defconfigs: use the new headers-version-same-as-kernel-version option

Now that we can say that the linux headers version should match that of
the kernel to be built, we inverse the logic in our defconfigs, as it is
more sensible that way.

And also because we'll get rid of the former, converse kernel-same-as-headers
option.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

bind: fix intermittent build issues with high BR2_JLEVEL

Build sometimes breaks with:

libtool: link: `unix/os.lo' is not a valid libtool object
make[3]: *** [rndc-confgen] Error 1
make[3]: *** Waiting for unfinished jobs....
make[4]: Leaving directory `/scratch/peko/build/bind-9.6-ESV-R4/bin/rndc/unix'

So disable parallel builds.

This patch was removed with commit
c36b5d89c5616f7ca0a7295cbb5c231606beb71e by Gustavo Zacarias
<gustavo@zacarias.com.ar> but the problem still occurs, so disabling
parallel builds again.

Fixes:
http://autobuild.buildroot.org/results/220/2201f04170ea8ef0961e907efce07c041a57c229/

Signed-off-by: Jan Heylen <heyleke@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libbsd: upgrade to v0.8.2 and expand architecture support

When libbsd still supported uClibc as a C library, we added the architecture
exclusions because uClibc's a.out.h includes linux/a.out.h. The latter only
exists for the specified architectures.

However, glibc doesn't include linux/a.out.h, it instead has its own
implementation and it adds a flag to indicate if a.out is supported on this
architecture or not.

Since libbsd currently only supports glibc-based toolchains, the architecture
exclusions are no longer valid.

On microblaze, the build still fails, but this time because of ELF support.
libbsd explicitly handles architectures and microblaze is not one of them (see
local-elf.h).

Signed-off-by: Carlos Santos <casantos@datacom.ind.br>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/php: security bump version to 5.6.18

Changelog: http://www.php.net/ChangeLog-5.php#5.6.18
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Acked-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xdotool: needs MMU support

Fixes:
http://autobuild.buildroot.net/results/d750919a0d5cfb624937c72253651a431212e238

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/mraa: bump version to 0.9.1

Removed patches applied upstream:
https://github.com/intel-iot-devkit/mraa/commit/86a0e10c037e5c069c48012f169f9e8e2587a0ef
https://github.com/intel-iot-devkit/mraa/commit/a87a104efeb83d4e762992e5be3b8908e6d2bf40

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-psutil: bump to 3.4.2

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/rsyslog: fix musl buil

The build error was not yet being found by the autobuilders:

omfile.c: In function ‘prepareFile’:
omfile.c:580:3: error: implicit declaration of function ‘open’ [-Werror=implicit-function-declaration]
   fd = open((char*) newFileName, O_WRONLY|O_APPEND|O_CREAT|O_NOCTTY|O_CLOEXEC,
   ^

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/lirc-tools: not available on musl

GLOB_BRACE is not supported by the musl libc:
http://www.openwall.com/lists/musl/2013/09/16/1

Upstream correctly rejected the hack proposed by me:
http://patchwork.ozlabs.org/patch/576188/
http://sourceforge.net/p/lirc/tickets/174/#b087

Fixes
http://autobuild.buildroot.net/results/603/603c4426c47220808305a416f3d60fec2d8c02a3/
http://autobuild.buildroot.net/results/c08/c08abddeb5f0d0d1adec63ab4390562e3cfe1087/

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/sconeserver: location support needs host-pkgconf

Patch suggested by Arnout: http://patchwork.ozlabs.org/patch/576212/

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-setuptools: enable Python 3.x support

Add missing XML dependency.

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: add a upstream patch to fix vlc runtime issues

IDN related declaration without implementation is a bad idea.

Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: fix MIPS builds with disabled threads

There was a regression found while building for MIPS
with disabled threads. Upstream patch fixed it.

Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Acked-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Reviewed-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Tested-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain: bump ARC toolchain components to arc-2015.12

This change introduces newer ARC toolchain in Buildroot.

That new arc-2015.12 release doesn't bring any significant changes
but mostly is focused on fixes and minor improvements here and there.
Most noticeable changes are:
 * GCC: Source update to v4.8.5
 * GDB: Updated to upstream 7.10 release.

You may find more info on fixes and improvements in that release at:
https://github.com/foss-for-synopsys-dwc-arc-processors/toolchain/releases/tag/arc-2015.12

Signed-off-by: Lada Trimasova <ltrimas@synopsys.com>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

grep: bump to version 2.23

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

network-manager: bump to version 1.0.10

Signed-off-by: Petr Vorel <petr.vorel@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

dbus-glib: bump version to 0.106

Signed-off-by: Petr Vorel <petr.vorel@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

protobuf: introduce BR2_PACKAGE_PROTOBUF_ARCH_SUPPORTS

The architecture dependencies of protobuf are going to change, and
they are already duplicated between protobuf, ola and mosh. In order
to factorize the expression of those dependencies, this commit
introduces BR2_PACKAGE_PROTOBUF_ARCH_SUPPORTS.

