platform/upstream/libffi.git
4 years agoChange s390x options
Anthony Green [Thu, 7 Nov 2019 17:17:29 +0000 (12:17 -0500)]
Change s390x options

4 years agoMark xfail for m68k and alpha.
Anthony Green [Thu, 7 Nov 2019 17:03:19 +0000 (12:03 -0500)]
Mark xfail for m68k and alpha.

4 years agoClean up. Debug s390x cpu support.
Anthony Green [Thu, 7 Nov 2019 17:03:01 +0000 (12:03 -0500)]
Clean up.  Debug s390x cpu support.

4 years agoTry coldfire for m68k builds
Anthony Green [Thu, 7 Nov 2019 16:50:08 +0000 (11:50 -0500)]
Try coldfire for m68k builds

4 years agoAdd GCC_OPTIONS flag
Anthony Green [Thu, 7 Nov 2019 16:37:25 +0000 (11:37 -0500)]
Add GCC_OPTIONS flag

4 years agoCompile m68k target for m5208
Anthony Green [Thu, 7 Nov 2019 16:26:53 +0000 (11:26 -0500)]
Compile m68k target for m5208

4 years agoDump supported CPUs
Anthony Green [Thu, 7 Nov 2019 16:15:13 +0000 (11:15 -0500)]
Dump supported CPUs

4 years agoFix shell script
Anthony Green [Thu, 7 Nov 2019 16:00:09 +0000 (11:00 -0500)]
Fix shell script

4 years agoPrint debug output. Use podman when available.
Anthony Green [Thu, 7 Nov 2019 15:56:40 +0000 (10:56 -0500)]
Print debug output. Use podman when available.

4 years agoSet QEMU_CPU within the build containers
Anthony Green [Thu, 7 Nov 2019 14:23:10 +0000 (09:23 -0500)]
Set QEMU_CPU within the build containers

4 years agoSet QEMU_CPU for m68k
Anthony Green [Thu, 7 Nov 2019 12:36:11 +0000 (07:36 -0500)]
Set QEMU_CPU for m68k

4 years agoAdd alpha and m68k testing
Anthony Green [Thu, 7 Nov 2019 03:02:11 +0000 (22:02 -0500)]
Add alpha and m68k testing

4 years agoSett CC and CXX for cross builds
Anthony Green [Wed, 6 Nov 2019 16:31:17 +0000 (11:31 -0500)]
Sett CC and CXX for cross builds

4 years agoautogen before cross building
Anthony Green [Wed, 6 Nov 2019 16:21:46 +0000 (11:21 -0500)]
autogen before cross building

4 years agoFix typo. Don't double test.
Anthony Green [Wed, 6 Nov 2019 16:10:59 +0000 (11:10 -0500)]
Fix typo. Don't double test.

4 years agoUse the right cross build container image
Anthony Green [Wed, 6 Nov 2019 15:51:33 +0000 (10:51 -0500)]
Use the right cross build container image

4 years agoTry s390x and sh4 cross builds/tests
Anthony Green [Wed, 6 Nov 2019 15:49:04 +0000 (10:49 -0500)]
Try s390x and sh4 cross builds/tests

4 years agoRevert to older test cross compilers
Anthony Green [Wed, 6 Nov 2019 11:47:05 +0000 (06:47 -0500)]
Revert to older test cross compilers

4 years agoUse older test compiler. Enable s390x testing.
Anthony Green [Wed, 6 Nov 2019 11:44:19 +0000 (06:44 -0500)]
Use older test compiler.  Enable s390x testing.

