi_go_closure_SYSV
+
+#endif /* FFI_GO_CLOSURES */
+
+
+/* void extend_hfa_type (void *dest, void *src, int h) */
+
+	LEAF_ENTRY	extend_hfa_type
+
+	adr	x3, extend_hfa_type_jump_base
+	and	w2, w2, #AARCH64_RET_MASK
+	sub	x2, x2, #AARCH64_RET_S4
+	add	x3, x3, x2, lsl #4
+	br	x3
+
+	ALIGN	4
+extend_hfa_type_jump_base
+	ldp	s16, s17, [x1]		/* S4 */
+	ldp	s18, s19, [x1, #8]
+	b	extend_hfa_type_store_4
+	nop
+
+	ldp	s16, s17, [x1]		/* S3 */
+	ldr	s18, [x1, #8]
+	b	extend_hfa_type_store_3
+	nop
+
+	ldp	s16, s17, [x1]		/* S2 */
+	b	extend_hfa_type_store_2
+	nop
+	nop
+
+	ldr	s16, [x1]		/* S1 */
+	b	extend_hfa_type_store_1
+	nop
+	nop
+
+	ldp	d16, d17, [x1]		/* D4 */
+	ldp	d18, d19, [x1, #16]
+	b       extend_hfa_type_store_4
+	nop
+
+	ldp     d16, d17, [x1]		/* D3 */
+	ldr     d18, [x1, #16]
+	b	extend_hfa_type_store_3
+	nop
+
+	ldp	d16, d17, [x1]		/* D2 */
+	b	extend_hfa_type_store_2
+	nop
+	nop
+
+	ldr	d16, [x1]		/* D1 */
+	b	extend_hfa_type_store_1
+	nop
+	nop
+
+	ldp	q16, q17, [x1]		/* Q4 */
+	ldp	q18, q19, [x1, #16]
+	b	extend_hfa_type_store_4
+	nop
+
+	ldp	q16, q17, [x1]		/* Q3 */
+	ldr	q18, [x1, #16]
+	b	extend_hfa_type_store_3
+	nop
+
+	ldp	q16, q17, [x1]		/* Q2 */
+	b	extend_hfa_type_store_2
+	nop
+	nop
+
+	ldr	q16, [x1]		/* Q1 */
+	b	extend_hfa_type_store_1
+
+extend_hfa_type_store_4
+	str	q19, [x0, #48]
+extend_hfa_type_store_3
+	str	q18, [x0, #32]
+extend_hfa_type_store_2
+	str	q17, [x0, #16]
+extend_hfa_type_store_1
+	str	q16, [x0]
+	ret
+
+	LEAF_END	extend_hfa_type
+
+
+/* void compress_hfa_type (void *dest, void *reg, int h) */
+
+	LEAF_ENTRY	compress_hfa_type
+
+	adr	x3, compress_hfa_type_jump_base
+	and	w2, w2, #AARCH64_RET_MASK
+	sub	x2, x2, #AARCH64_RET_S4
+	add	x3, x3, x2, lsl #4
+	br	x3
+
+	ALIGN	4
+compress_hfa_type_jump_base
+	ldp	q16, q17, [x1]		/* S4 */
+	ldp	q18, q19, [x1, #32]
+	st4	{ v16.s, v17.s, v18.s, v19.s }[0], [x0]
+	ret
+
+	ldp	q16, q17, [x1]		/* S3 */
+	ldr	q18, [x1, #32]
+	st3	{ v16.s, v17.s, v18.s }[0], [x0]
+	ret
+
+	ldp	q16, q17, [x1]		/* S2 */
+	st2	{ v16.s, v17.s }[0], [x0]
+	ret
+	nop
+
+	ldr	q16, [x1]		/* S1 */
+	st1	{ v16.s }[0], [x0]
+	ret
+	nop
+
+	ldp	q16, q17, [x1]		/* D4 */
+	ldp	q18, q19, [x1, #32]
+	st4	{ v16.d, v17.d, v18.d, v19.d }[0], [x0]
+	ret
+
+	ldp	q16, q17, [x1]		/* D3 */
+	ldr	q18, [x1, #32]
+	st3	{ v16.d, v17.d, v18.d }[0], [x0]
+	ret
+
+	ldp	q16, q17, [x1]		/* D2 */
+	st2	{ v16.d, v17.d }[0], [x0]
+	ret
+	nop
+
+	ldr	q16, [x1]		/* D1 */
+	st1	{ v16.d }[0], [x0]
+	ret
+	nop
+
+	ldp	q16, q17, [x1]		/* Q4 */
+	ldp	q18, q19, [x1, #32]
+	b	compress_hfa_type_store_q4
+	nop
+
+	ldp	q16, q17, [x1]		/* Q3 */
+	ldr	q18, [x1, #32]
+	b	compress_hfa_type_store_q3
+	nop
+
+	ldp	q16, q17, [x1]		/* Q2 */
+	stp	q16, q17, [x0]
+	ret
+	nop
+
+	ldr	q16, [x1]		/* Q1 */
+	str	q16, [x0]
+	ret
+
+compress_hfa_type_store_q4
+	str	q19, [x0, #48]
+compress_hfa_type_store_q3
+	str	q18, [x0, #32]
+	stp	q16, q17, [x0]
+	ret
+
+	LEAF_END	compress_hfa_type
+
+	END

src/arm/ffi.c

@@ -28,6 +28,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
+#if defined(__arm__) || defined(_M_ARM)
 #include <fficonfig.h>
 #include <ffi.h>
 #include <ffi_common.h>
@@ -35,6 +36,11 @@
 #include <stdlib.h>
 #include "internal.h"
 
+#if defined(_MSC_VER) && defined(_M_ARM)
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+#endif
+
 #if FFI_EXEC_TRAMPOLINE_TABLE
 
 #ifdef __MACH__
@@ -42,7 +48,11 @@
 #endif
 
 #else
+#ifndef _M_ARM
 extern unsigned int ffi_arm_trampoline[2] FFI_HIDDEN;
+#else
+extern unsigned int ffi_arm_trampoline[3] FFI_HIDDEN;
+#endif
 #endif
 
