system: introduce BR2_SYSTEM_ENABLE_NLS

Until now, the option BR2_ENABLE_LOCALE was more-or-less controlling
whether NLS support was enabled in packages. More precisely, if
BR2_ENABLE_LOCALE=y, we were not doing anything (so some packages
could have NLS support enabled, some not). And only when
BR2_ENABLE_LOCALE was disabled we were explicitly passing
--disable-nls to packages.

This doesn't make much sense, and there is no reason to tie NLS
support to locale support. You may want locale support, but not
necessarily NLS support. Therefore, this commit introduces
BR2_SYSTEM_ENABLE_NLS, which allows to enable/disable NLS support
globally. When this option is enabled, we pass --enable-nls to
packages, otherwise we pass --disable-nls.

In addition, when this option is enabled and the C library doesn't
provide a full-blown implementation of gettext, we select the gettext
package, which will provide the full blown implementation.

It is worth mentioning that this commit has a visible impact for users:

 - Prior to this commit, as soon as BR2_ENABLE_LOCALE=y, packages
   *could* provide NLS support. It was up to each package to decide
   whether they wanted to provide NLS support or not (we were not
   passing --enable-nls nor --disable-nls).

 - After this commit, it's BR2_SYSTEM_ENABLE_NLS that controls whether
   NLS is enabled or disabled, and this option is disabled by default.

Bottom line: with the default of BR2_SYSTEM_ENABLE_NLS disabled, some
packages may lose NLS support that they used to provide. But we
believe it's a reasonable default behavior for Buildroot, where
generally NLS support is not necessary.

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

package: rename DISABLE_NLS to NLS_OPTS

We are going to use DISABLE_NLS not only for disabling NLS support,
but also to explicitly enable it. Therefore, this preparatory commit
renames it to NLS_OPTS, which is consistent with other foo_OPTS
variables defined in package/Makefile.in.

While at it, we replace the := assignments by regular = assignments.

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

gettext: force build libintl if needed

This commit introduces a hidden option
BR2_PACKAGE_GETTEXT_PROVIDES_INTL that is used by gettext.mk to force
the build of libintl if the toolchain doesn't provide a full gettext
implementation.

For now, this hidden option is not selected by anything, but a
follow-up commit will introduce BR2_ENABLE_NLS, which will make use of
it.

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

toolchain: introduce BR2_TOOLCHAIN_HAS_FULL_GETTEXT

This new boolean is true if the toolchain provides a built-in
full-featured implementation of gettext (glibc), and false if only a
stub implementation is provided (uclibc, musl).

This will be used in follow-up commits to decide whether libintl needs
to be built by gettext or not.

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

uclibc: enable libintl stubs

In order to avoid having to bring libintl from gettext whenever you
want to build a program that needs it, uClibc-ng now provides a stub
implementation of the gettext functions in uClibc-ng itself. This
brings uClibc-ng in the same situation as musl.

This will fix a lot of build failures related to static linking with
libintl, and generally is a first step to simplify our gettext
handling.

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

ushare: do not use the DISABLE_NLS variable

ushare is not an autoconf/automake based package, so it does not have
the standard behavior of autoconf that every enable/disable option
supports both --enable-foo and --disable-foo.

For example, it supports --disable-nls, but not
--enable-nls. Currently, DISABLE_NLS is either empty or --disable-nls,
but this will be changed in a follow-up commit. As a preparation, this
commit modifies ushare to no longer use DISABLE_NLS, but instead
calculate itself whether --disable-nls should be passed.

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

lvm2: force disable NLS support

We are going to pass --enable-nls to all autotools packages when NLS
support is enabled globally. However, lvm2 NLS support doesn't build,
and anyway lvm2 doesn't provide any useful translation files, so we
force disable NLS support for this package.

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

flex: remove bogus comment

The comment that explains why we were defining HOST_FLEX_DEPENDENCIES
instead of inheriting from FLEX_DEPENDENCIES no longer makes sense,
since we remove such automatic inheritance of host dependencies from
target dependencies a while ago.

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

libglib2: disable compiler warnings

By default, libglib enables some fairly aggressive warnings, treated
as errors. In particular, the -Wformat=2 warning triggers a warning
due to the return value of the ngettext() macro from uClibc libintl
stub not being understood as being potentially a format string.