4 years agoTest cross-builds for hppa-linux-gnu
Anthony Green [Wed, 6 Nov 2019 11:30:52 +0000 (06:30 -0500)]
Test cross-builds for hppa-linux-gnu

4 years agoReference FFI_BUILDING_DLL
Anthony Green [Wed, 6 Nov 2019 11:30:20 +0000 (06:30 -0500)]
Reference FFI_BUILDING_DLL

4 years agorc2. hack as per: https://github.com/travis-ci/travis-ci/issues/6934
Anthony Green [Sun, 3 Nov 2019 11:28:51 +0000 (06:28 -0500)]
rc2. hack as per: https://github.com/travis-ci/travis-ci/issues/6934

4 years agoAdd sparc64-linux-gnu testing back
Anthony Green [Thu, 31 Oct 2019 23:07:04 +0000 (19:07 -0400)]
Add sparc64-linux-gnu testing back

4 years agoDon't build sparc. Build mips.
Anthony Green [Thu, 31 Oct 2019 19:32:48 +0000 (15:32 -0400)]
Don't build sparc.  Build mips.

4 years agoTest mips64el-linux-gnu
Anthony Green [Thu, 31 Oct 2019 18:35:43 +0000 (14:35 -0400)]
Test mips64el-linux-gnu

4 years agoDon't run autogen.sh all the time.
Anthony Green [Thu, 31 Oct 2019 16:33:22 +0000 (12:33 -0400)]
Don't run autogen.sh all the time.

4 years agoTest sparc64-linux-gnu on travis
Anthony Green [Thu, 31 Oct 2019 16:23:11 +0000 (12:23 -0400)]
Test sparc64-linux-gnu on travis

4 years agoTry new cfarm build for power
Anthony Green [Thu, 31 Oct 2019 14:53:37 +0000 (10:53 -0400)]
Try new cfarm build for power

4 years agoAdd make_sunver.pl to distribution.
Anthony Green [Sat, 26 Oct 2019 14:13:20 +0000 (10:13 -0400)]
Add make_sunver.pl to distribution.

4 years agoFix FFI_STDCALL ABI (#514)
Andreas Schwab [Sat, 26 Oct 2019 12:57:05 +0000 (14:57 +0200)]
Fix FFI_STDCALL ABI (#514)

Even for a stdcall function, the stack alignment is still the
responsibility of the caller.  Remember the original, not stack-aligned
argument size, but align when setting up a stack frame.  In
ffi_closure_inner, return the true argument size, so that
ffi_[go_]closure_STDCALL doesn't adjust too much.

4 years agoAdd missing build script, make_sunver.pl.
Anthony Green [Sat, 26 Oct 2019 11:26:30 +0000 (07:26 -0400)]
Add missing build script, make_sunver.pl.

4 years agoUpdate versions to 3.3-rc1
Anthony Green [Thu, 24 Oct 2019 11:03:43 +0000 (07:03 -0400)]
Update versions to 3.3-rc1

4 years agoRemove some debugging output
Anthony Green [Thu, 24 Oct 2019 10:41:11 +0000 (06:41 -0400)]
Remove some debugging output

4 years agoDon't test ppc64le with buggy qemu.
Anthony Green [Thu, 24 Oct 2019 10:39:00 +0000 (06:39 -0400)]
Don't test ppc64le with buggy qemu.

4 years agoStrip build status from README.md at 'make dist' time.
Anthony Green [Thu, 24 Oct 2019 10:24:14 +0000 (06:24 -0400)]
Strip build status from README.md at 'make dist' time.

4 years agoAdd missing dist files.
Anthony Green [Thu, 24 Oct 2019 09:25:11 +0000 (05:25 -0400)]
Add missing dist files.

4 years agoTest on arm32v7-linux-gnu, ppc64le-linux-gnu and aarch64-linux-gnu.
Anthony Green [Wed, 16 Oct 2019 20:05:46 +0000 (16:05 -0400)]
Test on arm32v7-linux-gnu, ppc64le-linux-gnu and aarch64-linux-gnu.

Use docker images and qemu to test libffi for non-x86 architectures on
travis-ci.
Use the LIBFFI_TEST_OPTIMIZATION environment variable to
force specific optimization levels at test time.

4 years agoFix comments.
Anthony Green [Mon, 14 Oct 2019 10:46:52 +0000 (06:46 -0400)]
Fix comments.

