gcc: rename option for ARC gcc

The ARC gcc version is now based on gcc 6.x and no longer gcc 4.8.x,
which makes the option BR2_GCC_VERSION_4_8_ARC a bit irrelevant, as is
the prompt of this option.

This commit therefore renames this option to BR2_GCC_VERSION_ARC, and
adjust its prompt as well.

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

package/nvidia-driver: update version

This new version brings in support for egl-wayland, the EGL extensions
aimed at making it possible to implement Wayland servers and clients. As
such, nvidia-driver becomes the second EGL implementation in Buildroot
that can act as a libegl provider with egl-wayland extensions.

In this version, it becomes possible to use our kernel-module infra,
with just a little few minor tricks: we need just specify the Linux
source and build trees (they are the same for us) and the list of
modules to build. We still need a little patch against the Kbuild files.

We also get rid of the LIBS_NO_VERSION trick and always use complete
filenames, as more libs are now packaged with different version in their
filenames, and even some with no version at all.

When installing libs, we switch from a shell loop to a make foreach
loop, which is easier to handle. It has the side-effect (and advantage)
of displaying the install commands for each library, rather than a
single biggish one, so it is easier to see what goes wrong. This also
means that an error in each phase of the install (the copy of the files
then each symlink) can be caught more easily (it was not previously):
each sequence is now its own make command; we need not use "|| exit 1"
after each command, even in a if block, because the if blocks returns
with the exit code of the last command in it; e.g. if an ln fails, the
if-block in which it is enclosed will return the exit code of ln, and
make will catch it.

Similarly for the X driver modules and each of the programs installed:
we now can catch any failure in the isntall of those.

All of this somewhat simplifies the .mk. It is a little bit longer, but
the structure is saner and more explicit.

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

python-pyqt5: new package

Signed-off-by: Manuel Vögele <develop@manuel-voegele.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/socat: fix build issue

Fix two build issue when socat is build with musl:
  - Set NETDB_INTERNAL locally since it's not provided by musl
  - remove if_tun.h kernel header

Fixes:
http://autobuild.buildroot.net/results/029/02977d3844ef62b1e9cbfbab244593632757c5e2

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

uclibc: add upstream patch to fix glog build on bfin

glog has been failing on bfin because of issue in uclibc which is now
fixed in upstream.

Fixes:
  http://autobuild.buildroot.net/results/a10/a10ed48e6eb8411a3d8372f57c05fd11130da0e0/

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

vala: bump to version 0.34.0

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

atk: bump to version 2.22.0

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

gdk-pixbuf: bump to version 2.36.0

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

libsoup: bump to version 2.56.0

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

gvfs: bump to version 1.30.0

Disable admin since it requires polkit-gobject + libcap.

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

glib-networking: bump to version 2.50.0

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

libglib2: bump to version 2.50.0

It now requires util-linux libmount so add the dependency/select.

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

ushare: fix build with NLS support

Add patch to fix build errors when NLS support is enabled.

Fixes:

  http://autobuild.buildroot.net/results/19d/19d67dd43e5a313c77e4be97ecb9811ffa52f797/

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/mesa3d: add support for Intel Vulkan driver

The Vulkan intel driver depends on the i965 dri driver:
https://cgit.freedesktop.org/mesa/mesa/tree/configure.ac?h=12.0#n1653

and a sha1 implementation:
https://cgit.freedesktop.org/mesa/mesa/tree/configure.ac?h=12.0#n1656

The Vulkan driver needs linux/memfd.h
https://cgit.freedesktop.org/mesa/mesa/tree/src/intel/vulkan/anv_allocator.c?h=12.0#n30
which is not available in kernel headers older than 3.18:
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/log/include/uapi/linux/memfd.h?id=refs/tags/v3.18.36

The Vulkan driver makes use of ifunc
https://cgit.freedesktop.org/mesa/mesa/tree/src/intel/vulkan/anv_entrypoints_gen.py?h=12.0#n287
which is not available on uClibc:
https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config.gcc;h=82cc9a9959b5ab57c0b8779e054b80cdb95f169b;hb=gcc-6-branch#l1485
https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=e6cdd6b1755033e8f416efaa4334d1294c0a43c6

The Vulkan driver makes use of static_assert
https://cgit.freedesktop.org/mesa/mesa/tree/src/intel/vulkan/anv_private.h?h=12.0#n153

Compiling the Vulkan driver with uClibc and musl fails, therefore this
driver is glibc-only.

Although the configure script does not check for dri3 support if the
Intel Vulkan driver is enabled it needs it nonetheless:
https://cgit.freedesktop.org/mesa/mesa/tree/src/intel/vulkan/anv_wsi_x11.c?h=12.0#n682

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

package/mesa3d: provides libgl not only with DRI drivers

This patch is a follow-up for
https://github.com/buildroot/buildroot/commit/999d4a263118ab8581190d3e3c8bdc471896628f

libGL.so and support files are not only available with DRI drivers but
also with Gallium drivers, in fact mesa3d can be built without any
driver enabled - unless X.org is disabled it will provide libGL.so.

