support/testing: move BRTest initialisation to __init__

BRTest's setUp() method contains a few assignments that initialize its
member variables. Since we will want to use these in test case
overrides, move them to the __init__ function.

Also allow the config member to be overridden, rather than always
taking the class member.

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

support/testing: strip /usr/ part from HOST_DIR

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

toolchain-wrapper: remove remaining references to HOST_DIR/usr

Commit 14151d77af20ec50eeba6e30465debf87b35faaa that eliminated
$(HOST_DIR)/usr seriously missed the toolchain-wrapper - only a single
reference was updated, the other three were missed. Commit
015d68c84c9c6ad6f6d41f181d19d813f309088b removed one more. This commit
finally removes the two remaining ones.

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

Makefile: properly create $(HOST_DIR)/usr compatibility symlink

Up to now we created the $(HOST_DIR)/usr compatibility symlink as part
of the creation of $(HOST_DIR) itself. However, when the user specifies
a custom BR2_HOST_DIR, it is possible that the directory already exists
so this rule will never trigger.

Therefore, add an explicit rule for creating $(HOST_DIR)/usr and add
this rule to the dependencies of the dirs target. HOST_DIR itself goes
back to the standard rule for directories. The order-only dependency of
STAGING_DIR isn't needed any more either: HOST_DIR is implicitly
created if needed by mkdir -p, and we don't need to trigger the
HOST_DIR rule any more if the directory already exists.

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

i2c-tools: use upstream tarball

Tarballs of the releases are now again available:

https://www.spinics.net/lists/linux-i2c/msg30349.html

So change back to that instead of getting the source code from git.

While we're at it, add a hash for the license file.

[Peter: Also update Config.in homepage URL as pointed out by Baruch]
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libssh2: fix link with libgcrypt

libssh2 uses the implicit 'yes' argument of the --with-libgcrypt parameter as
a library path prefix, which breaks the build. Pass the library path as
--with-libgcrypt argument to fix that. Drop the unneeded
--with-libgcrypt-prefix.

Fixes:
http://autobuild.buildroot.net/results/ac3/ac3a3f5871a6c20f5407a468809682ce491b9cb3/
http://autobuild.buildroot.net/results/ba3/ba31fa11b2ce2875649414dd9093e039dfc129e8/
http://autobuild.buildroot.net/results/b07/b07bbbf7fe164ba6dc435de6d8b41fadff7a3b08/

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

pcre2: new package

Signed-off-by: Naoki Matsumoto <n-matsumoto@melcoinc.co.jp>
Signed-off-by: Brent Sink <brents_3@hotmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas:
 - reassign authorship to Naoki Matsumoto
 - add entry to DEVELOPERS file
 - fix broken patch
 - add comment about where the PGP signature was coming from]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

uclibc: fix issues with udp.h, missing stdint.h

Fixes:
  http://autobuild.buildroot.net/results/3e4/3e468c7b1be376024422e2f06750571f64f433a2/
  http://autobuild.buildroot.net/results/8c6/8c629a1c5c4a1d78c8bf09a091a332a424a0f757/

Reported-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

opencv: enable NEON optimizations if possible

Signed-off-by: Andre Renaud <andre@ignavus.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

manual: patches are not applied for SITE_METHOD = local

We had several remarks on the mailing list of users that were surprised
that patches were not applied for packages whose SITE_METHOD is local.
So document this.

Note that for OVERRIDE_SRCDIR itself it is already documented:

  When Buildroot finds that for a given package, an
  <pkg>_OVERRIDE_SRCDIR has been defined, it will no longer attempt to
  download, extract and patch the package. Instead, it will directly use
  the source code available in in the specified directory.

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

uboot: apply xtensa overlay

Xtensa core configuration must be added to U-Boot before it can be
built for that xtensa CPU variant. Extract configuration files from the
xtensa overlay as is done for other packages that need to be configured
for a specific xtensa core.

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

linux: apply xtensa overlay

Xtensa core configuration must be added to linux before it can be
built for that xtensa CPU variant. Extract configuration files from the
xtensa overlay as is done for other packages that need to be configured
for a specific xtensa core.

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

configs/qemu-xtensa: use overlay from the github repository

Rather than bundle the overlay with us, directly fetch it from its
upstream location.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

arch/xtensa: accept the overlay to be an URL

It can be interesting to get the overlay from a remote server, rather
than expect it to be present locally.