 /* Forward declares. */
@@ -88,10 +98,20 @@ ffi_put_arg (ffi_type *ty, void *src, void *dst)
     case FFI_TYPE_SINT32:
     case FFI_TYPE_UINT32:
     case FFI_TYPE_POINTER:
+#ifndef _MSC_VER
     case FFI_TYPE_FLOAT:
+#endif
       *(UINT32 *)dst = *(UINT32 *)src;
       break;
 
+#ifdef _MSC_VER
+    // casting a float* to a UINT32* doesn't work on Windows
+    case FFI_TYPE_FLOAT:
+        *(uintptr_t *)dst = 0;
+        *(float *)dst = *(float *)src;
+        break;
+#endif
+
     case FFI_TYPE_SINT64:
     case FFI_TYPE_UINT64:
     case FFI_TYPE_DOUBLE:
@@ -210,7 +230,7 @@ ffi_prep_args_VFP (ffi_cif *cif, int flags, void *rvalue,
 }
 
 /* Perform machine dependent cif processing */
-ffi_status
+ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep (ffi_cif *cif)
 {
   int flags = 0, cabi = cif->abi;
@@ -301,7 +321,7 @@ ffi_prep_cif_machdep (ffi_cif *cif)
 }
 
 /* Perform machine dependent cif processing for variadic calls */
-ffi_status
+ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep_var (ffi_cif * cif,
 			  unsigned int nfixedargs, unsigned int ntotalargs)
 {
@@ -571,15 +591,28 @@ ffi_prep_closure_loc (ffi_closure * closure,
   config[0] = closure;
   config[1] = closure_func;
 #else
-  memcpy (closure->tramp, ffi_arm_trampoline, 8);
+
+#ifndef _M_ARM
+  memcpy(closure->tramp, ffi_arm_trampoline, 8);
+#else
+  // cast away function type so MSVC doesn't set the lower bit of the function pointer
+  memcpy(closure->tramp, (void*)((uintptr_t)ffi_arm_trampoline & 0xFFFFFFFE), FFI_TRAMPOLINE_CLOSURE_OFFSET);
+#endif
+
 #if defined (__QNX__)
   msync(closure->tramp, 8, 0x1000000);	/* clear data map */
   msync(codeloc, 8, 0x1000000);	/* clear insn map */
+#elif defined(_MSC_VER)
+  FlushInstructionCache(GetCurrentProcess(), closure->tramp, FFI_TRAMPOLINE_SIZE);
 #else
   __clear_cache(closure->tramp, closure->tramp + 8);	/* clear data map */
   __clear_cache(codeloc, codeloc + 8);			/* clear insn map */
 #endif
+#ifdef _M_ARM
+  *(void(**)(void))(closure->tramp + FFI_TRAMPOLINE_CLOSURE_FUNCTION) = closure_func;
+#else
   *(void (**)(void))(closure->tramp + 8) = closure_func;
+#endif
 #endif
 
   closure->cif = cif;
@@ -781,7 +814,7 @@ place_vfp_arg (ffi_cif *cif, int h)
 	}
       /* Found regs to allocate. */
       cif->vfp_used |= new_used;
-      cif->vfp_args[cif->vfp_nargs++] = reg;
+      cif->vfp_args[cif->vfp_nargs++] = (signed char)reg;
 
       /* Update vfp_reg_free. */
       if (cif->vfp_used & (1 << cif->vfp_reg_free))
@@ -803,7 +836,7 @@ place_vfp_arg (ffi_cif *cif, int h)
 static void
 layout_vfp_args (ffi_cif * cif)
 {
-  int i;
+  unsigned int i;
   /* Init VFP fields */
   cif->vfp_used = 0;
   cif->vfp_nargs = 0;
@@ -817,3 +850,5 @@ layout_vfp_args (ffi_cif * cif)
 	break;
     }
 }
+
+#endif /* __arm__ or _M_ARM */

src/arm/ffitarget.h

@@ -43,7 +43,7 @@ typedef enum ffi_abi {
   FFI_SYSV,
   FFI_VFP,
   FFI_LAST_ABI,
-#ifdef __ARM_PCS_VFP
+#if defined(__ARM_PCS_VFP) || defined(_M_ARM)
   FFI_DEFAULT_ABI = FFI_VFP,
 #else
   FFI_DEFAULT_ABI = FFI_SYSV,
@@ -57,7 +57,9 @@ typedef enum ffi_abi {
   signed char vfp_args[16]			\
 
 #define FFI_TARGET_SPECIFIC_VARIADIC
+#ifndef _M_ARM
 #define FFI_TARGET_HAS_COMPLEX_TYPE
+#endif
 
 /* ---- Definitions for closures ----------------------------------------- */
 
@@ -74,8 +76,13 @@ typedef enum ffi_abi {
 #error "No trampoline table implementation"
 #endif
 
+#else
+#ifdef _MSC_VER
+#define FFI_TRAMPOLINE_SIZE 16
+#define FFI_TRAMPOLINE_CLOSURE_FUNCTION 12
 #else
 #define FFI_TRAMPOLINE_SIZE 12
+#endif
 #define FFI_TRAMPOLINE_CLOSURE_OFFSET FFI_TRAMPOLINE_SIZE
 #endif
 

src/arm/sysv.S

@@ -25,6 +25,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
+#ifdef __arm__
 #define LIBFFI_ASM
 #include <fficonfig.h>
 #include <ffi.h>
@@ -377,6 +378,7 @@ ARM_FUNC_START(ffi_arm_trampoline)
 ARM_FUNC_END(ffi_arm_trampoline)
 