Tested using this defconfig:
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_TOOLCHAIN_BUILDROOT_CXX=y
BR2_PACKAGE_KODI=y
BR2_PACKAGE_KODI_VISUALISATION_WAVEFORHUE=y
BR2_PACKAGE_MESA3D=y
BR2_PACKAGE_XORG7=y

Fixes
http://autobuild.buildroot.net/results/d97/d97a80e6de066a7dea08b284eab24fc44b11d661/

kodi-visualisation-waveforhue detects OpenGL installed by mesa3d but the
kodi package itself does not select libglew before because mesa3d was
not listed as a provider for libgl in the buildroot build system.

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

aufs-util: new package

Signed-off-by: Christian Stewart <christian@paral.in>
[Atul:
 - Updated the default to empty.
 - Updated the license to GPLv2.
 - Used patches instead of SED commands.
 - CC and LD are replaced by TARGET_CC and TARGET_LD.]
Signed-off-by: Atul Singh <atul.singh.mandla@rockwellcollins.com>
[yann.morin.1998@free.fr:
  - drop the headers_install command
  - instead, instruct the kernel to install them
  - add patches 0002..0004
  - properly build the host tools
  - don't depend on the aufs extension, for kernel already patched
  - fix comment]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
[Thomas:
 - Remove the patch 0004-no-override-readdir64.patch, since it no longer
   applies to the latest version of aufs-util. Instead, add a post-patch
   hook that tweaks the Makefile to remove the problematic rdu64.o from
   the build process.
 - Use directly the kernel headers from the kernel source tree, rather
   than having to install them. This only requires a minor trick to
   define "__user" to empty.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

aufs: new kernel extension

Signed-off-by: Christian Stewart <christian@paral.in>
[Atul:
 - Removed the duplicate conditional block.
 - Updated the license to GPLv2.
 - Removed the visibilty of package from menuconfig.
 - Removed dependencies.
 - Removed the comment.
 - Changed the name of variable from BR2_PACKAGE_AUFS_STANDALONE_VERSION
   to BR2_LINUX_KERNEL_EXT_AUFS_VERSION.
 - Removed the AUFS_INSTALL_STAGING  and AUFS_INSTALL_TARGET  variables.
 - Removed the BR2_PACKAGE_AUFS_3X and BR2_PACKAGE_AUFS_4X variables.]
Signed-off-by: Atul Singh <atul.singh.mandla@rockwellcollins.com>
[yann.morin.1998@free.fr:
  - do not fail on version check if aufs ext is disabled
  - check for empty version
  - squash aufs package and linux extension in one patch
  - fail if the kernel already has aufs support
  - simplify handling of version]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
[Thomas:
 - Fix the apply patch logic, it was using a non-existent
   AUFS_VERSION_MAJOR variable. BR2_PACKAGE_AUFS_SERIES is used
   instead.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs/warp7: Add Wifi support

Warp7 has a BRCM43430 Wifi chipset. Add support for it.

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-psutil: bump to version 4.3.1

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

supertuxkart: Constrain to x86 architecture

Our build system has detected that angelscript can't be
cross-compiled properly. This has been also reported to
supertuxkart [1], and although the developers were open
to take patches fixing STK embedded angelscript, it was
pointed out that non-x86 architectures were not officially
supported and of little interest [2].

Hence, let's constrain the package selection and the host
architecture to x86.

[1] https://github.com/supertuxkart/stk-code/issues/2569
[2] https://github.com/supertuxkart/stk-code/pull/2195

Fixes: http://autobuild.buildroot.net/results/655/655f046b3bb5f7a4ea492328b608e409d6769689/
Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-cffi: bump to version 1.8.3

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-netifaces: bump to version 0.10.5

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-lxml: bump to version 3.6.4

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

manual: get rid of the lists of packages

We currently have four lists of packages in the manual:
  - the non-virtual target packages,
  - the virtual target packages,
  - the host packages,
  - the deprecated features.

Those list take more than half of the manual. They do not serve much
purpose except to show off.

After the recent discussion on the list [0], remove them all.

We can now get rid of our biggish and complex generating script (and its
companion library kconfiglib).

[0] http://lists.busybox.net/pipermail/buildroot/2016-September/171199.html

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>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/Makefile.in: remove unused STRIP_STRIP_ALL variable

This variable has been unused for a long time, so we can get rid of its
definition.

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

linux: use INSTALL_MOD_STRIP=1 to strip modules

We used to do a special handling of Linux kernel modules when stripping
target binaries because there's some special precious data in modules
that we must keep for them to properly operate. This is for example true
for stack unwinding data etc.

It turned out there're cases when our existing "strip --strip-unneeded"
doesn't work well. For example this removes .debug_frame section used by
Linux on ARC for stack unwinding, refer to [1] and [2] for more details.

Now Linux kernel may strip modules as a part of "modules_install" target
if INSTALL_MOD_STRIP=1 is passed in command line. And so we'll do
allowing kernel decide how to strip modules in the best way.

Still note as of today Linux kernel strips modules uniformly for all
arches with "strip" command, so this commit alone doesn't solve
mentioned problem but it opens a possibility to add later a patch to the
kernel which will strip modules for ARC differently - and that's our
plan for mainline kernel.