Note that we include in this hidden Config.in option both the target
architecture dependency and the host architecture dependency.

Finally, this commit also fixes a real mistake in the mosh Config.in
file, where the condition on the architecture dependency for the
Config.in comment was inverted: we only want to show the Config.in
comment when we are on supported architectures.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-cffi: bump to 1.5.0

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

board: add support for ARM Juno r0 and r1 Development Boards

The Juno ARM Development Platform (ADP) is a software development
platform for ARMv8-A, it includes the Juno Versatile Express board
and an ARMv8-A reference software port available through Linaro.

Signed-off-by: Joao Pinto <jpinto@synopsys.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/misc: Adding Vagrant file for provisioning

This patch adds a Vagrant file to buildroot. With this file
you can provision a complete buildroot developing environment
in minutes on all major platforms (Linux/Mac/Windows).

[Peter: bump to 2GB RAM, hardcode Buildroot release, add unzip,
drop website update and tweak manual text as suggested by Yann]
Signed-off-by: Angelo Compagnucci <angelo.compagnucci@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain/external: newer Linaro toolchains do not provide source code

Currently, we have a pattern-matching that automatically derives the
the source tarball filename from the binary tarball filename.

However, the latest Linaro toolchains no longer follow that scheme (and
do not even readily provide the sources...).

Remove the generic pattern-matching, and explicitly set the source
tarball name for those toolchains that do have a source tarball readily
available.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain: copy_toolchain_lib_root: rename LIBSPATH to LIBPATHS

LIBSPATH is populated based on a find with a pattern that can look like:
    libfoo*.so
and thus the output of the find will contain all file paths that match this
pattern.

Unfortunately, the name LIBSPATH suggests that only one entry is returned,
rather than possibly multiple.

As this code is quite complex, use the more accurate name LIBPATHS iso
LIBSPATH.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

qt5: bump version to 5.5.1

Reduce hash lists to strongest only (sha256).

Remove upstream committed patches (qt5base):
 - 0004-no-gold-linker-for-host-build.patch (see [1], [2])
 - 0007-big_endian-fix_destformat.patch (see [3])
 - 0008-QtGui-force-image-dir-in-include_path.patch (see [4], [5])

[1] https://bugreports.qt.io/browse/QTBUG-46125
[2] https://codereview.qt-project.org/#/c/113563/
[3] https://codereview.qt-project.org/#/c/120654/
[4] https://bugreports.qt.io/browse/QTBUG-47400
[5] https://codereview.qt-project.org/#/c/122145/

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

binutils: fix .loc processing in xtensa gas

Fixes:
  http://autobuild.buildroot.net/results/bff7436a800eeea92c0c92bd2846b0f2b31947fd/

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gnutls: bump to version 3.4.9

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-pyparted: new package

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-smbus-cffi: new package

[Thomas:
 - fix license, it's GPLv2, not GPLv2+.]

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-cffi: add host variant

Host variant is needed to cross-compile CFFI based C library wrappers.

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas:
 - add dependency on host-pkgconf for the host package variant.
 - add various environment variables needed to make pkg-config behave
   properly when building host-python-cffi. Otherwise, pkg-config
   returns values that are appropriate to build things for the target,
   and the build fails.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-pycparser: new package

pycparser requires C preprocessor in order to parse C related files,
hence available only as host package.

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libffi: fix headers location for host-libffi

There is a special hook for target-libffi, that moves headers
from /usr/lib/libffi-version/include to /usr/include. This patch
adds the same procedure for host-libffi.

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas:
 - factorize code between the staging headers cleanup and the host
   headers cleanup.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libffi: fix staging headers cleanup

The LIBFFI_MOVE_STAGING_HEADERS function is responsible for moving
around libffi headers to a standard location. Once this is done, it
removes the no longer used directory, but does so in $(TARGET_DIR) and
not $(STAGING_DIR). This directory is already cleaned up in
$(TARGET_DIR) in the LIBFFI_REMOVE_TARGET_HEADERS post-install target
hook.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-netaddr: new package

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas:
  - remove excessive empty new line at the end of Config.in
  - specify BSD-3c as the license, since it's the variant being used,
    and it's more specific than just BSD.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-mistune: new package

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas:
 - rename the .mk file to carry the proper name
 - rewrap Config.in help text to use the proper length and avoid a
   trailing space
 - remove excessive empty new line at the end of Config.in
 - specify BSD-3c as the license, since it's the variant being used,
   and it's more specific than just BSD.]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-click: new package

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas:
 - rewrap Config.in help text, lines were slightly too long
 - remove excessive empty line at end of Config.in
 - specify BSD-3c as the license, since BSD is not specific enough.]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-cbor: new package