4 years agoUpdate copyright year
Anthony Green [Mon, 14 Oct 2019 09:47:07 +0000 (05:47 -0400)]
Update copyright year

4 years agoAdd long double test (#492)
zhanhb [Wed, 9 Oct 2019 10:59:32 +0000 (18:59 +0800)]
Add long double test (#492)

Required to fix build error on macos with gcc-9

4 years agoMaking the change to correct the comment when SUN (#521)
pichikaudaykiran [Wed, 9 Oct 2019 10:56:06 +0000 (16:26 +0530)]
Making the change to correct the comment when SUN (#521)

and GCC are used together

4 years agohandle compilation warnings with ftruncate API (#508)
pnallan [Tue, 8 Oct 2019 13:16:47 +0000 (18:46 +0530)]
handle compilation warnings with ftruncate API (#508)

* fix me: avoid warning while handle ftruncate API

Signed-off-by: Prasad Nallani <prasad.nallani@intel.com>
* Update closures.c

4 years agoClean up line endings (#509)
John Ericson [Tue, 8 Oct 2019 10:58:52 +0000 (06:58 -0400)]
Clean up line endings (#509)

The CLRF visual studio files can be kept that way, but recognized as
text. The assembly file can be converted to LF.

4 years agoPowerPC bugfixes (#520)
Samuel Holland [Tue, 8 Oct 2019 10:57:28 +0000 (05:57 -0500)]
PowerPC bugfixes (#520)

* powerpc: Silence warnings about unused labels

* powerpc: Fix a couple of comments

* powerpc: Fix alignment after float structs

* powerpc: Don't pad rvalues copied from FP regs

* powerpc: Add missing check in struct alignment

* powerpc: Support homogeneous long double structs

4 years agoUse https for moxie toolchain repo
Anthony Green [Tue, 8 Oct 2019 10:24:55 +0000 (06:24 -0400)]
Use https for moxie toolchain repo

4 years agofix mingw build and crashing bugs for Python Windows ARM64 (#496)
Paul Monson [Wed, 7 Aug 2019 18:57:45 +0000 (11:57 -0700)]
fix mingw build and crashing bugs for Python Windows ARM64 (#496)

* fix mingw build and crashing bugs for Python Windows ARM64

* Fix issues found in PR review

4 years agoClear the apt cache
Anthony Green [Wed, 26 Jun 2019 10:18:48 +0000 (06:18 -0400)]
Clear the apt cache

4 years agoMore debugging output
Anthony Green [Wed, 26 Jun 2019 10:01:15 +0000 (06:01 -0400)]
More debugging output

4 years agoDebug moxie builds in travis
Anthony Green [Wed, 26 Jun 2019 03:39:52 +0000 (23:39 -0400)]
Debug moxie builds in travis

4 years ago libffi: added ARM64 support for Windows (#486)
ossdev07 [Wed, 26 Jun 2019 02:01:22 +0000 (07:31 +0530)]
 libffi: added ARM64 support for Windows (#486)

*  libffi: added ARM64 support for Windows

    1. ported sysv.S to win64_armasm.S for armasm64 assembler
    2. added msvc_build folder for visual studio solution
    3. updated README.md for the same
    4. MSVC solution created with the changes, and below test suites are tested
       with test script written in python.

       libffi.bhaible
       libffi.call
    5. Basic functionality of above test suites are getting passed

Signed-off-by: ossdev07 <ossdev@puresoftware.com>
* Update README.md

5 years agouuencode compressed log files for travis
Anthony Green [Sun, 28 Apr 2019 11:36:24 +0000 (07:36 -0400)]
uuencode compressed log files for travis

5 years agohppa: avoid TEXTREL in .eh_frame section (#447)
Sergei Trofimovich [Sat, 27 Apr 2019 19:53:29 +0000 (20:53 +0100)]
hppa: avoid TEXTREL in .eh_frame section (#447)

Before the change hand-crafted .eh_frame section contained
ABS relocation and caused TEXTREL tag to be emitted:

```
$ ./configure --host=hppa2.0-unknown-linux-gnu LDFLAGS=-Wl,-z,text
$ make
...
/usr/libexec/gcc/hppa2.0-unknown-linux-gnu/ld:
    read-only segment has dynamic relocations.
```

Link failure is caused by absolute address of FDEs
encoded into .eh_frame entries.