So, before we enable the stub libintl in uClibc, we disable such
warnings. A bug will be reported to upstream uClibc to get the actual
bug fixed, but disabling compiler warnings treated as errors is anyway
a good thing in the context of Buildroot.

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

package/strace: fix nios2_sp_ptr definition

The patch [1] introduced in strace 1.17 uses the old way of accessing
SP register as defined for kernel headers < 4.0. The kernel headers
for nios2 has been changed in commit [2].

This patch updates strace to use the new kernel/userspace interface.

Fixes:
http://autobuild.buildroot.net/results/b9f/b9fc25b82f3280872fe1593ac252a8529ba83576

[1] https://github.com/strace/strace/commit/6117728aacf431a3e3c05761766c776874cc0861
[2] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?id=92d5dd8cd6e2b211d32d8fbc6cf4b7470765a09f

Signed-off-by: Romain Naour <romain.naour@smile.fr>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: improve a bit the commit log and patch description.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

syslinux: drop patch 0003, not needed after gnu-efi update

Signed-off-by: Benoît Allard <benoit.allard@greenbone.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libusb-compat: fix a build issue with undefined PATH_MAX

libsigrok build failed because of PATH_MAX not being defined in usb.h.
On linux it is defined in linux/limits.h, but usb.h only includes
limit.h. Add a patch fixing the header in libusb-compat.

While we're at it: remove the patch numbering from previous patch.

Fixes: http://autobuild.buildroot.net/results/535/5353c2418c0c3311ef9ecb1f1ddc3ce769369b96/
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-fusion: fix build issue with Linux >= 4.7

The size parameter from sock_recvmsg() was removed from Linux kernel
API since 4.7. This commit adjusts the existing
0004-Port-one-one_udp.c-to-Linux-4.1.patch to fix the build with Linux
>= 4.7.

Signed-off-by: Matthew Shyu <matthew.shyu@amlogic.com>
[Thomas: improved commit title/log.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/systemd: needs timezone info

systemd does not like being booted without any timezone info (especially
on a R/O filesystem), so we forcibly enable that.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
[Thomas: fix alphabetic ordering.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

boot/syslinux: disable syslinux legacy-BIOS for broken toolchains

Since [1] syslinux is built with the target toolchain in order to
properly build with gnu-efi package. But toolchains built with
binutils 2.26 break the syslinux legacy-BIOS build as reported at [2],
due to binutils bug #19615.

Thanks to Benoît Allard for the investigation and the link to the
binutils bug [3].

[1] 6e432d5ecb46959f3c27dbbca48e41512bb40a70
[2] http://lists.busybox.net/pipermail/buildroot/2017-July/196253.html
[3] https://sourceware.org/bugzilla/show_bug.cgi?id=19615

Signed-off-by: Romain Naour <romain.naour@smile.fr>
Cc: Benoît Allard <benoit.allard@greenbone.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

toolchain: CodeSourcery AMD64 affected by PR19615

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

package/qt5base: provide "qt.conf" to make "qmake" relocatable

The file "qt.conf" can be used to override the hard-coded paths that are
compiled into the Qt library. We need it to make "qmake" relocatable.

CC: Julien Corjon <corjon.j@ecagroup.com>
CC: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
[Thomas:
 - use | instead of \ as sed separator, suggested by Arnout.
 - fix indentation, suggested by Arnout.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/scripts: add relocate-sdk.sh script for SDK relocation

It will install the script "relocate-sdk.sh" in the HOST_DIR
allowing to adjust the path to the SDK directory in all text
files after it has been moved to a new location.

Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
[Thomas:
 - Fix shebang to be /bin/sh instead of /bin/bash, suggested by Arnout
 - Use | instead of \ as a separator for sed expressions, suggested by
   Arnout, discussed with Wolfgang and others
 - Remove ./ at the beginning of LOCFILE, suggested by Arnout
 - Fix comment about the path check being made before doing the
   replacement, suggested by Arnout
 - Fix indentation, suggested by Arnout.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

spidev-test: s/BR2_DL_DIR/DL_DIR/

BR2_DL_DIR is not to be used by the package's Makefile.

Signed-off-by: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Acked-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

kodi: s/BR2_DL_DIR/DL_DIR/

BR2_DL_DIR is not to be used by the package's Makefile.