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

kodi, mpd: fix samba4 related dependency mistakes

This commit fixes the following kconfig warning:

   warning: (BR2_PACKAGE_KODI_LIBSMBCLIENT && BR2_PACKAGE_MPD_LIBSMBCLIENT) selects BR2_PACKAGE_SAMBA4 which has unmet direct dependencies (!BR2_PACKAGE_SAMBA && BR2_USE_MMU && BR2_USE_WCHAR && BR2_TOOLCHAIN_HAS_NATIVE_RPC && BR2_TOOLCHAIN_HAS_THREADS && !BR2_STATIC_LIBS && !BR2_nios2)

by updating the kodi and mpd Config.in files to properly take into
account the dependencies of the BR2_PACKAGE_SAMBA4 option that they
select.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

iptables: fix libpcap static linking

libpcap can depend on other libraries like libusb and doesn't support
pkgconfig, so add those libraries to LIBS when building statically.
Fixes:
http://autobuild.buildroot.net/results/b4a/b4a3d00e9673a7aacc663c81de1d8e887a17951d/

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/ola: fix musl build

Fixes
http://autobuild.buildroot.net/results/17f/17fe88868acdbba074561da14c4417e31d506823/
http://autobuild.buildroot.net/results/d99/d99ca60cc4bf86c338fd61f9fd2de0bb9598d670/

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
[Thomas: switch to using the upstreamed Github patch.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/ola: add hash file

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/ola: bump version to 0.10.0

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

kernel-module-imx-gpu-viv: add new package

This is the Vivante kernel driver split from the kernel source code in
order to make it possible to be used in any kernel source since 3.10.53.

The driver source code provided by Freescale needs fixes so the
community forked the code to allow faster development and easier
integration of fixes from the community.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/32cf391
https://github.com/Freescale/meta-fsl-arm/commit/4249193

This package has been tested with the following commands:
 # modprobe galcore
 # cd /usr/share/examples/viv_samples/vdk/
 # ./tutorial7

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/linux-headers: add option to use same sources as the kernel

Some heavily (and most often improperly) modified Linux kernels may export
new APIs to userland, so as to speak to custom hardware or custom kernel
facilities.

However, we currently have no easy way to use such kernels as a source
for the linux-headers package, which precludes having those userland
headers intalled for userland applications to use them.

We do have a way for the kernel to use the same version as for the
headers, but that is definitely not enough, as the linux-headers package
has a version choice that is far less versatile and capable than that of
the linux package.

Add a new option for the linux-headers package, for the user to specify
that the version (really, the sources) of the kernel be used to install
the headers from.

We do that by making linux-headers patch-depend on the linux package.

We can't have linux-header simply depend on linux, because the simple
dependency means the the dependee will be configured, built and installed
before the dependent is configured. And since linux is a target package,
it depends on the toolchain, which internally dependes on linux-headers,
which would depend on linux, and we'd get a circular dependency.

Using patch-depend will ensure that linux is extracted and patched
before linux-headers is extracted, which is really all we need.

Then, we install the headers from the linux source tree, rather than
from linux-headers' source tree (as there's nothing in there!).

Since we need to install a private set for uClibc (see cde947f, uclibc:
prevent rebuilding after installation to staging), we explicitly set
INSTALL_HDR_PATH when calling the kernel' install-headers rule in
LINUX_HEADERS_CONFIGURE_CMDS, so that the headers are installed in
linux-headers' $(@D) instead of linux' $(@D).

Finally, as there is no way to know the kernel version in this case, we
must still prompt the user for the kernel series the headers are from
(like we do for a custom version) and check for consistency at build
time.

Note however that this still leaves users that want to built their
such-kernel outside of Buildroot out in the cold.

[Peter: drop comment as suggested by Thomas]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Karoly Kasza <kaszak@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

util-linux: fix static builds with uClibc

In case of uClibc librt depends on libpthread. In particular
timer_create() function uses pthread_XXX(). That means in case
of static builds it's required to link not librt alone but
together with libpthread. So if checking timer_create function
in librt fails, it is necessary to check if timer_create function
successfully links with "-lpthread".

That issues was spotted in Buldroot autobuilder failures:
http://autobuild.buildroot.net/results/759/75960db671807091fe9155aee9e46a6245e32590/
http://autobuild.buildroot.org/results/112/112e8b85783f5aaba42a937a6eb064317615a21b/

0005-build-sys-use-REALTIME_LIBS.patch is a back-port of upstream
https://github.com/karelzak/util-linux/commit/b97edfe731fdf3d3e92e40494b22658207ab6d3c
that won't apply cleanly on v2.27.1

0006-buildsys-fix-static-configuration-and-building.patch was just
applied upstream, see https://github.com/karelzak/util-linux/commit/71a77ca1ddfdd7e0d14caad9e8a607a83d61a45c

Both patches must be removed on util-linux version bump.

Signed-off-by: Lada Trimasova <ltrimas@synopsys.com>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gstreamer1: don't rely on a recent coreutils for relative symlink support

Fixes:
http://autobuild.buildroot.net/results/3ae/3ae368cfd0ecd7be4c7e864590ec5dbf79d576ac/
http://autobuild.buildroot.net/results/4c9/4c951f31984e0eff21fc165cb68065f527c237e6/

And many more.