Since that file can be any URL, we can't know its hash, so we just
exclude it.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
[Thomas: use DL_DIR instead of BR2_DL_DIR.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

arch/xtensa: accept the overlay to be compressed

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

arch/xtensa: allow specifying path to tarball file

currently, specifying a custom Xtrensa core is done with two variables:
  - the core name
  - the directory containing the overlay tarball

However, the core name only serves to construct the tarball name, and is
not used whatsoever to configure any of the toolchain components
(binutils, gcc or gdb), except through the files that are overlayed in
their respective source trees.

This has two main drawbacks:
  - the overlay file must be named after the core,
  - the tarball can not be compressed.

Furthermore, it also makes it extremely complex to implement a download
of that tarball.

So, those two variables can be squeezed into a single variable, that is
the complete path of the overlay tarball.

Update the qemu-xtensa defconfig accordingly.

Note: we do not add a legacy entry for BR2_XTENSA_CORE_NAME, since it
was previously a blind option in the last release, and there's been no
release since we removed BR2_XTENSA_CUSTOM_NAME. So, we just update the
legacy comments for BR2_XTENSA_CUSTOM_NAME, since that's all the user
could have seen in any of our releases so far.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

arch/xtensa: get rid of an intermediate blind kconfig option

It is not needed to have an intermediate blind option, we can just
hide the prompt behind the same dependency as the non-blind symbol.

Update our qemu-xtensa defconfig acordingly (note: it was using
different values for both options, which is not possible; the blind
option was just set to the non-blind one in the .config).

Also remove an unneeded empty default for the BR2_XTENSA_OVERLAY_DIR
string option (strings are empty by default).

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: Add BR2_TARGET_ROOTFS_EXT2_MKFS_OPTIONS option

This option lets the user specify mke2fs options and/or ext2/3/4
features.

Signed-off-by: Sébastien Szymanski <sebastien.szymanski@armadeus.com>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Cc: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: improve wording a bit, fix typos.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fs/ext2: rename BR2_TARGET_ROOTFS_EXT2_BLOCKS to BR2_TARGET_ROOTFS_EXT2_SIZE

This change deprecates the ext2/3/4 rootfs size in blocks symbol in
favor of one that mimic the fs-size argument behavior of mkfs (i.e.
size in a human readable format accepting k, m, g or t suffix or their
upper-case variants).

This change also updates the defconfigs that used to set
BR2_TARGET_ROOTFS_EXT2_BLOCKS symbol.

Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libssh2: bump to version 1.8.0

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libssh: bump to version 0.7.5

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libressl: new package

Libressl is a fork of openssl from OpenSSL in 2014.  Its goal is to
modernize the OpenSSL codebase, improve security, and apply best
practice development processes.

Right now, libressl is API compatible with OpenSSL 1.0.1, but does not
yet include all new APIs from OpenSSL 1.0.2 and later.

Signed-off-by: Adam Duskett <aduskett@codeblue.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

irssi: security bump to version 1.0.4

>From the advisory:
https://irssi.org/security/irssi_sa_2017_07.txt

Two vulnerabilities have been located in Irssi.

(a) When receiving messages with invalid time stamps, Irssi would try
    to dereference a NULL pointer. Found by Brian 'geeknik' Carpenter
    of Geeknik Labs. (CWE-690)

    CVE-2017-10965 [2] was assigned to this bug

(b) While updating the internal nick list, Irssi may incorrectly use
    the GHashTable interface and free the nick while updating it. This
    will then result in use-after-free conditions on each access of
    the hash table. Found by Brian 'geeknik' Carpenter of Geeknik
    Labs. (CWE-416 caused by CWE-227)

    CVE-2017-10966 [3] was assigned to this bug

Impact
------

(a) May result in denial of service (remote crash).

(b) Undefined behaviour.

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

ccache: make default host-ccache cache dir fit for multi-user setups

While building I noticed:

>>> host-ccache 3.3.4 Building
conf.c: In function 'conf_create':
conf.c:314:2: warning: too many arguments for format [-Wformat-extra-args]
  conf->cache_dir = format("/home/peko/.buildroot-ccache", get_home_directory());
  ^

As host-ccache gets installed into $(HOST_DIR) and is part of the SDK,
hardcoding the build user homedir isn't really nice for the relocatable
SDK feature (or simply for a SDK used by multiple users).