Fixed TEXTREL by using pcrel (instead of ABS) encoding
for absolute addresses (__PIC__ code) by adding augmentation
information ("zR" CIE type).

All tests still pass on hppa2.0. The specific tests that still pass
and exercise this code path:
    testsuite/libffi.call/unwindtest.cc
    testsuite/libffi.call/unwindtest_ffi_call.cc

Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
5 years agofix x86/x64 MSVC build (#487)
driver1998 [Sat, 27 Apr 2019 19:21:44 +0000 (03:21 +0800)]
fix x86/x64 MSVC build (#487)

5 years agoadd support for 32-bit ARM on Windows (#477)
Paul Monson [Fri, 26 Apr 2019 11:58:58 +0000 (04:58 -0700)]
add support for 32-bit ARM on Windows (#477)

* add support for 32-bit ARM on Windows

* fix mismatched brace in appveyor.yml

* remove arm platform from appveyor.yml for now

* fix arm build

* fix typo

* fix assembler names

* try Visual Studio 2017

* add windows arm32 to .appveyor.yml

* update README.md

5 years agoMerge pull request #482 from sharkcz/aarch64
Tom Tromey [Tue, 9 Apr 2019 20:00:10 +0000 (14:00 -0600)]
Merge pull request #482 from sharkcz/aarch64

fix check for Linux/aarch64

5 years agofix check for Linux/aarch64
Dan Horák [Fri, 29 Mar 2019 13:19:20 +0000 (14:19 +0100)]
fix check for Linux/aarch64

fixes #473

5 years agoCleanup symbol exports on darwin and add architecture preprocessor checks to assist...
Jeremy Huddleston Sequoia [Tue, 19 Feb 2019 12:11:28 +0000 (04:11 -0800)]
Cleanup symbol exports on darwin and add architecture preprocessor checks to assist in building fat binaries (eg: i386+x86_64 on macOS or arm+aarch64 on iOS) (#450)

* x86: Ensure _efi64 suffixed symbols are not exported

* x86: Ensure we do not export ffi_prep_cif_machdep

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* x86: Ensure we don't export ffi_call_win64, ffi_closure_win64, or ffi_go_closure_win64

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* closures: Silence a semantic warning

libffi/src/closures.c:175:23: This function declaration is not a prototype

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* aarch64: Ensure we don't export ffi_prep_cif_machdep

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* arm: Ensure we don't export ffi_prep_cif_machdep

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* aarch64, arm, x86: Add architecture preprocessor checks to support easier fat builds (eg: iOS)

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* x86: Silence some static analysis warnings

libffi/src/x86/ffi64.c:286:21: The left operand of '!=' is a garbage value due to array index out of bounds
libffi/src/x86/ffi64.c:297:22: The left operand of '!=' is a garbage value due to array index out of bounds

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* aarch: Use FFI_HIDDEN rather than .hidden

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
* ffi.h: Don't advertise ffi_java_rvalue_to_raw, ffi_prep_java_raw_closure, and ffi_prep_java_raw_closure_loc when FFI_NATIVE_RAW_API is 0

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
5 years agoFix cfi checks for old compiler (#453)
dabrain34 [Tue, 19 Feb 2019 12:09:09 +0000 (13:09 +0100)]
Fix cfi checks for old compiler (#453)

cfi_sections can be unsupported when cfi_startproc
and cfi_endproc are.