The recent change to create the legacy gstconfig.h symlink using ln -srf
unfortunately breaks on older distributions as it was only added in
coreutils 8.16 (2012).

Instead, "manually" create the relative symlink.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-msgpack: bump to 0.4.7

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-jinja2: bump to 2.8

Signed-off-by: Mauro Condarelli <mc5686@mclink.it>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/faifa: fix musl build

Fixes
http://autobuild.buildroot.net/results/dc6/dc6e4d7327d2031ef921cccd814605dec73a0f9c/
http://autobuild.buildroot.net/results/868/868642325dbcd581373eaa9e1ef900f8a27f9121/
and others

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

unrar: new package

[Peter: also pass TARGET_LDFLAGS]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/ffmpeg: bump version to 2.8.6

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

binutils: ensure TARGET_CONFIGURE_ARGS is taken into consideration for subdirs

Binutils runs the configure scripts of subdirs at make time, so we need a
custom build command to ensure TARGET_CONFIGURE_ARGS are taken into
consideration for those (E.G. for musl gettext handling).

Reported-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/python-spidev: fix musl build

The build error was not yet being found by the autobuilders.

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/gstreamer/gst-plugins-base: Fix build on x86 if emmintrin.h is available but can't be used

I stumbled across this error using a musl-based allyesconfig but I
doubt it is related to musl. The autobuilders did not catch this
problem yet but the patch I found fits my config (BR2_i386=y,
BR2_x86_i586=y) and it solved the problem right away.

[Peter: pick patches from upstream git instead of downloading from alphine]
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core/legal-info: update the legal-info report header

In some cases, the toolchain sources are now recovered and available in
the legal-info output. So, adapt the header to use conditional instead
of an definitive negation.

Also update the part about saving the sources: it's not the license list
that defines whether sources are installed, but rather whether the
package is redistributable or not.

Update the header accordingly.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs/nitrogen6*: bump kernel version

Update to latest boundary-imx_3.14.28_1.0.0_ga commit which brings the
following additions:
- Fix cache functions export when !MULTI_CACHE
https://github.com/Freescale/linux-fslc/commit/6325d57
  - Mandatory for Vivante kernel module
- Add TW686x video decoder support
- Fix OV5642 subdev/capture driver conflict
- Add GSLX680 touchscreen support

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xdriver_xf86-video-imx-viv: bump to version 5.0.11 p7.1

The license is now provided in the package.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/1195b71

The following xorg.conf was used in order to force the use of the
vivante module:
https://github.com/Freescale/meta-fsl-arm/blob/master/recipes-graphics
/xorg-xserver/xserver-xf86-config/mx6/xorg.conf

Note that the X server must be started with the "noreset" option to
avoid crashes when closing X apps.
 # X -noreset

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

imx-gpu-viv: bump version to 5.0.11 p7.1

Upgrade to 5.0.11 P7.1 for i.MX6 with GPU

i.MX6QP qualified graphics driver(5.0.11P7.1).
Includes many of the bug fixes and stability improvements.
For more information refer to release notes from
http://www.freescale.com/imx6qp

libGAL_egl is no longer provided in p7.1.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/b3ef125

This package has been tested with both X11 and Framebuffer backends:
 # cd /usr/share/examples/viv_samples/vdk/
 # apitrace trace --api egl ./tutorial7
 # gmem_info
 ... display memory use per PID ...
 # apitrace replay tutorial7.trace
 # eglretrace tutorial7.trace

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gst1-plugins-bad: add install to staging directory

gst1-imx package links against libgstphotography for its imxv4l2src
plugin.

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libfslvpuwrap: bump to version 1.0.61

Changelog:
- Enable resolution change for H.263 and VP8

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/f11936b

This package has been implicitely tested through gstreamer as the
plugins rely on it for vpu decoding:
 # gst-launch-0.10 playbin uri=file:///root/tears_of_steel_1080p.webm
 # gst-launch-1.0 playbin uri=file:///root/tears_of_steel_1080p.webm
 # gst-launch-1.0 filesrc location=/root/tears_of_steel_1080p.webm ! \
   matroskademux ! imxvpudec ! imxipuvideosink

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libfslparser: bump to version 4.0.7

Changelog:
- Add h264 parser version.
- Support wide and 'ms 0x0 0x11' adpcm tag.
- Support hev1 tag for hevc video codec.
- TS segment fault.
- Improve mkv parser's loading time when create parser with large size
file.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/6406ad4
https://github.com/Freescale/meta-fsl-arm/commit/09a532c
https://github.com/Freescale/meta-fsl-arm/commit/371df94

This package has been implicitely tested through gstreamer as the 0.10
plugin relies on it:
 # gst-launch-0.10 playbin uri=file:///root/tears_of_steel_1080p.webm

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libfslcodec: bump to version 4.0.7

Changelog since 4.0.3:
- Add mx7 SoC support.
- Add test binary to release package.
- Fix segmentation fault issue with only standard build.
- Build std only for arm11 and arm12.
- Fix some mp3_dec noise issue of c code for arm11 and arm9.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/47b2677
https://github.com/Freescale/meta-fsl-arm/commit/c88fd22
https://github.com/Freescale/meta-fsl-arm/commit/3220a4c

This package has been implicitely tested through gstreamer as the 0.10
plugin relies on it:
 # gst-launch-0.10 playbin uri=file:///root/tears_of_steel_1080p.webm

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

freescale-imx: remove FREESCALE_IMX_VERSION

As all the packages now have a version which doesn't depend on the BSP
version.