[1] https://github.com/foss-for-synopsys-dwc-arc-processors/toolchain/issues/86
[2] http://lists.busybox.net/pipermail/buildroot/2016-September/172161.html

Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Daniel Mentz <danielmentz@google.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

sam-ba: correct license

The sam-ba binary is distributed under a proprietary license. TCL and some
applets code is under a BSD-2c like license, while some other applets are
under BSD-4c.

Also correct license files.

Cc: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

util-linux: add -lcurses to LIBS when static linking

When linking statically with -lreadline we also need to specify
-lcurses. We were already passing LIBS=-intl to fix gettext related
issues, so we refactor how LIBS is passed so that several values can be
passed.

It is worth mentioning that:

 1. Passing LIBS= at configure time is not sufficient, because the
    configure script does LIBS="" at its end. So in order for the LIBS
    value to also be effective during the build, it needs to be passed
    in MAKE_OPTS as well.

 2. LIBS=-lintl was only passed at build time (it was not needed for any
    of the configure tests). However, LIBS=-lncurses is needed both at
    configure time (for configure tests) and at build time. Therefore,
    the new variable UTIL_LINUX_LIBS gets passed as the LIBS value at
    both configure time and build time.

Fixes:
  http://autobuild.buildroot.net/results/70e0a273e699c73c6b2ab2a19c8f7ebb1ddee6cde
  http://autobuild.buildroot.net/results/6a7faf8adbffd9437c0117b9f0e89684c24dad23

Signed-off-by: Chris Packham <judge.packham@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

google-breakpad: needs thread support

When google-breakpad was enabled in uClibc configurations, the
dependency on threads was forgotten.

Fixes:

  http://autobuild.buildroot.net/results/6e6bc845314100f2deb8ed06d2a6373ccf715ffe/

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

package/kvm-unit-tests: new package

The unit tests are tiny guest operating systems that generally execute
only tens of lines of C and assembler test code in order to obtain its
PASS/FAIL result. Unit tests provide KVM and virt hardware functional
testing by targeting the features through minimal implementations of
their use per the hardware specification. The simplicity of unit tests
make them easy to verify they are correct, easy to maintain, and easy
to use in timing measurements. Unit tests are also often used for
quick and dirty bug reproducers. The reproducers may then be kept as
regression tests. It's strongly encouraged that patches implementing
new KVM features are submitted with accompanying unit tests.

Signed-off-by: Cyril Bur <cyrilbur@gmail.com>
[Thomas:
 - order architecture dependencies in Config.in alphabetically.
 - rewrap Config.in help text, lines were too long
 - add an empty line between the package description and the upstream
   project URL in the Config.in help text
 - don't make KVM_UNIT_TESTS_ARCH default to $(ARCH). This was not
   correct for i386 for example. Instead, just handle the few
   architectures that the package supports.
 - remove useless double quotes in variable definitions.
 - remove --prefix="$(TARGET_DIR)" from CONF_OPTS. It was installing
   everything in /share/ and not /usr/share/, and setting the prefix to
   TARGET_DIR at configure time is not good. Instead, pass DESTDIR at
   installation time.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/php: set MySQL socket path

Set the PHP default MySQL socket path to match what the
mysql package is using.

Signed-off-by: Floris Bos <bos@je-eigen-domein.nl>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/mysql: add MYSQL_SOCKET variable

Add MYSQL_SOCKET variable with MySQL socket location

Signed-off-by: Floris Bos <bos@je-eigen-domein.nl>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

raspberrypi3: fix serial console (load pi3-miniuart-bt overlay)

- enable BR2_PACKAGE_RPI_FIRMWARE_INSTALL_DTB_OVERLAYS in raspberrypi3_defconfig
- add copy of rpi-firmware/overlays directory to boot partition in genimage-raspberrypi3.cfg
- enhance post-image.sh script to add 'dtoverlay=pi3-miniuart-bt' on request
- add BR2_ROOTFS_POST_SCRIPT_ARGS="--add-pi3-miniuart-bt-overlay" in raspberrypi3_defconfig

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

package/cryptsetup: move depends on above select

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

python3: do not use hg when available

During the execution of its configure script, Python 3 tries to find an
available "hg" installation, and if available, will try to use it to get
information from the version control system. To do this, it tries to
communicate over the network, potentially over ports that are blocked,
causing the build to halt. This was reported by a user as part of bug
7802.

To solve this, we simply make the Python script use /bin/false as the
"hg" program.

Fixes bug #7802 for the python3 package.

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

python: do not use hg or svn when available

During the execution of its configure script, Python tries to find an
available "hg" and "svn" installation, and if available, will try to use
them to get information from the version control system. To do this, it
tries to communicate over the network, potentially over ports that are
blocked, causing the build to halt. This was reported by a user as part
of bug #7802.

To solve this, we simply make the Python script use /bin/false as the
"hg" and "svn" programs.

Fixes bug #7802 for the python package.

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

configs: add defconfig for TS-4900

The TS-4900 uses a custom Linux 4.X from Technologic Systems:
https://github.com/embeddedarm/linux-3.10.17-imx6/tree/imx_4.1.15_1.0.0_ga
Indeed, it is not supported by mainline Linux but this work is ongoing.

This patch is inspired by a previous work for the TS-4800.
A post image script is provided to generate an image that can be
directly written to an SD card. It also adds an fpga wifi module.