Signed-off-by: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Acked-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

manual: add documentation about DL_DIR

And at the same time don't encourage use of BR2_DL_DIR

Signed-off-by: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Acked-by: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gst1-rtsp-server: bump version to 1.12.1

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

gst1-validate: bump version to 1.12.1

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

gst1-plugins-bad: bump version to 1.12.1

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

gst1-plugins-ugly: bump version to 1.12.1

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

gst1-plugins-good: bump version to 1.12.1

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

gst1-plugins-base: bump version to 1.12.1

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

gst1-libav: bump version to 1.12.1

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

gst-omx: bump version to 1.12.1

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

gst1-vaapi: bump version to 1.12.1

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

gstreamer1: bump to version 1.12.1

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

libmad: add security patch from debian

Fixes:

CVE-2017-8372 - The mad_layer_III function in layer3.c in Underbit MAD
libmad 0.15.1b, if NDEBUG is omitted, allows remote attackers to cause a
denial of service (assertion failure and application exit) via a crafted
audio file.

CVE-2017-8373 - The mad_layer_III function in layer3.c in Underbit MAD
libmad 0.15.1b allows remote attackers to cause a denial of service
(heap-based buffer overflow and application crash) or possibly have
unspecified other impact via a crafted audio file.

CVE-2017-8374 - The mad_bit_skip function in bit.c in Underbit MAD libmad
0.15.1b allows remote attackers to cause a denial of service (heap-based
buffer over-read and application crash) via a crafted audio file.

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

icu: propagate host gcc dependency to reverse dependencies

Fixes:
http://autobuild.buildroot.net/results/821/821eae7add252e3711ed069a2d292a04427fa43b/

Commit 87f43c3403 (package/icu: needs host gcc >= 4.8) added a new toolchain
dependency to icu, but didn't propagate it to the reverse dependencies of
icu.

Some of these have been fixed in later commits, but not all.

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

gst1-rtsp-server: bump version to 1.12.0

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

gst1-validate: bump version to 1.12.0

Remove non-existant configure options:

  --disable-sphinx-doc:
  Even if it looks like there is some code for that in configure.ac, it
  seems that is not taken into account and that configure option is not
  even recognized.

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

gst1-plugins-bad: move dtls plugin

The dtls plugin needs openssl, so move it down to the "plugins with external
dependencies" section to match the rest of the plugins.

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

gst1-plugins-bad: add new Timed Text Markup Language (TTML) plugin

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

gst1-plugins-bad: add new dependency-less audiobuffersplit / audiomixmatrix plugins

These plugins are new in 1.12.x

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

gst1-plugins-bad: explicitly disable new msdk plugin

As we do not have the needed dependencies packaged.

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

gst1-plugins-bad: add Image Quality Assessment (IQA) plugin

This plugin replaces the recently removed videomeasure plugin.

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

gst1-plugins-bad: bump version to 1.12.0

Remove non-existant configure options:

  --disable-mimic:
    https://github.com/GStreamer/gst-plugins-bad/commit/ad661999ad2c79a5d7721f0eb1e7f1f83a304224

  --enable-dataurisrc:
    https://github.com/GStreamer/gst-plugins-bad/commit/de4bb6e1d4793637fd72d7316d06ebba4148583e

  --enable-videomeasure:
    https://github.com/GStreamer/gst-plugins-bad/commit/ca3020cf46d49d8e3fd54ec323b3a17cd389d780

  --enable-apexsink, --disable-nas, --disable-linsys, --enable-sdl,
  --disable-sdltest, --disable-libvisual, --disable-timidity,
  --disable-sndio, --disable-wininet, --disable-xvid,
  --enable-cdxaparse, --enable-dccp, --enable-hdvparse, --enable-mve,
  --enable-nuvdemux, --enable-patchdetect, --enable-sdi, --enable-tta,
  --disable-pvr:
    https://github.com/GStreamer/gst-plugins-bad/commit/9b5de053995488d5ddc78c1bf4df651101271d70

[Peter: split out webrtc rename to a separate patch, tweak dataurisrc help]
Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gst1-plugins-ugly: bump version to 1.12.0

Remove non-existant configure options:

  --enable-mad:
    https://github.com/GStreamer/gst-plugins-ugly/commit/2f767fb24ac8ee001bde7e1452b46a4f8fe3a719

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

gst1-plugins-good: bump version to 1.12.0

Remove non-existant configure options:

  --disable-esd, --disable-esdtest:
  https://github.com/GStreamer/gst-plugins-good/commit/3d6955d68624c089d7066bfdbf38ef69c1dc07b6

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

gst1-plugins-base: bump version to 1.12.0

Remove non-existant or unnecessary configure options:

  --disable-oggtest, --disable-vorbistest:
    It looks like these options will be taken into account by ogg.m4 and
    vorbis.m4 only when ogg and vobis are enabled. However, according to
    the code on those m4 files these two options are disabled by default.