 #endif /* FFI_EXEC_TRAMPOLINE_TABLE */
+#endif /* __arm__ */
 
 #if defined __ELF__ && defined __linux__
 	.section	.note.GNU-stack,"",%progbits

src/arm/sysv_msvc_arm32.S

@@ -0,0 +1,311 @@
+/* -----------------------------------------------------------------------
+   sysv.S - Copyright (c) 1998, 2008, 2011 Red Hat, Inc.
+        Copyright (c) 2011 Plausible Labs Cooperative, Inc.
+        Copyright (c) 2019 Microsoft Corporation.
+
+   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>
+#include <ffi_cfi.h>
+#include "internal.h"
+#include "ksarm.h"
+
+
+        ; 8 byte aligned AREA to support 8 byte aligned jump tables
+        MACRO
+        NESTED_ENTRY_FFI $FuncName, $AreaName, $ExceptHandler
+
+        ; compute the function's labels
+        __DeriveFunctionLabels $FuncName
+
+        ; determine the area we will put the function into
+__FuncArea   SETS    "|.text|"
+        IF "$AreaName" != ""
+__FuncArea   SETS    "$AreaName"
+        ENDIF
+
+        ; set up the exception handler itself
+__FuncExceptionHandler SETS ""
+        IF "$ExceptHandler" != ""
+__FuncExceptionHandler SETS    "|$ExceptHandler|"
+        ENDIF
+
+        ; switch to the specified area, jump tables require 8 byte alignment
+        AREA    $__FuncArea,CODE,CODEALIGN,ALIGN=3,READONLY
+
+        ; export the function name
+        __ExportProc $FuncName
+
+        ; flush any pending literal pool stuff
+        ROUT
+
+        ; reset the state of the unwind code tracking
+        __ResetUnwindState
+
+        MEND
+
+;        MACRO
+;        TABLE_ENTRY $Type, $Table
+;$Type_$Table
+;        MEND
+
+#define E(index,table) return_##index##_##table
+
+    ; r0:   stack
+    ; r1:   frame
+    ; r2:   fn
+    ; r3:   vfp_used
+
+    ; fake entry point exists only to generate exists only to 
+    ; generate .pdata for exception unwinding
+    NESTED_ENTRY_FFI ffi_call_VFP_fake
+    PROLOG_PUSH  {r11, lr}          ; save fp and lr for unwind
+
+    ALTERNATE_ENTRY ffi_call_VFP
+    cmp    r3, #3                   ; load only d0 if possible
+    vldrle d0, [r0]
+    vldmgt r0, {d0-d7}
+    add    r0, r0, #64              ; discard the vfp register args
+    b ffi_call_SYSV
+    NESTED_END ffi_call_VFP_fake
+
+    ; fake entry point exists only to generate exists only to 
+    ; generate .pdata for exception unwinding
+    NESTED_ENTRY_FFI ffi_call_SYSV_fake
+    PROLOG_PUSH  {r11, lr}          ; save fp and lr for unwind
+
+    ALTERNATE_ENTRY ffi_call_SYSV
+    stm    r1, {fp, lr}
+    mov    fp, r1
+
+    mov    sp, r0                   ; install the stack pointer
+    mov    lr, r2                   ; move the fn pointer out of the way
+    ldr    ip, [fp, #16]            ; install the static chain
+    ldmia  sp!, {r0-r3}             ; move first 4 parameters in registers.
+    blx    lr                       ; call fn
+
+    ; Load r2 with the pointer to storage for the return value
+    ; Load r3 with the return type code
+    ldr    r2, [fp, #8]
+    ldr    r3, [fp, #12]
+
+    ; Deallocate the stack with the arguments.
+    mov    sp, fp
+
+    ; Store values stored in registers.
+    ALIGN 8
+    lsl     r3, #3
+    add     r3, r3, pc
+    add     r3, #8
+    mov     pc, r3
+
+
+E(ARM_TYPE_VFP_S, ffi_call)
+    ALIGN 8
+    vstr s0, [r2]
+    pop    {fp,pc}
+E(ARM_TYPE_VFP_D, ffi_call)
+    ALIGN 8
+    vstr d0, [r2]
+    pop    {fp,pc}
+E(ARM_TYPE_VFP_N, ffi_call)
+    ALIGN 8
+    vstm r2, {d0-d3}
+    pop    {fp,pc}
+E(ARM_TYPE_INT64, ffi_call)
+    ALIGN 8
+    str    r1, [r2, #4]
+    nop
+E(ARM_TYPE_INT, ffi_call)
+    ALIGN 8
+    str    r0, [r2]
+    pop    {fp,pc}
+E(ARM_TYPE_VOID, ffi_call)
+    ALIGN 8
+    pop    {fp,pc}
+    nop
+E(ARM_TYPE_STRUCT, ffi_call)
+    ALIGN 8
+    cmp r3, #ARM_TYPE_STRUCT
+    pop    {fp,pc}
+    NESTED_END ffi_call_SYSV_fake
+
+    IMPORT |ffi_closure_inner_SYSV|
+    /*
+    int ffi_closure_inner_SYSV
+    (
+        cif,        ; r0
+        fun,        ; r1
+        user_data,  ; r2
+        frame       ; r3
+    )
+    */
+
+    NESTED_ENTRY_FFI ffi_go_closure_SYSV
+    stmdb   sp!, {r0-r3}            ; save argument regs
+    ldr     r0, [ip, #4]            ; load cif
+    ldr     r1, [ip, #8]            ; load fun
+    mov     r2, ip                  ; load user_data
+    b       ffi_go_closure_SYSV_0
+    NESTED_END ffi_go_closure_SYSV
+
+    ; r3:    ffi_closure
+
+    ; fake entry point exists only to generate exists only to 