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

imx-lib: bump to version 5.1

Change version to be independent of BSP version.  This limits
confusion on future BSP updates that do not include imx-lib updates.

Changelog since 3.14.28-1.0.0_ga:
- Add i.MX 7D support.
- Change LICENSE file.
- Enable pxp libraries for i.MX 6UltraLite.
- Fix rng and shara2 to exclude the 6UL platform to avoid build breaks

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/b299454
https://github.com/Freescale/meta-fsl-arm/commit/5f28af0

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

imx-kobs: bump to version 5.1

Use a different versioning schema based on package versioning
instead of BSP version. This way, if a new BSP version does
not bump this package, we can keep this package unchanged as well.

Changelog since 3.14.28-1.0.0_ga:
- Add fixes for maximum ecc strength for each platform.
- Add mx7 boot config support.
- Add i.MX 6UltraLite support.
- Join the ecc with real FCB data when not byte-aligned.
- Generalize the encode_bch_ecc function for different configurations.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/8d47a2e
https://github.com/Freescale/meta-fsl-arm/commit/70edf16

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

firmware-imx: bump to version 5.2

Use a different versioning schema based on package versioning
instead of BSP version. This way, if a new BSP version does
not bump this package, we can keep this package unchanged as well.

VPU firmware upgraded to v3.1.1_r46067
- Low Latency fixes for H.264 encoder
- Wrong frame number fix for H.264 decoder
- Decode failure fix for H.264 decoder with multi-SPS/PPS headers

The package adds the EPDC waveforms and the needed changes to properly
install those were done in the firmware-imx.mk file.

This patch is based on the Yocto equivalent:
https://github.com/Freescale/meta-fsl-arm/commit/d869826
https://github.com/Freescale/meta-fsl-arm/commit/075a5b1

This package has been implicitely tested through gstreamer as the
plugins rely on them for vpu decoding for instance:
 # gst-launch-1.0 filesrc location=/root/tears_of_steel_1080p.webm ! \
   matroskademux ! imxvpudec ! imxipuvideosink

Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
[Thomas:
 - fix typo in commit logs
 - remove ; \ and ; at the end of commands, since they were not
   needed.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs/raspberrypi: add dtb for compute module

[Peter: describe what 'CM' stands for, update comment]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

rpi-firmware: version bump

To match the previous kernel version bump.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

configs/raspberrypi{,2}: bump to kernel version 4.1.15

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

strongswan: unavailable for static builds

Uses dlopen in plugin_loader (not optional), fixes:
http://autobuild.buildroot.net/results/601/601d8dc1654d8733db49b195139e12437663034c/

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

vala: bump to version 0.30.1

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/autoconf-archive: bump to version 2015.09.25

Signed-off-by: Thomas Claveirole <thomas.claveirole@green-communications.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-headers: bump 4.{1, 3, 4}.x series

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Update copyright year

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux: bump default to version 4.4.1

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain-external: bump Linaro ARMEB to 2015.11-2

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas: only add the symlink with the old 2014.09 Linaro toolchain,
for the newer ones, it is no longer needed.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain-external: bump Linaro ARM to 2015.11-2

Runtime tested with Qemu 2.3.1 using a configuration based on
qemu_arm_vexpress_defconfig with BR2_ARM_ENABLE_VFP and
BR2_ARM_EABIHF selected

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas: only add the symlink with the old 2014.09 Linaro toolchain,
for the newer ones, it is no longer needed. This has been runtime
tested in Qemu.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: update uClibc-ng to 1.0.12

New release contains a security fix for the resolver library parts.
It contains exp10(), so some patches in buildroot might be obsolete,
when the buildroot toolchains are rebuilt.
It also contains a fix for the symbol clashing with bind9.

Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

uclibc: remove version selection

Since we now only support uClibc-ng, remove the version selection from
the uclibc package.