5 years agochanges for win32 on windows (#468)
Paul Monson [Tue, 19 Feb 2019 11:58:25 +0000 (03:58 -0800)]
changes for win32 on windows (#468)

5 years agoaarch64: Flush code mapping in addition to data mapping (#471)
Florian Weimer [Tue, 19 Feb 2019 11:55:11 +0000 (12:55 +0100)]
aarch64: Flush code mapping in addition to data mapping (#471)

This needs a new function, ffi_data_to_code_pointer, to translate
from data pointers to code pointers.

Fixes issue #470.

5 years agoRemove -Os testing. No ABI impact, and helps trim log lengths.
Anthony Green [Tue, 12 Feb 2019 13:50:30 +0000 (08:50 -0500)]
Remove -Os testing. No ABI impact, and helps trim log lengths.

5 years agoMerge pull request #461 from NativeScript/bektchiev/fix-visibility-hidden-check-in...
Tom Tromey [Wed, 19 Dec 2018 21:18:40 +0000 (14:18 -0700)]
Merge pull request #461 from NativeScript/bektchiev/fix-visibility-hidden-check-in-configure

fix(configure): Correctly detect visibility("hidden") support on Darwin

5 years agofix(configure): Correctly detect visibility("hidden") support on Darwin
Martin Bektchiev [Wed, 19 Dec 2018 16:33:08 +0000 (18:33 +0200)]
fix(configure): Correctly detect visibility("hidden") support on Darwin

5 years agoMerge pull request #457 from NativeScript/bektchiev/fix-arm64-q3-q4-args
Tom Tromey [Fri, 30 Nov 2018 15:16:00 +0000 (08:16 -0700)]
Merge pull request #457 from NativeScript/bektchiev/fix-arm64-q3-q4-args

Fix Q registers parameter passing on ARM64

5 years agoFix Q registers parameter passing on ARM64
Martin Bektchiev [Wed, 31 Oct 2018 13:53:54 +0000 (15:53 +0200)]
Fix Q registers parameter passing on ARM64

The second two quads are located at offset 32 not 16

5 years agoMerge pull request #443 from jeremyhu/master
Tom Tromey [Wed, 19 Sep 2018 13:29:36 +0000 (07:29 -0600)]
Merge pull request #443 from jeremyhu/master

Update FFI_HIDDEN() to use .private_extern on Apple platforms and use the macro where appropriate

5 years agoMerge pull request #449 from gpakosz/align-macros
Tom Tromey [Tue, 18 Sep 2018 14:26:31 +0000 (08:26 -0600)]
Merge pull request #449 from gpakosz/align-macros

Prefix ALIGN_DOWN macro with FFI_

5 years agoPrefix ALIGN_DOWN macro with FFI_
Gregory Pakosz [Tue, 18 Sep 2018 13:19:53 +0000 (15:19 +0200)]
Prefix ALIGN_DOWN macro with FFI_

5 years agoMerge pull request #445 from andreas-schwab/master
Tom Tromey [Sat, 11 Aug 2018 15:53:15 +0000 (09:53 -0600)]
Merge pull request #445 from andreas-schwab/master

RISC-V go closures

5 years agoRISC-V go closures
Andreas Schwab [Thu, 9 Aug 2018 10:12:29 +0000 (12:12 +0200)]
RISC-V go closures

This implements go closures for RISC-V.  It has been tested on
riscv64-suse-linux and against the libgo testsuite.

5 years agoUpdate FFI_HIDDEN() to use .private_extern on Apple platforms and use the macro where...
Jeremy Huddleston Sequoia [Fri, 20 Jul 2018 16:37:43 +0000 (09:37 -0700)]
Update FFI_HIDDEN() to use .private_extern on Apple platforms and use the macro where appropriate

Fix issue #439

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
5 years agoAdd compact unwind for darwin/i386 (#440)
Jeremy Huddleston Sequoia [Mon, 25 Jun 2018 11:38:58 +0000 (04:38 -0700)]
Add compact unwind for darwin/i386 (#440)

* x86: Add implementation of compact unwind for ffi_call_i386, et al.