Signed-off-by: Mathieu Audat <mathieu.audat@savoirfairelinux.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

ts4900-fpga: add ts4900-fpga package

This package is responsible from downloading and deploying bitstream for
TS-4900's FPGA. It implements clocks, UART MUX, and GPIOs. One of these
GPIOs is used to enable the wifi module.

It is loaded by U-Boot during the boot sequence, the default U-boot
script expects to find it in the /boot folder.

Signed-off-by: Mathieu Audat <mathieu.audat@savoirfairelinux.com>
[Thomas:
 - Use "config" instead of "menuconfig" since there's a single option
   and no sub-options.
 - Add BR2_arm dependency, since it's only used on a specific i.MX6
   platform.
 - Implement TS4900_FPGA_EXTRACT_CMDS that copy the file to the package
   build directory $(@D)
 - Change the install command to copy from $(@D) instead of from
   $(DL_DIR)
 - Change the install command to remove the 'mkdir -p
   $(TARGET_DIR)/boot' and instead just use the -D option of the install
   program that creates the destination directory if it doesn't exist.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-firmware: add option for Marvell usb8801 firmware

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

package/efl: bump to 1.18.1

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

python-daemon: bump to version 2.1.1

Bump to the latest version, to get python3 support.

Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain/wrapper: extend paranoid check to -isystem

Some packages, like libbsd, use -isystem flags to provide so-called
overrides to the system include files. In this particular case, this
is used in a .pc file, then used by antoher package; pkgconf does not
mangle this path; and eventually that other package ends up using
/usr/include/bsd to search for headers.

Our current toolchain wrapper is limited to looking for -I and -L, so
the paranoid check does not kick in.

Furthermore, as noticed by Arnout, there might be a bunch of other
so-unsafe options: -isysroot, -imultilib, -iquote, -idirafter, -iprefix,
-iwithprefix, -iwithprefixbefore; even -B and --sysroot are unsafe.

Extend the paranoid check to be able to check any arbitrary number of
potentially unsafe options:

  - add a list of options to check for, each with their length,
  - iterate over this list until we find a matching unsafe option.

Compared to previously, the list of options include -I and -L (which we
already had) extended with -idirafter, -iquote and -isystem, but leaving
all the others noticed by Arnout away, until we have a reason for
handling them.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain/wrapper: display options leading to a paranoid failure

Current, we only display the path that causes the paranoid failure. This
is sufficient, as we can fail only for -I and -L options, and it is thus
easy to infer from the path, which option is the culprit.

However, we're soon to add a new test for the -isystem option, and then
when a failure occurs, we would not know whether it was because of -I or
-isystem. Being able to differentiate both can be hugely useful to
track down the root cause for the unsafe path.

Add two new arguments to the check_unsafe_path() function: one with the
current-or-previous argument, one to specify whether it has the path in
it or not. Print that in the error message, instead of just the path.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/systemd: fix build with old toolchains

Toolchains using glibc-2.18 or older do not define O_TMPFILE, which
causes build failures on some archs.

systemd has a definition for O_TMPFILE if it is missing, but only
defines it for i386 or x86_64. Furthermore, the header defining it is
not included everywhere O_TMPFILE is used.

Fix that with three patches backported from upstream:
  - include the needed header where it is needed (he!),
  - define O_TMPFILE for all archs, according to linux-4.8rc3,
  - no longer guard against undefined O_TMPFILE in fileio.

Upstream merge commit:
    https://github.com/systemd/systemd/commit/4a13100c6a5a0a4b793e90bd43d21c3696c42d46

Fixes:

  http://autobuild.buildroot.net/results/b0067e72ffcbbe1db9ef49ab297cece951345aeb/

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

olimex_a20_olinuxino_lime: bump u-boot to 2016.09

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain: Bump ARC tools to arc-2016.09-eng013

As described at:
4520524ba055706236db9f00dd79f1b2e2e87fde
this commit continues a series of updates of ARC tools.
This time we're updating tools to arc-2016.09-eng013.

This engenering build contains some important GCC updatesthat fixes
some failures for ARC, such as:
http://autobuild.buildroot.net/results/f80/f80ad8a07d9d58b46592c2c99b227197b0f808d3//
http://autobuild.buildroot.net/results/e59/e59e6592fc697cceda359e421387dd905ff04701//
http://autobuild.buildroot.net/results/521/52157aa2b5d75b5338f3bde7ae5beb3d300283d3//

We still keep GDB as it is of arc-2016.03 release because there're some
issues we'd like to resolve before releasing it to wider audience.

So again note this is next engineering builds of arc-2016.09 series
and it might have all kinds of breakages, please don't use it for
production builds.

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

boot/uboot: allow to build multiple U-Boot images

Sometimes it is desired to build multiple U-boot images. E.g. one to
save into flash memory and one for serial load. So far this was not
possible.

This change allows to select any combination of the target formats. They
are all copied to the image folder.

Signed-off-by: Petr Kulhavy <brain@jikos.cz>
[Thomas:
 - Handle the BR2_TARGET_UBOOT_FORMAT_BIN, which wasn't handled, and
   therefore u-boot.bin was not copied when
   BR2_TARGET_UBOOT_FORMAT_BIN=y.
 - Rename UBOOT_BIN to UBOOT_BINS, since it can now contain multiple
   values.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/efl: enable elput support

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas:
 - add --disable-elput in the !BR2_PACKAGE_EFL_ELPUT case
 - add missing dependency on libxkbcomm]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/efl: add OpenGL/OpenGLES support

Allow to enable graphic acceleration using OpenGL or OpenGLES with efl
libraries.