+    ; generate .pdata for exception unwinding
+    NESTED_ENTRY_FFI ffi_closure_SYSV_fake  
+    PROLOG_PUSH  {r11, lr}          ; save fp and lr for unwind
+    ALTERNATE_ENTRY ffi_closure_SYSV
+    ldmfd   sp!, {ip,r0}            ; restore fp (r0 is used for stack alignment)
+    stmdb   sp!, {r0-r3}            ; save argument regs
+
+    ldr     r0, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET]    ; ffi_closure->cif
+    ldr     r1, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET+4]  ; ffi_closure->fun
+    ldr     r2, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET+8]  ; ffi_closure->user_data
+
+    ALTERNATE_ENTRY ffi_go_closure_SYSV_0
+    add     ip, sp, #16             ; compute entry sp
+
+    sub     sp, sp, #64+32          ; allocate frame parameter (sizeof(vfp_space) = 64, sizeof(result) = 32)
+    mov     r3, sp                  ; set frame parameter
+    stmdb   sp!, {ip,lr}
+
+    bl      ffi_closure_inner_SYSV  ; call the Python closure
+
+                                    ; Load values returned in registers.
+    add     r2, sp, #64+8           ; address of closure_frame->result
+    bl      ffi_closure_ret         ; move result to correct register or memory for type
+
+    ldmfd   sp!, {ip,lr}
+    mov     sp, ip                  ; restore stack pointer
+    mov     pc, lr
+    NESTED_END ffi_closure_SYSV_fake
+
+    IMPORT |ffi_closure_inner_VFP|
+    /*
+    int ffi_closure_inner_VFP
+    (
+        cif,        ; r0
+        fun,        ; r1
+        user_data,  ; r2
+        frame       ; r3
+    )
+    */
+
+    NESTED_ENTRY_FFI ffi_go_closure_VFP
+    stmdb   sp!, {r0-r3}			; save argument regs
+    ldr	r0, [ip, #4]			; load cif
+    ldr	r1, [ip, #8]			; load fun
+    mov	r2, ip				; load user_data
+    b	ffi_go_closure_VFP_0
+    NESTED_END ffi_go_closure_VFP
+
+    ; fake entry point exists only to generate exists only to 
+    ; generate .pdata for exception unwinding
+    ; r3:    closure
+    NESTED_ENTRY_FFI ffi_closure_VFP_fake
+    PROLOG_PUSH  {r11, lr}          ; save fp and lr for unwind
+
+    ALTERNATE_ENTRY ffi_closure_VFP
+    ldmfd   sp!, {ip,r0}            ; restore fp (r0 is used for stack alignment)
+    stmdb   sp!, {r0-r3}            ; save argument regs
+
+    ldr     r0, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET]    ; load cif
+    ldr     r1, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET+4]  ; load fun
+    ldr     r2, [ip, #FFI_TRAMPOLINE_CLOSURE_OFFSET+8]  ; load user_data
+
+    ALTERNATE_ENTRY ffi_go_closure_VFP_0
+    add     ip, sp, #16             ; compute entry sp
+    sub     sp, sp, #32             ; save space for closure_frame->result
+    vstmdb  sp!, {d0-d7}            ; push closure_frame->vfp_space
+
+    mov     r3, sp                  ; save closure_frame
+    stmdb   sp!, {ip,lr}
+
+    bl      ffi_closure_inner_VFP
+
+    ; Load values returned in registers.
+    add     r2, sp, #64+8           ; load result
+    bl      ffi_closure_ret
+    ldmfd   sp!, {ip,lr}
+    mov     sp, ip                  ; restore stack pointer
+    mov     pc, lr
+    NESTED_END ffi_closure_VFP_fake
+
+/* Load values returned in registers for both closure entry points.
+   Note that we use LDM with SP in the register set.  This is deprecated
+   by ARM, but not yet unpredictable.  */
+
+    NESTED_ENTRY_FFI ffi_closure_ret
+    stmdb sp!, {fp,lr}
+
+    ALIGN 8
+    lsl     r0, #3
+    add     r0, r0, pc
+    add     r0, #8
+    mov     pc, r0
+
+E(ARM_TYPE_VFP_S, ffi_closure)
+    ALIGN 8
+    vldr s0, [r2]
+    b call_epilogue
+E(ARM_TYPE_VFP_D, ffi_closure)
+    ALIGN 8
+    vldr d0, [r2]
+    b call_epilogue
+E(ARM_TYPE_VFP_N, ffi_closure)
+    ALIGN 8
+    vldm r2, {d0-d3}
+    b call_epilogue
+E(ARM_TYPE_INT64, ffi_closure)
+    ALIGN 8
+    ldr    r1, [r2, #4]
+    nop
+E(ARM_TYPE_INT, ffi_closure)
+    ALIGN 8
+    ldr    r0, [r2]
+    b call_epilogue
+E(ARM_TYPE_VOID, ffi_closure)
+    ALIGN 8
+    b call_epilogue
+    nop
+E(ARM_TYPE_STRUCT, ffi_closure)
+    ALIGN 8
+    b call_epilogue
+call_epilogue
+    ldmfd sp!, {fp,pc}
+    NESTED_END ffi_closure_ret
+
+    AREA |.trampoline|, DATA, THUMB, READONLY
+    EXPORT |ffi_arm_trampoline|
+|ffi_arm_trampoline| DATA
+thisproc    adr    ip, thisproc
+            stmdb  sp!, {ip, r0}
+            ldr    pc, [pc, #0]
+            DCD    0
+            ;ENDP
+
+    END