Note that the BR2_UCLIBC_VERSION_SUPPORTS_* hidden booleans, which
were only used to allow each uClibc version to specify which thread
implementation they support and on which architecture are removed. Now
such architecture dependencies are directly encoded in the "Thread
library implementation" choice.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

uclibc: remove support for the snapshot version

The snapshot version points to the original uClibc project, which is
dead. Moreover, we no longer support "snapshot" versions for any other
Buildroot component, so there is no reason to keep it for uClibc.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

uclibc: remove 0.9.33 version

The upstream project is dead, the 0.9.33 version requires tons of
patches, and uclibc-ng has now successfully replaced uclibc. It is
time to get rid of the 0.9.33 version.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

samba4: remove dependency on specific C libraries

samba4 relies on the $ORIGIN feature of the dynamic linker, which used
to not be implemented in old uClibc versions. However:

 - this feature is supported by glibc
 - this feature is supported by uClibc-ng, which is the only uClibc
   version we are going to support
 - this feature is supported by musl

Consequently, we can completely remove the dependency of samba4 on
certain C libraries.

Note that despite this commit, samba4 still cannot be chosen when the
musl C library is used, because samba4 requires native RPC support,
which musl doesn't provide.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

vlc: remove dependency on C library

vlc uses <spawn.h> which was not available in old uClibc
versions. However, since we are removing support for uClibc 0.9.33, we
can get rid of such dependency. In addition, <spawn.h> is provided by
musl, and therefore VLC can be enabled with this C library.

Consequently, this commit completely removes any C library dependency for
the vlc package.  The only special case that needs to be handled is the
Blackfin external toolchain from Analog Devices, which still uses an old
uClibc version that doesn't provide <spawn.h>, but as vlc uses fork() we add
a depends on BR2_USE_MMU (which covers the blackfin toolchain).

[Peter: add BR2_USE_MMU dependency]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Romain Naour <romain.naour@gmail.com>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mongrel2: allow using uClibc but only on certain architectures

mongrel2 uses the {get,make,swap}context functions:

 - With glibc, no problem, they are available on all supported
   architectures

 - With uClibc, they are available only on a subset of the
   architectures. Until now, only BR2_UCLIBC_VERSION_SNAPSHOT
   configurations were allowed to select mongrel2, but we are going to
   get rid of the uClibc snapshot version, and uClibc-ng is as capable
   as the uClibc snapshot. However, only certain architectures have
   the *context() functions.

 - With musl, there is no *context() support.

Since this dependency is quite complicated, we introduce a
BR2_PACKAGE_MONGREL2_LIBC_SUPPORTS hidden boolean to encode which C
libraries are supported.

Also, listing the supported uClibc architectures would be too long in
the comment, so we simply indicate that the package needs uClibc or
glibc.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

lttng-babeltrace: make available for all C libraries

Due to posix_fallocate() being unavailable in uClibc 0.9.33,
lttng-babeltrace was marked as available only for uClibc snapshot and
glibc. However:

 - lttng-babeltrace builds fine with musl
 - lttng-babeltrace builds fine with uClibc-ng
 - we're going to get rid of uClibc 0.9.33 support
 - the only toolchain left with an old uClibc version is the Blackfin
   Analog Devices toolchain, and lttng-babeltrace cannot be enabled on
   non-MMU platforms

Conclusion: We can enable lttng-babeltrace on all C libraries, and no
longer require any condition. This commit adjusts the lttng-babeltrace
package accordingly.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Romain Naour <romain.naour@gmail.com>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libunwind: fix C library dependency

The libunwind package currently dependency on glibc *or* uclibc
snapshot. However, we are going to remove the support for uclibc
snapshot, and uClibc-ng has equivalent functionality as uclibc
snapshot. Moreover, musl is also capable of building libunwind for
certain architectures.

Therefore, this commit reworks the architecture dependencies of
libunwind, to make it available on all architectures for which it is
supported, depending on the capabilities of the different C libraries,
and the implementation of libunwind on each architecture.

On some architectures, libunwind uses the C library *context()
functions, which are not provided by musl at all, and not provided by
uClibc on all architectures. But on some other architectures,
libunwind does not use the C library *context() functions, which
explains why it can be built with musl on such architectures.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain-external: bump Linaro AArch64 to 2015.11-2

Runtime tested with Qemu 2.3.1 using qemu_aarch64_virt_defconfig.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas: only add the symlink with the old 2014.09 Linaro toolchain,
for the newer ones, it is no longer needed. This has been runtime
tested in Qemu.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

website: add reference to submitting-patches to Contribute tab

In the hope of improving the quality of patches send by newcomers,
add a reference to the submitting-patches section of the manual
to the Contribute tab of the website.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

README: add reference to submitting-patches

In the hope of improving the quality of patches sent by newcomers,
add a reference to the submitting-patches section of the manual
to the top-level README file.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs: qemu_arm_nuri: enable vfp

In order to be abble to select BR2_ARM_FPU_VFPV3D16, BR2_ARM_ENABLE_VFP
must be selected first.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs: qemu_arm_vexpress: enable vfp

In order to be abble to select BR2_ARM_FPU_VFPV3D16, BR2_ARM_ENABLE_VFP
must be selected first.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/Makefile.in: fix typo

Commit dc95d50fe3ee (correct gettext handling for musl) introduced a last
minute typo, fix that.