The full OpenGL option depend on X11, because full OpenGL means GLX,
which means X11. Also select efl xlib support when full OpenGL is
selected.

Enable OpenGL ES when EGL API is enabled, otherwise the build fail with
this error:
configure: error: OpenGL ES requires EGL, please add --enable-egl to your configure options to switch to EGL + OpenGL ES.

With the upcomming Wayland support in EFL Buildroot packaging,
OpenGL ES support is recommended.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas:
 - use a single ifeq/else ifeq/else ifeq/endif block
 - remove the "default" for the choice, move the "none" choice at the
   end, and simply rely on the first choice having its dependencies met
   being automatically enable by kconfig. So OpenGL first, then
   OpenGLES, and then none.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

core: distclean is a noconfig target

When the .config contains invalid configuration *(e.g. two providers for
a same virtual package), it is not possible to run "make distclean"
because the .config file is sourced and packages interpreted in this
case.

Add distclean to the noconfig list, so that we can run it in all cases.

However, this meand that DL_DIR is no longer set, and thus the default
download location never removed. We fix that by always removing the
download location, so that if it is the one configured we still remove
it (no change) and if it is not the one configured, we remove an
non-existing location and leave the user's location intact (no change
either).

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

core: don't build host-cmake if it is available on the build host

Currently all cmake packages depend on host-cmake. Unfortunately
host-cmake takes a long time to configure and build: almost 7 minutes
on a dual-core i5 with SSD. The time does not change even with ccache
enabled.

Indeed, building host-cmake is avoidable if it is already installed on
the build host: CMake is supposed to be quite portable, and the only
patch in Buildroot for the CMake package seems to only affect
target-cmake.

Thus we automatically skip building host-cmake and use the one on the
system if:
 - cmake is available on the system and
 - it is recent enough.

First, we leverage the existing infrastructure in
support/dependencies/dependencies.mk to find out whether there's a
suitable cmake executable on the system. Its path can be passed in the
BR2_CMAKE environment variable, otherwise it defaults to "cmake". If
it is enabled, found and suitable then we set BR2_CMAKE_HOST_DEPENDENCY
to empty; otherwise we set BR2_CMAKE_HOST_DEPENDENCY to 'host-cmake' and
override BR2_CMAKE with "$(HOST_DIR)/usr/bin/cmake" to revert to using
our own cmake (the old behaviour).

Then in pkg-cmake.mk we replace the hard-coded dependency on host-cmake
to using the BR2_CMAKE_HOST_DEPENDENCY variable, and we use $(BR2_CMAKE)
instead of $(HOST_DIR)/usr/bin/cmake.

Unlike what we do for host-tar and host-xzcat, for host-cmake we do
not add host-cmake to DEPENDENCIES_HOST_PREREQ. If we did, host-cmake
would be a dependency for _any_ package when it's not installed on the
host, even when no cmake package is selected.

Cmake versions older than 3.0 are affected by the bug described and
fixed in Buildroot in ef2c1970e4bf ("cmake: add patch to fix Qt mkspecs
detection"). The bug was fixed in upstream CMake in version 3.0 [0].

Amongst all the cmake packages currently in Buildroot, the currently
highest version mentioned in cmake_minimum_required() is 3.1 (grantlee
and opencv3).

Thus we use 3.1 as the lowest required cmake for now, until a package is
bumped, or a new package added, with a higher required version.

[0] https://cmake.org/gitweb?p=cmake.git;h=e8b8b37ef6fef094940d3384df5a1d421b9fa568

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Samuel Martin <s.martin49@gmail.com>
Cc: Davide Viti <zinosat@tiscali.it>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Tested-by: Romain Naour <romain.naour@gmail.com>
[yann.morin.1998@free.fr:
  - simplify logic in check-host-cmake.mk;
  - set and use BR2_CMAKE_HOST_DEPENDENCY, drop USE_SYSTEM_CMAKE;
  - bump to cmake 3.1 for grantlee and opencv;
]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xen: remove -Werror when building host tools

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

xvisor: new package

This new package provides Xvisor, an open-source type-1 hypervisor,
which aims at providing a monolithic, light-weight, portable, and
flexible virtualization solution.

Signed-off-by: Eric Le Bihan <eric.le.bihan.dev@free.fr>
[Thomas:
 - use MKIMAGE_ARCH instead of BR2_ARCH when calling mkimage
 - use $(MKIMAGE)
 - license is GPLv2+.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

docs/manual: update pkg-kconfig doc about <pkg>_KCONFIG_DOTCONFIG

Content provided by Yann E. Morin.

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

pkg-kconfig: support alternative .config location

Kconfig clones, such as openconf used by xvisor [1], do not look for
.config at the root of the build directory, but in a subdirectory (e.g.
build/openconf).

This patch introduces a new Makefile variable named
$(2)_KCONFIG_DOTCONFIG, which contains the path to .config relative to
the source directory of the package. It defaults to .config and can be
overridden in the package Makefile.