  --disable-gio_unix_2_0:
    I don't know where this option comes from.

  --disable-freetypetest:
    https://github.com/GStreamer/gst-plugins-base/commit/183610c035dd6955c9b3540b940aec50474af031

[Peter: add rawparse plugin]
Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gst1-libav: bump version to 1.12.0

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

gst-omx: bump version to 1.12.0

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

gst1-vaapi: bump version to 1.12.0

0001 patch already included in this release:
  https://github.com/GStreamer/gstreamer-vaapi/commit/159e3c3f08ef1c19da4ede8a437337dd8cf7f0b7
0002 patch already included in this release:
  https://github.com/GStreamer/gstreamer-vaapi/commit/152217064f1192bb836620ad7eaadbc031624f4c

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

gstreamer1: bump version to 1.12.0

0001 patch already included in this release:
  https://github.com/GStreamer/gstreamer/commit/c893882fd93238ce722096ecf4db87b6a9f87c5d

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

gst1-plugins-bad: fix webrtc option

When the WebRTC plugin option was introduced in commit
ee267886bc090335dd94c97dca1c294addccc090 ("gst1-plugins-bad: enable
webrtc plugin"), it was incorrect added as "webrtc", while the actual
name of the plugin and corresponding configure option is "webrtcdsp".

This commit therefore fixes the .mk file to use the correct name. And
also, since we want to keep Buildroot option consistent with the name
of the GStreamer plugins, it renames the Config.in option as well, and
introduces the necessary Config.in.legacy handling.

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

pixz: new package

Add a package for pixz (pronounced pixie), a parallel, indexing version of
xz.

See https://github.com/vasi/pixz

Signed-off-by: Vincent Stehlé <vincent.stehle@laposte.net>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

system: require a timezone to be set

Even though no timezone implicitly means UTC, systemd is not all that
happy when it does not have a timezone set. This is all fine on a RW
filesystem because systemd will create a symlink on its own (to
Etc/UTC), but not so much on a RO filesystem, causing all kind of
issues at boot time (up to the point that the system is unusable).

We fix that by requiring that the timezone is actually set. The check is
done by verifying that the timezone file is an actual file; if not set,
the test would find a directory and would thus fail.

Update the help entry accordingly.

Also fix indentation in tzdata.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Richard Braun <rbraun@sceen.net>
Cc: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Cc: Martin Bark <martin@barkynet.com>
Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/tz: fix variable name

The tz package mixes its own variable with the one from the tzdata
package...

Fix the variable name in tz.

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

package/mke2img: add option to specify block size

ext filesystems can use a block size of 1024, 2048, or 4096 bytes, the
former being interesting to store small files, while the latter being
more intersting to store bigger files.

So far, we were using the default, which was to use a 1024-byte block.
Continue doing so (for now...).

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: simplify code

No option is conditional anymore, so just collate them into a single
assignment.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: allow reserving zero block for root

The previous default, zero, just meant "use the default value of the
filesystem generator", which happened to be 5% (the traditional value
for all ext-creating tools we've ever seen).

So, change the new default accordingly to 5%.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: always pass the label option

... since passing an empty string is equivalent to not setting a label.

And fix the syntax highlighting in some editors...

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: always pass the number of inodes

... since zero also means auto for the mke2img script, now.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/mke2img: specifying zero inodes means auto

Currently, leaving the number of inodes to be autocalculated requires
the user to not specify the -i option at all.

Also accept zero as meaning auto; this will help simplify the ext2.mk
code a little bit.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libsepol: rework host installation

libsepol has a pretty peculiar interpretation of DESTDIR and PREFIX.
PREFIX is not consistently used: some installation paths are forced to
$(DESTDIR)/usr/... . Also, the shared lib symlink is forced to ../../lib
(i.e. assumes it's installed under /usr). For these reasons, the host
build and install commands are a bit unusual: they use DESTDIR instead
of PREFIX, DESTDIR points to $(HOST_DIR) without /usr, and some
additional fixup is needed to correct the installation directory of
libsepol.so.1 and the symlink to it.