As the warning shows, CCache replaces "%s" with the current user home
directory, so rewrite BR_CACHE_DIR to use this feature if it begins with
$HOME.

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

linux: bump default to version 4.12

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux-headers: bump to 4.12 kernel version

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain: add 4.12.x choice for headers

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

util-linux: remove leftover comment

We no longer have automatic derivation of DEPENDENCIES for host
packages, so the comment that we don't want a host-busybox dependency
is no longer valid.

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

execline: restore --shebangdir configure option

Passing the option --shebangdir=/usr/bin to the configuration script adds the
CPP definition EXECLINE_SHEBANGPREFIX to
execline-x.y.z/src/include/execline/config.h. It is used by `s6-rc-compile` from
the s6-rc package to set the path to the execline interpreter in the scripts it
generates.

So, when building the host variant of execline, this path will be used in the
target service scripts generated by the host variant of `s6-rc-compile`. If not
forced to /usr/bin, the location of the execline interpreter on the target, it
will default to $(HOST_DIR)/bin thus leading to non-working scripts on the
target.

So, restore this option for the host variant.

Signed-off-by: Eric Le Bihan <eric.le.bihan.dev@free.fr>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mtd: host variant needs host-util-linux, host-e2fsprogs

Since mtd was converted to the package infrastructure in commit
de4cf4e9135e198d4c3beefc8ad63c03870eb78a ("mtd: convert to gentargets,
add host package"), its host variant depended on host-e2fsprogs. At
the time, only a host variant of the mtd package was available.

When a target variant of mtd was introduced in commit
b50e0fa113bf641a3764ae99b94bb7ba4e1e8f85 ("mtd: add option to build
mkfs.ubifs for target"), it depended on util-linux.

So today, the target variant continues to depend on util-linux, while
the host variant depends on e2fsprogs. What mkfs.ubifs really needs
is libuuid, which is provided by util-linux. It was in fact provided
by the fact that host-e2fsprogs depends on host-util-linux.

But really, host-e2fsprogs is not needed as a dependency, so use
host-util-linux to be consistent with the target variant.

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>

mtd: fix build issues on older hosts

Building the MTD test programs requires the MS_DIRSYNC, which is not
necessarily available on old build machines. But obviously, MTD test
programs are not needed, so we can simply disable them, as they were
prior to the migration to mtd 2.0.

Fixes:

  http://autobuild.buildroot.net/results/21e1ad2a4560b6d3ba6490d20ae064246e66d5c1/

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-wrapper: fix breakage after host/usr removal

The toolchain wrapper, when called through PATH, strips the last three
levels of /proc/self/exe to find HOST_DIR. However, after the host/usr
removal, this should be just two levels.

The toolchain wrapper has different logic for when it is called with a
full path (i.e. $HOST_DIR/usr/bin/arm-linux-gcc) then when it is called
through the PATH (i.e. just arm-linux-gcc). The latter is never used
internally in Buildroot, that's why this wasn't discovered through
testing.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Cc: Mark Jackson <mpfj-list@newflow.co.uk>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

azure-iot-sdk-c: Bump to version 2017-06-30

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

webkitgtk: select libgcrypt

Libgrcrypt is a direct dependency of WebKitGTK+, and as such it
should be selected.

Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: add missing dependency on BR2_PACKAGE_LIBGPG_ERROR_ARCH_SUPPORTS.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

pkg-cmake.mk: Set CMAKE_SYSTEM_PROCESSOR_ARM_VARIANT for ARMv8

This is needed for correctly building some CMake-based packages which
use this variable. For example, this is needed for WebKitGTK+ 2.16.x
to build correctly when an ARMv8 target is configured.

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

cjson: bump to version 1.5.6

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linuxptp: remove trailing backslashes

Thanks to Yann E. Morin for spotting these.

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

atop: add hash for license file

Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

pkg-stats: add exception for nvidia-tegra23.mk

This file is not a package per-se, it includes other .mk files that
are packages.

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

pkg-stats: take into account waf packages

The waf package infrastructure was not known by the pkg-stats script,
so let's add it.