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
* x86: Use __text as the section name to avoid deprecated section name warnings.

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
* darwin: Add missing regular,debug attributes for compact unwind sections

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
5 years agoMark sysv.S as SafeSEH compatible (#438)
Shoaib Meenai [Mon, 18 Jun 2018 00:04:24 +0000 (17:04 -0700)]
Mark sysv.S as SafeSEH compatible (#438)

It contains no exception handler, so we can just emit the special
@feat.00 symbol to indicate that it's trivially SafeSEH compatible.
SafeSEH only applies to x86 and not x86-64, hence its inclusion in the
x86-specific block. See [1] for details.

[1] https://msdn.microsoft.com/en-us/library/windows/desktop/ms680547(v=vs.85).aspx#the_.sxdata_section_

5 years agoi386: Fix missing break; in case statement leading to incorrectly returned FFI_BAD_AB...
Jeremy Huddleston Sequoia [Mon, 18 Jun 2018 00:01:50 +0000 (17:01 -0700)]
i386: Fix missing break; in case statement leading to incorrectly returned FFI_BAD_ABI (#437)

* i386: Add missing break triggering dead store static analyzer checks.

Register calling sequence is being reported as bad ABI instead of working as intended.

Found-by: Clang Static Analysis
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
* Mark ffi arm sysv entry points as private_extern.

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
* x86_64: Add implementation of compact unwind for ffi_call_unix64.

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
6 years agoHandle FFI_GNUW64 on non-Windows systems (EFI)
Anthony Green [Wed, 9 May 2018 17:21:02 +0000 (13:21 -0400)]
Handle FFI_GNUW64 on non-Windows systems (EFI)

6 years agoMark some cases as xfail due to GCC bug
Anthony Green [Wed, 9 May 2018 14:50:46 +0000 (10:50 -0400)]
Mark some cases as xfail due to GCC bug

6 years agoRevert "Remove some symbol exports and cleanup newline warnings (#433)"
Anthony Green [Sat, 5 May 2018 11:41:53 +0000 (07:41 -0400)]
Revert "Remove some symbol exports and cleanup newline warnings (#433)"

This reverts commit a5a0f3cf36dfb4d64316414a872288c3170e6c1d.

6 years agoRemove some symbol exports and cleanup newline warnings (#433)
Jeremy Huddleston Sequoia [Sat, 5 May 2018 10:44:33 +0000 (03:44 -0700)]
Remove some symbol exports and cleanup newline warnings (#433)

* build: Ensure darwin generated sources end with a new line

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
* build: Use .private_extern where missing to prevent exporting symbols that are not API

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@macports.org>
6 years agoRe-enable msabi testing (#436)
hjl-tools [Wed, 2 May 2018 13:19:58 +0000 (06:19 -0700)]
Re-enable msabi testing (#436)

* Revert "disable msabi testing for now"

This reverts commit 7b7638eb0eac2adfa72f7ec9f254ba287c9947e2.

* x86: Correct testing for 64-bit x86-64

Since passing -m32 to x86-64 compiler will generate i386 code, we
need to check both __ILP32__ and __i386__ for 64-bit x86-64.

* x86: Check __i386__ instead of i?86-*-* targets

Since passing -m32 to x86-64 compiler will generate i386 code, we need
to check __i386__ instead of i?86-*-* targets for i386 targets.

* i386: Properly passing integer parameters in registers

For thiscall and fastcall, if the paramter passed as 64-bit integer or
struct, all following integer paramters will be passed on stack.

* test: Add ABI_ATTR to callback_code

Add ABI_ATTR to callback_code to properly test different ABIs.