This can be simplified quite a lot by passing both DESTDIR and PREFIX,
and including the /usr part in both. The symlink still has to be fixed,
but that's much simpler.

Note that a side effect is that the man pages are now installed under
$(HOST_DIR)/usr/usr/man. While not very nice, we don't really care
about man pages. In addition, this will resolve itself when we later
move $(HOST_DIR)/usr to $(HOST_DIR).

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

pkg-rebar: remove redundant / in front of $($(PKG)_ERLANG_LIBDIR)

Sometimes $(1)/$($(PKG)_ERLANG_LIBDIR) is used, sometimes
$(1)$($(PKG)_ERLANG_LIBDIR). Make this consistent, with /. Therefore,
remove the leading / from $($(PKG)_ERLANG_LIBDIR).

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

package/libunwind: fix build failure due to asm()

The gcc documentation [1] suggest to use __asm__ instead of asm.

Fixes:
http://autobuild.buildroot.net/results/3ef/3efe156b6494e4392b6c31de447ee2c72acc1a53

[1] https://gcc.gnu.org/onlinedocs/gcc/Alternate-Keywords.html#Alternate-Keywords

Signed-off-by: Romain Naour <romain.naour@smile.fr>
Cc: Bernd Kuhls <bernd.kuhls@t-online.de>
Tested-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

busybox: applets as individual binaries

The individual binaries option of busybox allows for the applets
that would usually be symlinks to be built as individual applications
that link against a shared library.

This feature is needed for SELinux to allow the applications to run
under the correct SELinux context.

The patch being added allows the individual applications to be
installed and will be upstreamed to the busybox developers.

The initial work for this change was done by Thomas Petazzoni
<thomas.petazzoni@free-electrons.com>.

Signed-off-by: Clayton Shotwell <clayton.shotwell@rockwellcollins.com>
Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Niranjan Reddy <niranjan.reddy@rockwellcollins.com>
Signed-off-by: Bryce Ferguson <bryce.ferguson@rockwellcollins.com>
[Thomas:
 - add help text in Config.in option.
 - rename BUSYBOX_CONFIGURE_INDIVIDUAL_BINARIES to
   BUSYBOX_SET_INDIVIDUAL_BINARIES to be consistent with other
   variables.
 - call BUSYBOX_INSTALL_INDIVIDUAL_BINARIES in
   BUSYBOX_INSTALL_TARGET_CMDS, not in BUSYBOX_INSTALL_INIT_SYSV.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
[Arnout:
 - revert to the "optional file" approach in makedevs;
 - reword Config.in comment text to match our usual pattern;
 - abbreviate comment about how suid applets are found a little;
 - all applets are optional (each one may have been unselected from the
   busybox config).]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: remove /usr/share/udhcpc/default.script from BUSYBOX_PERMISSIONS.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

makedevs: support optional files

Add the 'F' file type to makedevs, that allows a file to be optional.
With this option, the line is just silently skipped if the file doesn't
exist (or is not a regular file).

This is useful for _PERMISSIONS where the file can be configured out
by package-specific configuration that is not directly handled by
Buildroot, like busybox.

Cc: Clayton Shotwell <clayton.shotwell@rockwellcollins.com>
Cc: Matthew Weber <matthew.weber@rockwellcollins.com>
Cc: Niranjan Reddy <niranjan.reddy@rockwellcollins.com>
Cc: Bryce Ferguson <bryce.ferguson@rockwellcollins.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/libva-utils: new package

The utils provided by this package were formerly packaged with libva.

Libva contained a patch not to compile mpeg2vldemo to reduce
dependencies by avoiding C++, this patch is not moved to this package.

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/lua-stdlib: new package

Signed-off-by: Marcin Niestroj <m.niestroj@grinn-global.com>
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

pngquant: doesn't support building with ccache

pngquant's homegrown configure script doesn't understand CC with a
space, so we can't use ccache with it. Only the host variant is
affected: for the target, we call ccache from the toolchain-wrapper so
it's transparent to the configure script.

This wasn't seen in the autobuilders since they never enable
BR2_CCACHE.