src/closures.c

@@ -122,7 +122,7 @@ ffi_closure_free (void *ptr)
 #  define FFI_MMAP_EXEC_WRIT 1
 #  define HAVE_MNTENT 1
 # endif
-# if defined(X86_WIN32) || defined(X86_WIN64) || defined(__OS2__)
+# if defined(X86_WIN32) || defined(X86_WIN64) || defined(_M_ARM64) || defined(__OS2__)
 /* Windows systems may have Data Execution Protection (DEP) enabled, 
    which requires the use of VirtualMalloc/VirtualFree to alloc/free
    executable memory. */
@@ -172,7 +172,7 @@ struct ffi_trampoline_table
 
 struct ffi_trampoline_table_entry
 {
-  void *(*trampoline) ();
+  void *(*trampoline) (void);
   ffi_trampoline_table_entry *next;
 };
 
@@ -385,7 +385,7 @@ ffi_closure_free (void *ptr)
 #endif
 #include <string.h>
 #include <stdio.h>
-#if !defined(X86_WIN32) && !defined(X86_WIN64)
+#if !defined(X86_WIN32) && !defined(X86_WIN64) && !defined(_M_ARM64)
 #ifdef HAVE_MNTENT
 #include <mntent.h>
 #endif /* HAVE_MNTENT */
@@ -511,7 +511,7 @@ static int dlmalloc_trim(size_t) MAYBE_UNUSED;
 static size_t dlmalloc_usable_size(void*) MAYBE_UNUSED;
 static void dlmalloc_stats(void) MAYBE_UNUSED;
 
-#if !(defined(X86_WIN32) || defined(X86_WIN64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX)
+#if !(defined(X86_WIN32) || defined(X86_WIN64) || defined(_M_ARM64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX)
 /* Use these for mmap and munmap within dlmalloc.c.  */
 static void *dlmmap(void *, size_t, int, int, int, off_t);
 static int dlmunmap(void *, size_t);
@@ -525,7 +525,7 @@ static int dlmunmap(void *, size_t);
 #undef mmap
 #undef munmap
 
-#if !(defined(X86_WIN32) || defined(X86_WIN64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX)
+#if !(defined(X86_WIN32) || defined(X86_WIN64) || defined(_M_ARM64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX)
 
 /* A mutex used to synchronize access to *exec* variables in this file.  */
 static pthread_mutex_t open_temp_exec_file_mutex = PTHREAD_MUTEX_INITIALIZER;
@@ -789,7 +789,13 @@ dlmmap_locked (void *start, size_t length, int prot, int flags, off_t offset)
 	  close (execfd);
 	  goto retry_open;
 	}
-      ftruncate (execfd, offset);
+      if (ftruncate (execfd, offset) != 0)
+      {
+        /* Fixme : Error logs can be added here. Returning an error for
+         * ftruncte() will not add any advantage as it is being
+         * validating in the error case. */
+      }
+
       return MFAIL;
     }
   else if (!offset
@@ -801,7 +807,12 @@ dlmmap_locked (void *start, size_t length, int prot, int flags, off_t offset)
   if (start == MFAIL)
     {
       munmap (ptr, length);
-      ftruncate (execfd, offset);
+      if (ftruncate (execfd, offset) != 0)
+      {
+        /* Fixme : Error logs can be added here. Returning an error for
+         * ftruncte() will not add any advantage as it is being
+         * validating in the error case. */
+      }
       return start;
     }
 
@@ -896,7 +907,7 @@ segment_holding_code (mstate m, char* addr)
 }
 #endif
 
-#endif /* !(defined(X86_WIN32) || defined(X86_WIN64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX) */
+#endif /* !(defined(X86_WIN32) || defined(X86_WIN64) || defined(_M_ARM64) || defined(__OS2__)) || defined (__CYGWIN__) || defined(__INTERIX) */
 
 /* Allocate a chunk of memory with the given size.  Returns a pointer
    to the writable address, and sets *CODE to the executable
@@ -921,6 +932,13 @@ ffi_closure_alloc (size_t size, void **code)
   return ptr;
 }
 
+void *
+ffi_data_to_code_pointer (void *data)
+{
+  msegmentptr seg = segment_holding (gm, data);
+  return add_segment_exec_offset (data, seg);
+}
+
 /* Release a chunk of memory allocated with ffi_closure_alloc.  If
    FFI_CLOSURE_FREE_CODE is nonzero, the given address can be the
    writable or the executable address given.  Otherwise, only the
@@ -960,6 +978,12 @@ ffi_closure_free (void *ptr)
   free (ptr);
 }
 
+void *
+ffi_data_to_code_pointer (void *data)
+{
+  return data;
+}
+
 # endif /* ! FFI_MMAP_EXEC_WRIT */
 #endif /* FFI_CLOSURES */
 

src/pa/linux.S

@@ -297,10 +297,18 @@ ffi_closure_pa32:
 .LSCIE1:
 	.word   0x0     ;# CIE Identifier Tag
 	.byte   0x1     ;# CIE Version
+#ifdef __PIC__
+	.ascii  "zR\0"  ;# CIE Augmentation: 'z' - data, 'R' - DW_EH_PE_... data
+#else
 	.ascii "\0"     ;# CIE Augmentation
+#endif
 	.uleb128 0x1    ;# CIE Code Alignment Factor
 	.sleb128 4      ;# CIE Data Alignment Factor
 	.byte   0x2     ;# CIE RA Column
+#ifdef __PIC__
+	.uleb128 0x1    ;# Augmentation size
+	.byte	0x1b    ;# FDE Encoding (DW_EH_PE_pcrel|DW_EH_PE_sdata4)
+#endif
 	.byte   0xc     ;# DW_CFA_def_cfa
 	.uleb128 0x1e
 	.uleb128 0x0
@@ -310,9 +318,15 @@ ffi_closure_pa32:
 	.word   .LEFDE1-.LASFDE1        ;# FDE Length
 .LASFDE1:
 	.word   .LASFDE1-.Lframe1       ;# FDE CIE offset
-	.word   .LFB1   ;# FDE initial location
+#ifdef __PIC__
+	.word	.LFB1-. ;# FDE initial location
+#else
+	.word	.LFB1   ;# FDE initial location
+#endif
 	.word   .LFE1-.LFB1     ;# FDE address range
-
+#ifdef __PIC__
+	.uleb128 0x0	;# Augmentation size: no data
+#endif
 	.byte   0x4     ;# DW_CFA_advance_loc4
 	.word   .LCFI11-.LFB1
 	.byte	0x83	;# DW_CFA_offset, column 0x3
@@ -338,8 +352,15 @@ ffi_closure_pa32:
 	.word   .LEFDE2-.LASFDE2        ;# FDE Length
 .LASFDE2:
 	.word   .LASFDE2-.Lframe1       ;# FDE CIE offset
+#ifdef __PIC__
+	.word   .LFB2-. ;# FDE initial location
+#else
 	.word   .LFB2   ;# FDE initial location
+#endif
 	.word   .LFE2-.LFB2     ;# FDE address range
+#ifdef __PIC__
+	.uleb128 0x0	;# Augmentation size: no data
+#endif
 	.byte   0x4     ;# DW_CFA_advance_loc4
 	.word   .LCFI21-.LFB2
 	.byte   0x83    ;# DW_CFA_offset, column 0x3