6 years agoPowerPC64 ELFv1 fp arg fixes
Alan Modra [Wed, 2 May 2018 04:25:29 +0000 (13:55 +0930)]
PowerPC64 ELFv1 fp arg fixes

The ELFv1 ABI says: "Single precision floating point values are mapped
to the second word in a single doubleword" and also "Floating point
registers f1 through f13 are used consecutively to pass up to 13
floating point values, one member aggregates passed by value
containing a floating point value, and to pass complex floating point
values".

libffi wasn't expecting float args in the second word, and wasn't
passing one member aggregates in fp registers.  This patch fixes those
problems, making use of the existing ELFv2 homogeneous aggregate
support since a one element fp struct is a special case of an
homogeneous aggregate.

I've also set a flag when returning pointers that might be used one
day.  This is just a tidy since the ppc64 assembly support code
currently doesn't test FLAG_RETURNS_64BITS for integer types..

* src/powerpc/ffi_linux64.c (discover_homogeneous_aggregate):
Compile for ELFv1 too, handling single element aggregates.
(ffi_prep_cif_linux64_core): Call discover_homogeneous_aggregate
for ELFv1.  Set FLAG_RETURNS_64BITS for FFI_TYPE_POINTER return.
(ffi_prep_args64): Call discover_homogeneous_aggregate for ELFv1,
and handle single element structs containing float or double
as if the element wasn't wrapped in a struct.  Store floats in
second word of doubleword slot when big-endian.
(ffi_closure_helper_LINUX64): Similarly.

6 years agoAdd license for build-time tools
Anthony Green [Tue, 1 May 2018 10:46:30 +0000 (06:46 -0400)]
Add license for build-time tools

6 years agoFix two "return" issues in x86/ffi64.c (#431)
Tom Tromey [Sat, 28 Apr 2018 10:46:10 +0000 (04:46 -0600)]
Fix two "return" issues in x86/ffi64.c (#431)

Issue #70 pointed out that at least one compiler didn't like:

    return ffi_call_efi64(cif, fn, rvalue, avalue);

... where the return type is "void".  This patch splits the statement
into two.

I also noticed that ffi_call_go here seems to do a double call.  I
suspect a "return" is missing here, so this patch adds it as well.

6 years agoMerge pull request #429 from yousong/m4subst
Tom Tromey [Fri, 27 Apr 2018 16:48:51 +0000 (10:48 -0600)]
Merge pull request #429 from yousong/m4subst

build: fix subst toolexecdir, toolexeclibdir

6 years agobuild: fix subst toolexecdir, toolexeclibdir
Yousong Zhou [Thu, 26 Apr 2018 11:27:54 +0000 (19:27 +0800)]
build: fix subst toolexecdir, toolexeclibdir

On CentOS 7.4, configure script generated by autogen.sh will output the
following values when compiling GCC source code

    toolexecdir='NONE/$(target_alias)'
    toolexeclibdir='NONE/$(target_alias)/lib'

and cause build error

    ... -o libffi.la -rpath NONE/riscv64-bs-linux-gnu/lib ...
    checking for shl_load in -ldld... libtool: link: only absolute run-paths are allowed

6 years agoupdate to https for cygwin download (#428)
Stephen [Wed, 25 Apr 2018 20:17:32 +0000 (13:17 -0700)]
update to https for cygwin download (#428)

6 years agoxtensa-linux: use cache flush instruction only if it is available (#426)
Lucas Pluvinage [Fri, 20 Apr 2018 22:24:50 +0000 (00:24 +0200)]
xtensa-linux: use cache flush instruction only if it is available (#426)

6 years agoVarious MIPS Fixes (#425)
James Cowgill [Thu, 19 Apr 2018 00:28:23 +0000 (01:28 +0100)]
Various MIPS Fixes (#425)

* mips: simplify closure #defines

This commit should have no visible effect.

* mips: add special handling of variadic functions

MIPS requires special handling of variadic functions which pass floating
point arguments:
* In the o32 ABI, all float arguments are passed in integer registers.
* In the n32/n64 ABIs, float arguments after the ellipsis are passed in
  integer registers.