It also creates the full directory hierarchy where the .config file is
stored, in case it does not exist (e.g. in xvisor).

This allows the use of the kconfig-package infrastructure with packages
relying on such clones.

[1] https://github.com/xvisor/xvisor/tree/master/tools/openconf

Signed-off-by: Eric Le Bihan <eric.le.bihan.dev@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[Thomas: use "install -D" instead of "mkdir + cp", as suggested by
Yann.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

infra/pkg-kconfig: Be sure to reconfigure the package on foo-reconfigure

Currently, calling foo-reconfigure for a kconfig-based package will not
re-trigger the configuration (kconfig-wise) step for the package.

This can be problematic when using an override-srcdir suring development
and the content of the veride-srcdir changes (e.g. on a rebase, a merge,
or during a bisect).

This is because the configuration (kconfig-wise) of the package is not
done in the _CONFIGURE_CMDS block, but as a separate action that is not
part of any step [0].

So, be sure to remove the stamp file .stamp_kconfig_fixup_done during
the foo-clean-for-reconfigure rule, so that the configuration is applied
again with the new source tree.

We use another rule, foo-clean-kconfig-for-reconfigure, because we do
not want to override the default foo-clean-for-reconfigure rule, and we
have no way to add conditional commands to it.

[0] The reasons it was not done are not entirely clear in my head, but
IIRC that was not working at the time we tried with Thomas DS.

Reported-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/autoconf-archive: bump to version 2016.03.20

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

openal: Disable build also for examples and tests

Fixes:

  http://autobuild.buildroot.net/results/a11eab534b8642077705d891329c1fd40d394419

Signed-off-by: André Hentschel <nerv@dawncrow.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/{mesa3d, mesa3d-headers}: bump version to 12.0.3

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

php: bump to 7.0.11

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

package/upmpdcli: add patch to fix include path of log

The "log.{h,hxx}" header files are included from libupnpp.

As we are using the pkg-config file of libupnpp to get the header
include path, instead of the hardcoded `{prefix}/include/libupnpp`
these files are not found.

Upstream-status: https://github.com/medoc92/upmpdcli/issues/46

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/upmpdcli: add patch to use pkg-config for libupnpp

Building upmpdcli aborts because of an unsafe header inclusion of
`/usr/include/libupnpp`. Using `{prefix}` to get the header path is
wrong for cross-compilation.

Instead of hard-coding the flags for libupnpp use pkg-config to get the
correct ones.

Note that this patch adds the host-pkgconfig dependency unconditionally and
removes the necessity to pass the libupnpp manually to LIBS for a static
build.

Upstream-status: https://github.com/medoc92/upmpdcli/pull/49

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/upmpdcli: add patch to fix jsoncpp detection

Although jsoncpp is present, the build system of upmpdcli does not detect it.

The default installation path for the header files of jsoncpp is
{prefix}/json, but some systems (Debian) prefers to install the headers
in {prefix}/jsoncpp/json.

The build system only checks for the header files in
{prefix}/jsoncpp/json, so jsoncpp is not found.

This patch from upstream handles both cases now when checking for the jsoncpp
library.

The patch is squashed of two commits from upstream:
1/ aba96db0cfe8a2e5c825f4448c0ad184e21b50e1
2/ ce144f19a8c2e8c1e0ec3fe53699e874e5eb43fe

Fixes:
http://autobuild.buildroot.net/results/5bc6c558fccea0e1e1936894ce372f2b7b8ecac2
http://autobuild.buildroot.net/results/9c2d3892f17c240c44f7255d821d66acbed361f0
http://autobuild.buildroot.net/results/4d22729535ca4e6acbb48a25145611bb49c10149
http://autobuild.buildroot.net/results/fda321774c3eb225f966dd9f83def5329b703a03
http://autobuild.buildroot.net/results/f5eb306b0032e830f7f72ce80e318aa11774a5d1

.. and many more!

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/upmpdcli: fix missing dependencies

Starting with version 1.2 upmpdcli depends on libmicrohttpd and jsoncpp.

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/upmpdcli: requires GCC 4.7

upmpdcli now uses C++11 features, meaning the minimum usable GCC version
is 4.7.

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/libupnpp: add patch to get pkg-config file

Apply a patch from upstream to install a pkg-config file for libupnpp.

Fetch from:
https://github.com/medoc92/libupnpp/commit/9f03bb0e7b47e2843edea6f25ed9eabbfb6412df

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs: nitrogen*: enable VFP and NEON support

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

blktrace: needs MMU support

iowatcher/main.c uses system() which uses fork().

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

blktrace: add dependency on NPTL

blkiomon.c assumes that clock_nanosleep() is always available. But it
is only available with NPTL. Since it unconditionally uses it, we add
dependency on NPTL.

Fixes:
  http://autobuild.buildroot.net/results/c87/c87cd4a7b0bed938506bcf7e88d4cda60e28e6e5/

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

nginx-naxsi: clarify pcre dependency

It is not directly obvious why we select pcre but not add it to the .mk
file, so add a comment to clarify.

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

breakpad: enable for uClibc

There's nothing holding it off for uClibc-based toolchains so enable it.

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

package/urg: fix build caused by missing header inclusion

Update select-h patch to also fix the C++ serial driver.