To test, use any configuration and do
make BR2_CCACHE=y host-pngquant

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

azure-iot-sdk-c: Package depends on NPTL

Fixes:
http://autobuild.buildroot.org/results/1ad/1ad1c02eb866dd9a1b586308b11b4242f4321355/
http://autobuild.buildroot.org/results/fd0/fd0a1f71fec062a530bcffe95e501095657c0d50/

Signed-off-by: André Hentschel <nerv@dawncrow.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

easy-rsa: new package

[Peter: Mention that openssl is a runtime dependency,
Add license info as noted by Romain, add gpl-2.0.txt,
Add comment explaining the missing build step,
Use install -t for x509-types files]
Signed-off-by: Julien Viard de Galbert <julien@vdg.name>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/devmem2: fix legal-info

Fixes:
http://autobuild.buildroot.net/results/1c6/1c69c77d7b378f6f76790fb5ca0078d0db0c40f8

devmem2 is a single C file that is downloaded as-is, and we patch it.
We also use that file as the license file.

Thus, the file when it is downloaded has a specific hash, but when
saved by legal-info, it has a different hash.

However, we can't store two different hashes for the same _filename_
with different content.

So, we extract the license file as a pre-patch hook, and store it in a
separate file, for which we can now add a hash.

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

package/cache-calibrator: fix legal-info

Fixes:
http://autobuild.buildroot.net/results/e1b/e1b3dee917f5ab3961f3dea006720431444d0ef5/

cache-calibrator is a single C file that is downloaded as-is, and we
patch it. We also use that file as the license file.

Thus, the file when it is downloaded has a specific hash, but when
saved by legal-info, it has a different hash.

However, we can't store two different hashes for the same _filename_
with different content.

So, we extract the license file as a pre-patch hook, and store it in a
separate file, for which we can now add a hash.

[Peter: add autobuilder reference]
Reported-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Stephan Hoffmann <sho@relinux.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mpg123: security bump to version 1.25.1

>From the release notes:

- Avoid memset(NULL, 0, 0) to calm down the paranoid.

- Fix bug 252, invalid read of size 1 in ID3v2 parser due to forgotten
  offset from the frame flag bytes (unnoticed in practice for a long time).
  Fuzzers are in the house again.  This one got CVE-2017-10683.

  https://sourceforge.net/p/mpg123/bugs/252/

- Avoid a mostly harmless conditional jump depending on uninitialised
  fr->lay in compute_bpf() (mpg123_position()) when track is not ready yet.

- Fix undefined shifts on signed long mask in layer3.c (worked in practice,
  never right in theory).  Code might be a bit faster now, even.  Thanks to
  Agostino Sarubbo for reporting.

dlopen() is now directly used to load output modules (and the
--with-modules-suffix option has been removed), so adjust the modules logic
to match.

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

vlc: add upstream security patches fixing CVE-2017-10699

avcodec 2.2.x, as used in VideoLAN VLC media player 2.2.7-x before
2017-06-29, allows out-of-bounds heap memory write due to calling memcpy()
with a wrong size, leading to a denial of service (application crash) or
possibly code execution.

https://trac.videolan.org/vlc/ticket/18467

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

libpwquality: select cracklib

libpwquality adds cracklib to its _DEPENDENCIES, but forgot to select it in
Config.in, leading to build failures if cracklib isn't explicitly enabled:

Makefile:536: *** cracklib is in the dependency chain of libpwquality that
has added it to its _DEPENDENCIES variable without selecting it or depending
on it from Config.in.  Stop.

No autobuilder references as this error happens before build-time.log is
written, causing the autobuilder to ignore the build result.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

docs/manual: document hashes for license files

[Peter: use sha256 in example]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core/pkg-utils: check hashes of license files

This will help catch a change of license even if the filename does
not change.

For now, a missing hash for the license files is not a fatal error, to
let people catch up and add them. When we switch to make it mandatory,
we can simplify the code by just removing the case statement.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core/pkg-util: pass package directory and name when saving license files

This will be useful when checking the hashes of the license files.

[Peter: use '.' as buildroot directory so /buildroot.hash isn't checked]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core/pkg-generic: call MESSAGE when saving package legal-info

Currently, the per-package legal-info is mostly silent, but we're soon
to add a check for the hashes of the license files.

In that case, and when there is a hash mis-match, we want a user to know
what package had a changed license file.

So, we add a call to MESSAGE to display the package we're currently
saving the legal-info of, like so:

    >>> busybox 1.26.2 Collecting legal info

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

support/check-package: don't check filenames of hashes

Currently, we check that the filenames in hash lists do not contain
a slash '/' character, because all we are checking so far are the
downloaded archives, and we explicitly need the filename to not contain
a directory component at all.