src/powerpc/ffi.c

@@ -121,8 +121,9 @@ ffi_call_int (ffi_cif *cif,
 # endif
 	/* The SYSV ABI returns a structure of up to 8 bytes in size
 	   left-padded in r3/r4, and the ELFv2 ABI similarly returns a
-	   structure of up to 8 bytes in size left-padded in r3.  */
-	if (rsize <= 8)
+	   structure of up to 8 bytes in size left-padded in r3. But
+	   note that a structure of a single float is not paddded.  */
+	if (rsize <= 8 && (cif->flags & FLAG_RETURNS_FP) == 0)
 	  memcpy (rvalue, (char *) smst_buffer + 8 - rsize, rsize);
 	else
 #endif

src/powerpc/ffi_linux64.c

@@ -63,10 +63,30 @@ ffi_prep_types_linux64 (ffi_abi abi)
 
 
 static unsigned int
-discover_homogeneous_aggregate (const ffi_type *t, unsigned int *elnum)
+discover_homogeneous_aggregate (ffi_abi abi,
+                                const ffi_type *t,
+                                unsigned int *elnum)
 {
   switch (t->type)
     {
+#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE
+    case FFI_TYPE_LONGDOUBLE:
+      /* 64-bit long doubles are equivalent to doubles. */
+      if ((abi & FFI_LINUX_LONG_DOUBLE_128) == 0)
+        {
+          *elnum = 1;
+          return FFI_TYPE_DOUBLE;
+        }
+      /* IBM extended precision values use unaligned pairs
+         of FPRs, but according to the ABI must be considered
+         distinct from doubles. They are also limited to a
+         maximum of four members in a homogeneous aggregate. */
+      else
+        {
+          *elnum = 2;
+          return FFI_TYPE_LONGDOUBLE;
+        }
+#endif
     case FFI_TYPE_FLOAT:
     case FFI_TYPE_DOUBLE:
       *elnum = 1;
@@ -79,7 +99,7 @@ discover_homogeneous_aggregate (const ffi_type *t, unsigned int *elnum)
 	while (*el)
 	  {
 	    unsigned int el_elt, el_elnum = 0;
-	    el_elt = discover_homogeneous_aggregate (*el, &el_elnum);
+	    el_elt = discover_homogeneous_aggregate (abi, *el, &el_elnum);
 	    if (el_elt == 0
 		|| (base_elt && base_elt != el_elt))
 	      return 0;
@@ -112,7 +132,7 @@ ffi_prep_cif_linux64_core (ffi_cif *cif)
   unsigned bytes;
   unsigned i, fparg_count = 0, intarg_count = 0;
   unsigned flags = cif->flags;
-  unsigned int elt, elnum;
+  unsigned elt, elnum, rtype;
 
 #if FFI_TYPE_LONGDOUBLE == FFI_TYPE_DOUBLE
   /* If compiled without long double support..  */
@@ -138,7 +158,11 @@ ffi_prep_cif_linux64_core (ffi_cif *cif)
 #endif
 
   /* Return value handling.  */
-  switch (cif->rtype->type)
+  rtype = cif->rtype->type;
+#if _CALL_ELF == 2
+homogeneous:
+#endif
+  switch (rtype)
     {
 #if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE
     case FFI_TYPE_LONGDOUBLE:
@@ -164,19 +188,18 @@ ffi_prep_cif_linux64_core (ffi_cif *cif)
 
     case FFI_TYPE_STRUCT:
 #if _CALL_ELF == 2
-      elt = discover_homogeneous_aggregate (cif->rtype, &elnum);
+      elt = discover_homogeneous_aggregate (cif->abi, cif->rtype, &elnum);
       if (elt)
-	{
-	  if (elt == FFI_TYPE_DOUBLE)
-	    flags |= FLAG_RETURNS_64BITS;
-	  flags |= FLAG_RETURNS_FP | FLAG_RETURNS_SMST;
-	  break;
-	}
+        {
+          flags |= FLAG_RETURNS_SMST;
+          rtype = elt;
+          goto homogeneous;
+        }
       if (cif->rtype->size <= 16)
-	{
-	  flags |= FLAG_RETURNS_SMST;
-	  break;
-	}
+        {
+          flags |= FLAG_RETURNS_SMST;
+          break;
+        }
 #endif
       intarg_count++;
       flags |= FLAG_RETVAL_REFERENCE;
@@ -224,7 +247,7 @@ ffi_prep_cif_linux64_core (ffi_cif *cif)
 		intarg_count = FFI_ALIGN (intarg_count, align);
 	    }
 	  intarg_count += ((*ptr)->size + 7) / 8;
-	  elt = discover_homogeneous_aggregate (*ptr, &elnum);
+	  elt = discover_homogeneous_aggregate (cif->abi, *ptr, &elnum);
 	  if (elt)
 	    {
 	      fparg_count += elnum;
@@ -391,7 +414,7 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack)
   valp rest;
   valp next_arg;
 