Fixes:
  http://autobuild.buildroot.net/results/cd7/cd76699188d2ee99b5f2d32b6a797de0d832a192/

Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libcap-ng: annotate licenses with components

libcap-ng.so library is released under LGPLv2.1+ and programs testcap,
filecap, netcap and pscap are released under GPLv2+.

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libcap-ng: bump version to 0.7.8

Drop upstream patch

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

audit: update legal info

Programs are released under GPLv2+. README file says libraries
libaudit.* and libauparse.* are released under LGPL but there are two
files auparse/strsplit.c and lib/strsplit.c which are released under
GPLv2+ which makes both libraries GPL. Until we review situation with
upstream, we simply mention "unclear (libraries)".

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
[Thomas: add "unclear (libraries)" as suggested by Arnout.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

audit: bump version to 2.6.7

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

audit: remove autoreconf

Commit upstream 51925b7dd6d0("package/audit: Bump to version 2.6.5")
removed patch 0002 as part of version bump. But forgot to disable
autoreconf.

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

util-linux: add optional dependency on audit

login and hwclock can be built with audit support. This patch adds
optional dependency on audit package if it is already selected.

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

util-linux: add optional dependency on readline

fdisk and sfdisk tools can be built with readline support. This patch
adds optional dependency on readline if it is already selected.

Signed-off-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libubox: update legal info

Reviewed-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Reviewed-by: Abhimanyu Vishwakarma <Abhimanyu.Vishwakarma@imgtec.com>
Signed-off-by: Rahul Jain <rahul.jain@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libubox: bump version

Reviewed-by: Rahul Bedarkar <rahul.bedarkar@imgtec.com>
Reviewed-by: Abhimanyu Vishwakarma <Abhimanyu.Vishwakarma@imgtec.com>
Signed-off-by: Rahul Jain <Rahul.Jain@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-firmware: bump version

And copy firmware for BCM43430 and BCM43438, hence rpi3 wifi support at
last.

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

linux-headers: bump 3.{2, 14}.x and 4.4.x series

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

linux-headers: bump 4.7.x series to 4.7.4

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain-external: fix potential entire root filesystem removal

This reverts commit a0aa7e0e1750f6ace2879ea8adb1425a41431b79 and reworks
the code to fix a major and potentially catastrophic bug when the
following conditions are met:

 - The user has selected a "known toolchain profile", such as a Linaro
   toolchain, a Sourcery CodeBench toolchain etc. People using "custom
   toolchain profile" are not affected.

 - The user has enabled BR2_TOOLCHAIN_EXTERNAL_PREINSTALLED=y to
   indicate that the toolchain is already locally available (as
   opposed to having Buildroot download and extract the toolchain)

 - The user has left BR2_TOOLCHAIN_EXTERNAL_PATH empty, because his
   toolchain is directly available through the PATH environment
   variable. When BR2_TOOLCHAIN_EXTERNAL_PATH is non-empty, Buildroot
   will do something silly (remove the toolchain contents), but that
   are limited to the toolchain itself.

When such conditions are met, Buildroot will run "rm -rf /*" due to
TOOLCHAIN_EXTERNAL_INSTALL_DIR being empty.

This bug does not exist in 2016.05, and appeared in 2016.08 due to
commit a0aa7e0e1750f6ace2879ea8adb1425a41431b79.

Commit a0aa7e0e1750f6ace2879ea8adb1425a41431b79 removed the assignment
of TOOLCHAIN_EXTERNAL_SOURCE and TOOLCHAIN_EXTERNAL_SITE to empty, as
part of a global cleanup to remove such assignments that supposedly
had become unneeded following a fix of the package infrastructure
(75630eba22b20d6140a5b58a6d1e35598fb3c0d3: core: do not attempt
downloads with no _VERSION set).

However, this causes TOOLCHAIN_EXTERNAL_SOURCE to be non-empty even
for BR2_TOOLCHAIN_EXTERNAL_PREINSTALLED=y configuration, with the
following consequences:

 - Buildroot downloads the toolchain tarball (while we're saying the
   toolchain is already available). Not dramatic, but clearly buggy.

 - Buildroot registers a post-extract hook that moves the toolchain
   from its extract directory (output/build/toolchain-external-.../ to
   its final location in host/opt/ext-toolchain/). Before doing this,
   it removes everything in TOOLCHAIN_EXTERNAL_INSTALL_DIR (which
   should normally be host/opt/ext-toolchain/).

Another mistake that caused the bug is commit
b731dc7bfb9c8ce7be502711f0b44ccab5515f1d ("toolchain-external: make
extraction idempotent"), which introduce the dangerous call "rm -rf
$(var)/*", which can be catastrophic if by mistake $(var) is
empty. Instead, this commit should have just used rm -rf $(var) to
remove the directory instead: it would have failed without consequences
if $(var) is empty, and the directory was anyway already re-created
right after with a mkdir.

To address this problem, we:

 - Revert commit a0aa7e0e1750f6ace2879ea8adb1425a41431b79, so that
   _SOURCE and _SITE are empty in the pre-installed toolchain case.