Implement this in libffi. To support this in n32/n64 closures, we need to add
a new mips_nfixedargs field to ffi_cif which will break the libffi ABI.

This fixes the libffi.call/cls_longdouble_va.c test which was failing on
64-bit MIPS.

* mips: align argn for all 64-bit types in o32 closure handler

Ensure that argn is pre-aligned for all 64-bit argument types (including
doubles) and not just integer types.

This fixes closures of the form "f(float, double, <some integer args>)".
Previously the first integer argument would be read from a2 which is garbage
at this point (the float arguments have already "consumed" a0-a3). After
this commit, argn is correctly padded between the "float" and "double"
arguments so that the first integer argument is read from the stack.

Fixes "double f(float,double,int)" test in #371

* mips: do not read from floating point register if returning a struct

In the o32 ABI, the pointer passed in a0 used to return structures
indirectly is treated as the first argument for argument allocation purposes.
This means that it should inhibit floating point registers the same way that
other integer arguments do.

Fixes "Double f(float,Double,double)" test in #371

* mips: fix pointer cast warnings

Fix two pointer cast warnings when compiled on 64-bit mips by casting
through uintptr_t.

Fixes mips64el part of #404

6 years agoMerge branch 'master' of github.com:/libffi/libffi
Anthony Green [Sun, 8 Apr 2018 22:25:43 +0000 (18:25 -0400)]
Merge branch 'master' of github.com:/libffi/libffi

6 years agoFix case where callback arg value is split across regs and stack
Anthony Green [Sun, 8 Apr 2018 22:25:34 +0000 (18:25 -0400)]
Fix case where callback arg value is split across regs and stack

6 years agoFix issue #421 (#422)
Andreas Krebbel [Thu, 5 Apr 2018 12:27:32 +0000 (14:27 +0200)]
Fix issue #421 (#422)

Fantastic - thanks for digging into this.

6 years ago3.3 release candidate 0
Anthony Green [Mon, 2 Apr 2018 12:30:17 +0000 (08:30 -0400)]
3.3 release candidate 0

6 years agodisable msabi testing for now
Anthony Green [Mon, 2 Apr 2018 12:24:44 +0000 (08:24 -0400)]
disable msabi testing for now

6 years agoFix appveyor windows build (#420)
fwg [Mon, 2 Apr 2018 11:55:31 +0000 (13:55 +0200)]
Fix appveyor windows build (#420)

* Fix msvcc dll build by adding dllexport decorations to all API declarations

* Fix appveyor build for VS 2013

Use the new -DFFI_BUILDING_DLL for producing a working DLL. Update the
msvcc.sh wrapper script to successfully compile the testsuite files.

* MSVC build: suppress warnings in testsuite

* fix testsuite on appveyor

6 years agoTrim some optimization tests
Anthony Green [Thu, 29 Mar 2018 11:22:57 +0000 (07:22 -0400)]
Trim some optimization tests

6 years agoSimplify matrix
Anthony Green [Thu, 29 Mar 2018 11:22:38 +0000 (07:22 -0400)]
Simplify matrix

6 years agoRemove warning message from clang
Anthony Green [Thu, 29 Mar 2018 11:10:23 +0000 (07:10 -0400)]
Remove warning message from clang

6 years agoExpand ABI tests on x86. Testsuite bug fixes.
Anthony Green [Thu, 29 Mar 2018 11:01:14 +0000 (07:01 -0400)]
Expand ABI tests on x86.  Testsuite bug fixes.

6 years agodeal with msvc warnings
Anthony Green [Tue, 27 Mar 2018 18:45:40 +0000 (14:45 -0400)]
deal with msvc warnings

6 years agofix quoting issue
Anthony Green [Tue, 27 Mar 2018 18:37:50 +0000 (14:37 -0400)]
fix quoting issue

6 years agoDon't ignore a file
Anthony Green [Tue, 27 Mar 2018 18:35:10 +0000 (14:35 -0400)]
Don't ignore a file