-  /* 'fpr_base' points at the space for fpr3, and grows upwards as
+  /* 'fpr_base' points at the space for f1, and grows upwards as
      we use FPR registers.  */
   valp fpr_base;
   unsigned int fparg_count;
@@ -492,7 +515,9 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack)
 	  /* Fall through.  */
 #endif
 	case FFI_TYPE_DOUBLE:
+#if _CALL_ELF != 2
 	do_double:
+#endif
 	  double_tmp = **p_argv.d;
 	  if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs)
 	    {
@@ -511,7 +536,9 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack)
 	  break;
 
 	case FFI_TYPE_FLOAT:
+#if _CALL_ELF != 2
 	do_float:
+#endif
 	  double_tmp = **p_argv.f;
 	  if (fparg_count < NUM_FPR_ARG_REGISTERS64 && i < nfixedargs)
 	    {
@@ -548,9 +575,13 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack)
 	      if (align > 16)
 		align = 16;
 	      if (align > 1)
-		next_arg.p = FFI_ALIGN (next_arg.p, align);
+                {
+                  next_arg.p = FFI_ALIGN (next_arg.p, align);
+                  if (next_arg.ul == gpr_end.ul)
+                    next_arg.ul = rest.ul;
+                }
 	    }
-	  elt = discover_homogeneous_aggregate (*ptr, &elnum);
+	  elt = discover_homogeneous_aggregate (ecif->cif->abi, *ptr, &elnum);
 	  if (elt)
 	    {
 #if _CALL_ELF == 2
@@ -576,11 +607,9 @@ ffi_prep_args64 (extended_cif *ecif, unsigned long *const stack)
 		      fparg_count++;
 		    }
 		  while (--elnum != 0);
-		  if ((next_arg.p & 3) != 0)
-		    {
-		      if (++next_arg.f == gpr_end.f)
-			next_arg.f = rest.f;
-		    }
+		  if ((next_arg.p & 7) != 0)
+                    if (++next_arg.f == gpr_end.f)
+                      next_arg.f = rest.f;
 		}
 	      else
 		do
@@ -813,7 +842,7 @@ ffi_closure_helper_LINUX64 (ffi_cif *cif,
 	      if (align > 1)
 		pst = (unsigned long *) FFI_ALIGN ((size_t) pst, align);
 	    }
-	  elt = discover_homogeneous_aggregate (arg_types[i], &elnum);
+	  elt = discover_homogeneous_aggregate (cif->abi, arg_types[i], &elnum);
 	  if (elt)
 	    {
 #if _CALL_ELF == 2
@@ -915,7 +944,9 @@ ffi_closure_helper_LINUX64 (ffi_cif *cif,
 	  /* Fall through.  */
 #endif
 	case FFI_TYPE_DOUBLE:
+#if _CALL_ELF != 2
 	do_double:
+#endif
 	  /* On the outgoing stack all values are aligned to 8 */
 	  /* there are 13 64bit floating point registers */
 
@@ -930,7 +961,9 @@ ffi_closure_helper_LINUX64 (ffi_cif *cif,
 	  break;
 
 	case FFI_TYPE_FLOAT:
+#if _CALL_ELF != 2
 	do_float:
+#endif
 	  if (pfr < end_pfr && i < nfixedargs)
 	    {
 	      /* Float values are stored as doubles in the

src/powerpc/linux64_closure.S

@@ -143,7 +143,7 @@ ffi_closure_LINUX64:
 	stfd	%f12, -104+(11*8)(%r1)
 	stfd	%f13, -104+(12*8)(%r1)
 
-	# load up the pointer to the saved fpr registers */
+	# load up the pointer to the saved fpr registers
 	addi	%r8, %r1, -104
 
 	# load up the pointer to the result storage

src/prep_cif.c

@@ -129,7 +129,9 @@ ffi_status FFI_HIDDEN ffi_prep_cif_core(ffi_cif *cif, ffi_abi abi,
   cif->rtype = rtype;
 
   cif->flags = 0;
-
+#ifdef _M_ARM64
+  cif->is_variadic = isvariadic;
+#endif
 #if HAVE_LONG_DOUBLE_VARIANT
   ffi_prep_types (abi);
 #endif
@@ -199,7 +201,7 @@ ffi_status FFI_HIDDEN ffi_prep_cif_core(ffi_cif *cif, ffi_abi abi,
 	    bytes = 6*4;
 #endif
 
-	  bytes += STACK_ARG_SIZE((*ptr)->size);
+	  bytes += (unsigned int)STACK_ARG_SIZE((*ptr)->size);
 	}
 #endif
     }

src/x86/ffi.c

@@ -29,7 +29,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
-#ifndef __x86_64__
+#if defined(__i386__) || defined(_M_IX86)
 #include <ffi.h>
 #include <ffi_common.h>
 #include <stdint.h>
@@ -756,4 +756,4 @@ ffi_raw_call(ffi_cif *cif, void (*fn)(void), void *rvalue, ffi_raw *avalue)
   ffi_call_i386 (frame, stack);
 }
 #endif /* !FFI_NO_RAW_API */
-#endif /* !__x86_64__ */
+#endif /* __i386__ */

src/x86/ffi64.c

@@ -282,7 +282,7 @@ classify_argument (ffi_type *type, enum x86_64_reg_class classes[],
 
 	    /* The X86_64_SSEUP_CLASS should be always preceded by
 	       X86_64_SSE_CLASS or X86_64_SSEUP_CLASS.  */
-	    if (classes[i] == X86_64_SSEUP_CLASS
+	    if (i > 1 && classes[i] == X86_64_SSEUP_CLASS
 		&& classes[i - 1] != X86_64_SSE_CLASS
 		&& classes[i - 1] != X86_64_SSEUP_CLASS)
 	      {
@@ -293,7 +293,7 @@ classify_argument (ffi_type *type, enum x86_64_reg_class classes[],
 
 	    /*  If X86_64_X87UP_CLASS isn't preceded by X86_64_X87_CLASS,
 		everything should be passed in memory.  */
-	    if (classes[i] == X86_64_X87UP_CLASS
+	    if (i > 1 && classes[i] == X86_64_X87UP_CLASS
 		&& (classes[i - 1] != X86_64_X87_CLASS))
 	      {
 		/* The first one should never be X86_64_X87UP_CLASS.  */
@@ -394,7 +394,7 @@ extern ffi_status
 ffi_prep_cif_machdep_efi64(ffi_cif *cif);
 #endif
 