 - Rework the code to ensure that similar problems will no happen in the
   future, by:

   - Registering the TOOLCHAIN_EXTERNAL_MOVE hook only when
     BR2_TOOLCHAIN_EXTERNAL_DOWNLOAD=y, since moving the toolchain is
     only needed when Buildroot downloaded the toolchain.

   - Introduce a variable TOOLCHAIN_EXTERNAL_DOWNLOAD_INSTALL_DIR which
     is the path in which Buildroot installs external toolchains when it
     is in charge of downloading/extracting them. Then, the
     TOOLCHAIN_EXTERNAL_MOVE hook is changed to use this variable, which
     is guaranteed to be non-empty.

   - Replace the removal of the directory contents $(var)/* by removing
     the directory itself $(var). The directory was anyway already
     re-created if needed afterwards. Thanks to doing this, if $(var)
     ever becomes empty, we will do "rm -rf" which will fail and abort
     the build, and not the catastrophic "rm -rf /*".

Reported-by: Mason <slash.tmp@free.fr>
Cc: Mason <slash.tmp@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux: bump default to version 4.7.4

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libinput: bump to version 1.5.0

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

ccache: bump to version 3.3.1

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

configs: Add support for Solidrun's MX6 Cubox/Hummingboard

Add support for Solidrun's MX6 Cubox/Hummingboard.

For more information about these boards, please check:

https://www.solid-run.com/freescale-imx6-family/hummingboard/
https://www.solid-run.com/freescale-imx6-family/cubox-i/

U-Boot is capable of detecting the version of the board in run-time
and loads the correct dtb.

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-pyusb: bump to version 1.0.0

Remove dependency on Python 2 only.

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-tornado: bump to version 4.4.1

Remove upstream patch.

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

pango: bump to version 1.40.3

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

libcurl: bump version to 7.50.3

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

httping: bump to version 2.5

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

wandboard: Remove unneeded return code

There is no need to handle this return code, so just remove it.

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Revert "package/urg: fix build"

This reverts commit 57e61a91531bf8ea696b5613bd8d846cc645a1d5.

The patch doesn't apply, and makes the urg package fail in all
situations.

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

openal: Use upstream patch to fix aarch64 build

Signed-off-by: André Hentschel <nerv@dawncrow.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/libupnpp: add patch to fix build issue

Bumping libupnpp to version 0.15.0 introduced a build error for some
architectures:

```
./libupnpp/workqueue.h:308:29: error: field 'res' has incomplete type
         std::future<void *> res;
                             ^
./libupnpp/workqueue.h: In member function 'bool
WorkQueue<T>::start(int, void* (*)(void*), void*)':
./libupnpp/workqueue.h:81:25: error: invalid use of incomplete type
'class std::packaged_task<void*(void*)>'
             w.res = task.get_future();
```

The problem is that `std::future` is not available for all
architectures, e.g. it is missing for ARMv5 (soft-float). libstdc++
enables `std::future` only if `ATOMIC_INT_LOCK_FREE > 1`, which
according the the libstdc++ documentation means that operations on
atomic ints are guaranteed to be lock-free.

This check is not true for some target, e.g. the toolchain for the
ARMv5 target defines:

```
$ echo | /usr/bin/arm-linux-g++ -dM -E - | grep ATOMIC_INT_LOCK_FREE
#define __GCC_ATOMIC_INT_LOCK_FREE 1

```

We add a patch from upstream which detects if `std::future` is available
and opts-out the functionality for unsupported targets.

Fixes:
http://autobuild.buildroot.net/results/f8eb38ef79160e1f646dae0be19f27fae82a58bc
http://autobuild.buildroot.net/results/158e0ceeff009146b78a0d2bb278c76479bafc34
http://autobuild.buildroot.net/results/1760cf53c77e16bb2b5bc795bd0ce5eb65f258f2
http://autobuild.buildroot.net/results/5b73e52495aa1bbb2c704492237da784b9d74bc2
http://autobuild.buildroot.net/results/6d922a1bbace024fcf040beb49bbb119036865ef
http://autobuild.buildroot.net/results/bebb8983e7579ee4b4bcf18f037c7797ab86f8db
http://autobuild.buildroot.net/results/0f1a5832ced2a6ec2e18f9f14197446fda6af692
http://autobuild.buildroot.net/results/94a9cc349f23a47246b06619a019bcb917493415
http://autobuild.buildroot.net/results/19fba094098af5c1ce1e1371e5e88393c4a6ac9c
http://autobuild.buildroot.net/results/578935b5c04005d1f096059387d16911f953ec6f
http://autobuild.buildroot.net/results/2ad698f1665b7b886141672181cb76ec3c40a743

and more...

Note, that we enable autoreconf as the patch touches configure.ac.

Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs/wandboard: Remove BR2_LINUX_KERNEL_CUSTOM_GIT

Now that we use a mainline kernel for wandboard, we no longer
need to select BR2_LINUX_KERNEL_CUSTOM_GIT, so remove it.

This fixes the following warning:

make wandboard_defconfig
/configs/wandboard_defconfig:19:warning: override: BR2_LINUX_KERNEL_CUSTOM_VERSION changes choice state
#
# configuration written to /home/fabio/buildroot/.config

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

configs/orangepipc: Use Linux 4.7.3

There's now an sun8i-h3-orangepi-pc dts, so we can use that and drop the
hack in the post-image script.

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