Thanks to Thomas Petazzoni for noticing.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Revert "madplay: fix musl gettext support detection"

This reverts commit a0a244d26d8e7e7e5465c3e6e9fcd1c31e2c178d.

As this is now handled globally in TARGET_CONFIGURE_ARGS, this can be
reverted here.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/Makefile.in: correct gettext handling for musl

Based on a patch by Bernd Kuhls.

The AM_GNU_GETTEXT autotools macro misdetects musl gettext support as it
checks for internal glibc symbols.  Work around it by forcing libc gettext
support when musl is used for the supported gettext api levels.

As this is a generic issue for any package using AM_GNU_GETTEXT, add it to
the global TARGET_CONFIGURE_ARGS instead of for each affected package.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/shairport-sync: bump to version 2.8.0

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/unionfs: fix musl build

Fixes
http://autobuild.buildroot.net/results/020/020951a32ace3b4bca12a888eaaeb8b6c5633e4f/

[Peter: notice patch is now upstream:
 https://github.com/rpodgorny/unionfs-fuse/commit/f64b094eba93b042a1ed5964]
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

manual: minor fixes in the graph-size documentation

Fix some typos and references to a size-stats 'target' (the script is called
'size-stats' but the make target is 'graph-size').

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/squashfs: fix musl build

The build errors were not yet found by the autobuilders:

action.c: In function ‘name_fn’:
action.c:1911:29: error: ‘FNM_EXTMATCH’ undeclared (first use in this function)
     FNM_PATHNAME|FNM_PERIOD|FNM_EXTMATCH) == 0;)
                             ^
pseudo.c: In function ‘read_pseudo_def’:
pseudo.c:435:11: error: ‘S_IFBLK’ undeclared (first use in this function)
   mode |= S_IFBLK;
           ^

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain-external: create symlink ARCH_LIB_DIR->lib

Currently, following symbolic links are created in both target and
staging directories:
- lib(32|64) --> lib
- usr/lib(32|64) --> lib

The decision for lib32 or lib64 is based on the target architecture
configuration in buildroot (BR2_ARCH_IS_64).

In at least one case this is not correct: when building for a Cavium Octeon
III processor using the toolchain from the Cavium Networks SDK, and
specifying -march=octeon3 in BR2_TARGET_OPTIMIZATION, libraries are expected
in directory 'lib32-fp' rather than 'lib32' (ABI=n32; likewise for
lib64-fp in case of ABI=n64)

More generally the correct symbolic link is from (usr/)${ARCH_LIB_DIR}->lib.
However, feedback from Arnout Vandecappelle is that there are packages that
do depend on the lib32/lib64 symlink, even if ARCH_LIB_DIR is different.
Hence, these links must be kept.

Fix the problem as follows:
- For internal toolchains: no change
- For external toolchains: create a symlink ARCH_LIB_DIR->lib if
  (usr/)ARCH_LIB_DIR does not exist yet.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: "Yann E. Morin" <yann.morin.1998@free.fr>
Cc: Romain Naour <romain.naour@gmail.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Tested-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain-external: improve sysroot rsync if ARCH_LIB_DIR != lib/lib32/lib64

The copy_toolchain_sysroot helper in toolchain/helpers.mk performs an
rsync of various directories from the extracted external toolchain to the
corresponding directory in staging.

The relevant (simplified) snippet is:
for i in etc $${ARCH_LIB_DIR} sbin usr usr/$${ARCH_LIB_DIR}; do \
rsync -au --chmod=u=rwX,go=rX --exclude 'usr/lib/locale' \
--exclude '/lib/' --exclude '/lib32/' \
--exclude '/lib64/' \
$${ARCH_SYSROOT_DIR}/$$i/ $(STAGING_DIR)/$$i/ ; \
done ; \

The exclusion logic of lib/lib32/lib64 has originally been added by commit
5628776c4a4d29d0715633ea463b64cc19e19c5a with the purpose of only copying
the relevant usr/lib* directory from the toolchain to staging, instead of
all. For example, if ARCH_LIB_DIR is 'lib64', then only usr/lib64 would be
copied and usr/lib and usr/lib32 are ignored. It works by ignoring any
lib/lib32/lib64 subdirectory on the rsync of 'usr' and then separately
copying usr/{lib,lib32,lib64} as appropriate. (The exclusion rules only have
impact on the files beneath the main source directory.)

However, ARCH_LIB_DIR can take other values than (lib, lib32, lib64), for
example lib32-fp or lib64-fp (Octeon III toolchain with -march=octeon3). In
the existing code, the rsync for 'usr' would then already copy these lib
directories, and the next rsync for 'usr/$${ARCH_LIB_DIR}' does nothing.

By itself, this is not a very big problem: the staging directory simply has
some extra directories. However, a subsequent patch will create a staging
symlink from $${ARCH_LIB_DIR} to lib. The first rsync would then overwrite
that symlink with the real directory usr/$${ARCH_LIB_DIR} from the
toolchain, which is not correct.