-ffi_status
+ffi_status FFI_HIDDEN
 ffi_prep_cif_machdep (ffi_cif *cif)
 {
   int gprcount, ssecount, i, avn, ngpr, nsse;
@@ -451,9 +451,11 @@ ffi_prep_cif_machdep (ffi_cif *cif)
     case FFI_TYPE_DOUBLE:
       flags = UNIX64_RET_XMM64;
       break;
+#if FFI_TYPE_LONGDOUBLE != FFI_TYPE_DOUBLE
     case FFI_TYPE_LONGDOUBLE:
       flags = UNIX64_RET_X87;
       break;
+#endif
     case FFI_TYPE_STRUCT:
       n = examine_argument (cif->rtype, classes, 1, &ngpr, &nsse);
       if (n == 0)

src/x86/ffiw64.c

@@ -25,6 +25,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
+#if defined(__x86_64__) || defined(_M_AMD64)
 #include <ffi.h>
 #include <ffi_common.h>
 #include <stdlib.h>
@@ -33,7 +34,7 @@
 #ifdef X86_WIN64
 #define EFI64(name) name
 #else
-#define EFI64(name) name##_efi64
+#define EFI64(name) FFI_HIDDEN name##_efi64
 #endif
 
 struct win64_call_frame
@@ -48,7 +49,7 @@ struct win64_call_frame
 extern void ffi_call_win64 (void *stack, struct win64_call_frame *,
 			    void *closure) FFI_HIDDEN;
 
-ffi_status
+ffi_status FFI_HIDDEN
 EFI64(ffi_prep_cif_machdep)(ffi_cif *cif)
 {
   int flags, n;
@@ -306,3 +307,5 @@ ffi_closure_win64_inner(ffi_cif *cif,
   fun (cif, rvalue, avalue, user_data);
   return flags;
 }
+
+#endif /* __x86_64__ */

src/x86/sysv.S

@@ -26,7 +26,7 @@
    DEALINGS IN THE SOFTWARE.
    ----------------------------------------------------------------------- */
 
-#ifndef __x86_64__
+#ifdef __i386__
 #ifndef _MSC_VER
 
 #define LIBFFI_ASM	
@@ -1122,7 +1122,7 @@ L(EFDE9):
 #endif /* __APPLE__ */
 
 #endif /* ifndef _MSC_VER */
-#endif /* ifndef __x86_64__ */
+#endif /* ifdef __i386__ */
 
 #if defined __ELF__ && defined __linux__
 	.section	.note.GNU-stack,"",@progbits

src/x86/win64.S

@@ -1,3 +1,4 @@
+#ifdef __x86_64__
 #define LIBFFI_ASM
 #include <fficonfig.h>
 #include <ffi.h>
@@ -42,6 +43,7 @@
 
 	.align	8
 	.globl	C(ffi_call_win64)
+	FFI_HIDDEN(C(ffi_call_win64))
 
 	SEH(.seh_proc ffi_call_win64)
 C(ffi_call_win64):
@@ -169,6 +171,7 @@ E(0b, FFI_TYPE_SMALL_STRUCT_4B)
 
 	.align	8
 	.globl	C(ffi_go_closure_win64)
+	FFI_HIDDEN(C(ffi_go_closure_win64))
 
 	SEH(.seh_proc ffi_go_closure_win64)
 C(ffi_go_closure_win64):
@@ -188,6 +191,7 @@ C(ffi_go_closure_win64):
 
 	.align	8
 	.globl	C(ffi_closure_win64)
+	FFI_HIDDEN(C(ffi_closure_win64))
 
 	SEH(.seh_proc ffi_closure_win64)
 C(ffi_closure_win64):
@@ -226,6 +230,7 @@ C(ffi_closure_win64):
 
 	cfi_endproc
 	SEH(.seh_endproc)
+#endif /* __x86_64__ */
 
 #if defined __ELF__ && defined __linux__
 	.section	.note.GNU-stack,"",@progbits

testsuite/lib/libffi.exp

@@ -1,4 +1,4 @@
-#   Copyright (C) 2003, 2005, 2008, 2009, 2010, 2011, 2014 Free Software Foundation, Inc.
+#   Copyright (C) 2003, 2005, 2008, 2009, 2010, 2011, 2014, 2019 Free Software Foundation, Inc.
 
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -480,11 +480,11 @@ proc run-many-tests { testcases extra_flags } {
     switch $compiler_vendor {
       "clang" {
 	set common "-W -Wall"
-	set optimizations { "-O0" "-O2" "-Os" }
+	set optimizations { "-O0" "-O2" }
       }
       "gnu" {
         set common "-W -Wall -Wno-psabi"
-        set optimizations { "-O0" "-O2" "-Os" "-O2 -fomit-frame-pointer" }
+        set optimizations { "-O0" "-O2" "-O2 -fomit-frame-pointer" }
       }
       default {
         # Assume we are using the vendor compiler.

testsuite/libffi.call/float2.c

@@ -33,19 +33,20 @@ int main (void)
 
   f = 3.14159;
 
-#if 1
-  /* This is ifdef'd out for now. long double support under SunOS/gcc
-     is pretty much non-existent.  You'll get the odd bus error in library
-     routines like printf().  */
+#if defined(__sun) && defined(__GNUC__)
+  /* long double support under SunOS/gcc is pretty much non-existent.
+     You'll get the odd bus error in library routines like printf() */
+#else
   printf ("%Lf\n", ldblit(f));
 #endif
+
   ld = 666;
   ffi_call(&cif, FFI_FN(ldblit), &ld, values);
 
-#if 1
-  /* This is ifdef'd out for now. long double support under SunOS/gcc
-     is pretty much non-existent.  You'll get the odd bus error in library
-     routines like printf().  */
+#if defined(__sun) && defined(__GNUC__)
+  /* long double support under SunOS/gcc is pretty much non-existent.
+     You'll get the odd bus error in library routines like printf() */
+#else
   printf ("%Lf, %Lf, %Lf, %Lf\n", ld, ldblit(f), ld - ldblit(f), LDBL_EPSILON);
 #endif