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

pkg-stats: allow to sort by column

With 2000+ packages it's not trivial to identify i.e.:
- all packages that don't have a hash file;
- all packages that have patches;
- all packages that have code style warnings;

User experience can be improved by dynamically sorting the resulting
table.

There is an open-source solution that does that in the client-side and
requires minimal changes to our script: sorttable.js.  The script is
MIT licensed as stated in its website.

Also add a hint to the user that the table can be sorted.

Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-jinja2: don't remove async files for Python 3

The needed functionality is already included into Python 3.6.x,
so these files can be now compiled without errors.

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

python-txaio: bump to version 2.8.0

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

python-setproctitle: change setup type to setuptools

Changing setup type to setuptools avoids installing as zipped .egg

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

monit: bump to version 5.23.0

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

linuxptp: refactor with LINUXPTP_MAKE_{ENV,OPTS} variables

Since there is quite some duplication in the variables to be passed in
the make environment and as make options between the build and install
steps, this commit introduces LINUXPTP_MAKE_ENV and LINUXPTP_MAKE_OPTS
to avoid the duplication.

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

package/linuxptp: set KBUILD_OUTPUT to STAGING_DIR

incdefs.sh try to define some flags with user_flags() and kernel_flags()
functions. The later is looking at the kernel headers installed on the host
when KBUILD_OUTPUT is not set. If no kernel headers are installed on the host,
the grep fail and HAVE_ONESTEP_SYNC is not set on the command line:
see: grep: /usr/include/linux/net_tstamp.h: No such file or directory
So the missing.h define HWTSTAMP_TX_ONESTEP_SYNC which is also present in the
kernel headers installed in STAGING_DIR (toolchain w/ headers >= 3.2).

Indeed KBUILD_OUTPUT is empty because it's reset in the makefile, so move
KBUILD_OUTPUT in the enviroment while calling "make"/

Also set KBUILD_OUTPUT to STAGING_DIR to find net_tstamp.h.

While at it, use the same arguments for BUILD_CMDS and
INSTALL_TARGET_CMDS.

Thanks to Yann for the live review during the Buildroot summer camp.

Fixes:
http://autobuild.buildroot.net/results/364/36470db2c262d2e1fda5144a08cfe221831e093e

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

package/linuxptp: fix coding style issue

Thanks to Yann for the live review during the Buildroot summer camp.

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

dbus: bump to version 1.10.20

Change site to https to avoid a redirection.

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xvisor: fix bogus check on XVISOR_ARCH

The XVISOR_ARCH check added in commit
117fd5dfbc756c4f2b4aef97fc2b568528c66df7 ("xvisor: fix build on
AArch64") broke Buildroot entirely on all architectures except ARM,
AArch64 and x86-64, because the $(error ...) test was not enclosed
inside a condition that made sure the xvisor package was enabled.

This commit fixes that, and allows Buildroot to be usable again on all
architectures.

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

toolchain-external: drop reference to non-existing variable

Commit 32bec8ee2fb0 (toolchain-external: copy ld*.so* for all C libraries)
removed the definition of TOOLCHAIN_EXTERNAL_MUSL_LD_LINK. Remove also the
reference to it.

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

toolchain: replace absolute symlinks by relative symlinks during sysroot copy

In commit 32bec8ee2fb00c6750fa842bbb0eb79b0c081fa2
("toolchain-external: copy ld*.so* for all C libraries") we changed
how the musl dynamic linker symbolic link was being created. Instead
of having specific logic in Buildroot, we switched to simply copying
the ld*.so.* symbolic link from staging to target, as well as the
target of this symbolic link.

However, it turns out that by default, musl creates its dynamic linker
symbolic link with an absolute path as the target of the link:
/lib/libc.so.

Therefore, external Musl toolchains built with Buildroot look like
this:

  lrwxrwxrwx 1 thomas thomas      12 Jul  4 19:46 ld-musl-armhf.so.1 -> /lib/libc.so

The principle of the copy_toolchain_lib_root function, which is used
to copy libraries from staging to target, is to copy symbolic links
and follow their targets. In this case, it means we end up copying
/lib/libc.so (from the host machine) into the target folder. From
there on, there are two cases:

 1. /lib/libc.so exists in your host system. It gets copied to the
    target. But later on, Buildroot also copies /lib/libc.so from
    staging to target, overwriting the bogus libc.so. So everything
    works fine, even though it's admittedly ugly.