Assuming the patch that creates the symlink ARCH_LIB_DIR->lib is applied,
the original situation after 'make clean toolchain' with an
ARCH_LIB_DIR=lib32-fp is:

$ ls -ld output/staging/{,usr/}lib* output/target/{usr/,}lib*
drwxr-xr-x 2  4096 May 26  2015 output/staging/lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/staging/lib32 -> lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/staging/lib32-fp -> lib
drwxr-xr-x 2  4096 Jan 20 13:47 output/staging/usr/lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/staging/usr/lib32 -> lib
drwxr-xr-x 4  4096 May 26  2015 output/staging/usr/lib32-fp
drwxr-xr-x 4  4096 May 26  2015 output/staging/usr/lib64-fp
drwxr-xr-x 4  4096 May 26  2015 output/staging/usr/libexec
drwxr-xr-x 3  4096 May 26  2015 output/staging/usr/libexec32
drwxr-xr-x 3  4096 May 26  2015 output/staging/usr/libexec32-fp
drwxr-xr-x 3  4096 May 26  2015 output/staging/usr/libexec64-fp
drwxr-xr-x 2  4096 Jan 20 13:48 output/target/lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/target/lib32 -> lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/target/lib32-fp -> lib
drwxr-xr-x 2  4096 Jan 20 13:48 output/target/usr/lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/target/usr/lib32 -> lib
lrwxrwxrwx 1     3 Jan 20 13:47 output/target/usr/lib32-fp -> lib

Notice how usr/lib32-fp is not a symlink but a directory, and the presence
of an unnecessary directory usr/lib64-fp.

This patch improves the rsync exclusion rules by excluding any lib*
directory on the first rsync. As this would also exclude any
libexec/libexec32/... directory, explicitly include them first (first match
takes precedence). This (as is already the case today) results in more
usr/libexec* directories than needed, but it is not touched by this patch.

With the fix applied, the situation becomes:

drwxr-xr-x 2 4096 May 26  2015 output/staging/lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/staging/lib32 -> lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/staging/lib32-fp -> lib
drwxr-xr-x 4 4096 May 26  2015 output/staging/usr/lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/staging/usr/lib32 -> lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/staging/usr/lib32-fp -> lib
drwxr-xr-x 4 4096 May 26  2015 output/staging/usr/libexec
drwxr-xr-x 3 4096 May 26  2015 output/staging/usr/libexec32
drwxr-xr-x 3 4096 May 26  2015 output/staging/usr/libexec32-fp
drwxr-xr-x 3 4096 May 26  2015 output/staging/usr/libexec64-fp
drwxr-xr-x 2 4096 Jan 20 14:27 output/target/lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/target/lib32 -> lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/target/lib32-fp -> lib
drwxr-xr-x 2 4096 Jan 20 14:27 output/target/usr/lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/target/usr/lib32 -> lib
lrwxrwxrwx 1    3 Jan 20 14:27 output/target/usr/lib32-fp -> lib

For cases where ARCH_LIB_DIR is one of lib, lib32 or lib64 this fix
makes no difference, and likewise for internal toolchains.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Cc: Samuel Martin <s.martin49@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain-external: don't exclude too much lib in sysroot rsync

The copy_toolchain_sysroot helper in toolchain/helpers.mk performs an
rsync of various directories from the extracted external toolchain to the
corresponding directory in staging.

The relevant (simplified) snippet is:
    for i in etc $${ARCH_LIB_DIR} sbin usr usr/$${ARCH_LIB_DIR}; do \
            rsync -au --chmod=u=rwX,go=rX --exclude 'usr/lib/locale' \
                    --exclude lib --exclude lib32 --exclude lib64 \
                    $${ARCH_SYSROOT_DIR}/$$i/ $(STAGING_DIR)/$$i/ ; \
    done ; \

The exclusion logic of lib/lib32/lib64 has been added by commit
5628776c4a4d29d0715633ea463b64cc19e19c5a with the purpose of only copying
the relevant usr/lib* directory from the toolchain to staging, instead of
all. For example, if ARCH_LIB_DIR is 'lib64', then only usr/lib64 would be
copied and usr/lib and usr/lib32 are ignored. It works by ignoring any
lib/lib32/lib64 subdirectory on the rsync of 'usr' and then separately
copying usr/{lib,lib32,lib64} as appropriate. (The exclusion rules only have
impact on the files beneath the main source directory.)

However, on the rsync of 'usr', ANY of the following directories AND files
would be excluded:
    lib/
    lib
    lib32/
    foobar/something/lib/
    something-else/lib64/

while it is only the intention to skip directories directly under usr.

Therefore, add a leading (to restrict the scope to first-level) and trailing
(to restrict to directories) slash to the exclude pattern. From 'man rsync':

    - if the pattern starts with a / then it is anchored to [..] the root of
      the transfer.
    - if the pattern ends with a / then it will only match a directory, not
      a regular file, symlink, or device.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Cc: Samuel Martin <s.martin49@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>