However, we're soon to also check the hashes of the license files in
packages sources, and those license files may be at any arbitrary
directory-depth in the packages source tree.

[Peter: Remove reference to files with same basename]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Luca Ceresoli <luca@lucaceresoli.net>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Cc: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Acked-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

fakedate: simplify logic

Using -ef to check for the same file is nicer than relying on a magic
symlink-to-fakedate.

Notice that -ef isn't stricly posix (but supported by bash/dash/zsh), so
I've changed the shebang to /bin/bash.

While we are at it, restructure the logic to do a single exec at the end
instead of handling the epoch/!epoch cases differently for simplicity.

With that out of the way we can directly install it as $HOST/usr/bin/date
instead of the fakedate / date symlink.

[Peter: drop IFS=: change]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Makefile: generate wrapper makefile when running make without a .config

The recent change to error out instead of running menuconfig when no .config
is available broke an existing use case:

make O=output-foo; cd output-foo; br-init-conf (or similar to get a sensible .config); make

As there is no wrapper makefile in output-foo.

Fix it by ensuring the wrapper gets created if needed.

Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

tslib: use SPDX identifiers for licenses in use

[Peter: License is GPL-2.0+ / LGPL-2.1+]
Signed-off-by: Martin Kepplinger <martink@posteo.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

setlocalversion: fix detection of hg revision for untagged versions

By default, cut prints the entire line if the specified delimiter is not
present at all:

    $ printf "foo bar" | cut -d' ' -f2
    bar
    $ printf "foobar" | cut -d' ' -f2
    foobar

In setlocalversion, cut is presented with the output of 'hg id' which has
the format:

    "<revision> <tags-if-any>"

If the current revision is not tagged, the output of 'hg id' does not
contain the delimiter (space), cut prints the entire string, and
setlocalversion thinks the version is the tag.
As setlocalversion does not print anything for tagged versions, there is no
output overall, and no correct indication of the mercurial revision.

Fix by passing the extra cut option '--only-delimited', which suppresses
output if no delimiter is found.

This problem likely went unnoticed for so long, because the tag 'tip' (i.e.
most recent revision of the branch) is treated specially: in this case the
mercurial revision _is_ printed, i.e. the situation is treated as
'untagged'.
The problem is only seen when you are _not_ at the most recent revision in
your branch.

Signed-off-by: Thomas De Schampheleire <thomas.de_schampheleire@nokia.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

qt5virtualkeyboard: new package

This patch adds the Qt virtualkeyboard package.

Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com>
Reviewed-by: Peter Seiderer <ps.report@gmx.net>
[Thomas:
 - use SDPX license codes
 - add entry to DEVELOPERS file
 - drop BR2_PACKAGE_QT5BASE_LICENSE_APPROVED logic, since we removed this option
 - select BR2_PACKAGE_QT5DECLARATIVE and
   BR2_PACKAGE_QT5DECLARATIVE_QUICK instead of using a "depends on"
   and propagate the appropriate dependencies]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

lirc-tools: bump to version 0.9.4d

Drop 0001-tools-make_rel_symlink.py-can-also-use-python2.patch. Build requires
host-python3 since the previous version bump.

Drop upstream 0002-lircd-Remove-use-of-functions-killed-in-kernel-4.8.0.patch.

Add a patch fixing header guard macro collision with musl libc.

Set HAVE_WORKING_POLL to skip poll(2) run test that is not compatible with
cross compilation.

Add optional dependency on libftdi1.

Cc: Rhys Williams <github@wilberforce.co.nz>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

wavpack: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

trinity: don't download patches from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

tinycbor: on't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

systemd: don't download patches from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

softether: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

sngrep: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

For this package, the patch is not replaced 1:1 with what Github was
providing. Indeed, the URL
https://github.com/irontec/sngrep/pull/191/commits/4740f3341a99eaec105dee202a6fa7828212cdf1.patch
no longer exists. This pull request has been merged, so we instead use
the real upstream commits. However, the upstream developer apparently
messed up, and applied the change in two separate commits, which is
why one patch fetched from Github is replaced by this commit by two
patches in package/sngrep/.

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

scrub: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

rpm: don't download patches from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

qt: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

poco: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

openzwave: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

openssh: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

openal: don't download patches from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

numactl: don't download patches from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

mediastreamer: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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

lxc: don't download patch from Github

Patches downloaded from Github are not stable, so bring them in the
tree.

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