 2. /lib/libc.so doesn't exist in your host system. In this case, the
    build fails with no clear error message.

This problem does not happen with Musl toolchains built by
Crosstool-NG, because Crosstool-NG replaces the absolute target of the
dynamic linker symbolic link by a relative path.

However, since we want to support existing Buildroot Musl toolchains
and generally work with the fact that Musl by default installs an
absolute symlink, the following commit improves the
copy_toolchain_sysroot function to replace symbolic links with an
absolute destination to use a relative destination. I.e, in staging,
the ld-musl-armhf.so.1 symbolic link looks like this:

lrwxrwxrwx 1 thomas thomas 14 Jul  5 22:59 output/staging/lib/ld-musl-armhf.so.1 -> ../lib/libc.so

Fixes:

  http://autobuild.buildroot.net/results/ce80264575918a8f71d9eab1091c21df85b65b1a/

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

xvisor: fix build on AArch64

Xvisor was failing to build on AArch64 with:

package/xvisor/xvisor.mk:60: *** No Xvisor defconfig name specified, check your BR2_PACKAGE_XVISOR_DEFCONFIG setting.  Stop.

The first problem is that the Config.in file had a typo: it was using
BR2_AARCH64 instead of BR2_aarch64, and therefore the
BR2_PACKAGE_XVISOR_DEFCONFIG variable had no value.

Once this is fixed, another problem occurs: the ARCH variable needs to
be specified as "arm" for XVisor, for both ARM and AArch64. Therefore,
a XVISOR_ARCH variable is introduced, which is calculated according to
the Buildroot configuration options. Only x86-64, arm and aarch64 are
supported by Xvisor currently, so it remains simple.

Fixes:

  http://autobuild.buildroot.net/results/1719a63ff257f13634a06a14327abfb327984101/

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

cxxtest: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

nanomsg: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gupnp-av: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gupnp: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gssdp: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gflags: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

glog: add hash for license file

Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/kvazaar: disable on PowerPC, PowerPC64 and PowerPC64le

kvazaar is affected by three different build issues on PowerPC and
related architectures:

 - On PowerPC64, the build fails due to the use of a deprecated
   vec_lvsl() function:

strategies/altivec/picture-altivec.c: In function ‘reg_sad_altivec’:
strategies/altivec/picture-altivec.c:43:5: error: vec_lvsl is deprecated for little endian; use assignment for unaligned loads and stores [-Werror=deprecated]
perm1 = vec_lvsl(0, &data1[y * stride1]);
^~~~~

   This bug has been reported upstream at:
   https://github.com/ultravideo/kvazaar/issues/172

 - On PowerPC 8548, the build fails due to mixing AltiVec and E500
   instructions:

strategies/altivec/picture-altivec.c:1:0: error: AltiVec and E500 instructions cannot coexist

   This bug has been reported upstream at:
   https://github.com/ultravideo/kvazaar/issues/173

 - On PowerPC e500mc, because Altivec is not supported on this target:

strategies/altivec/picture-altivec.c:1:0: error: AltiVec not supported in this target

   This bug has been reported upstream at:
   https://github.com/ultravideo/kvazaar/issues/174

Fixes:

  http://autobuild.buildroot.net/results/486/486757f04bc73736b64d18c3298216280be897b1
  http://autobuild.buildroot.net/results/45d/45d22c9f2f5b593bed286319105c82cbc3396835
  http://autobuild.buildroot.net/results/76c/76c76191f84748326d76fe0cc89456a72738a06b

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

linux: bump default version to 4.11.9

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

linux-headers: bump 4.{4,9,11}.x series

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

DEVELOPERS: Add Adrian Perez de Castro for webkitgtk

[Peter: reword commit message]
Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

board: Add nanopi-m1 Support

Add initial support for nanopi-m1 board
with below features
- U-Boot 2017.07-rc1
- Linux 4.11.5
- Default packages from buildroot

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

toolchain/helpers.mk: fix creation of DESTDIR in copy_toolchain_lib_root

In commit b3cc7e65ee, the definition of the DESTDIR variable was moved
down into the loop that follows symlinks in the libraries that are
copied to target. However, the corresponding mkdir was not moved down,
so that no directories are ever created.

In practice, this mkdir is normally redundant since the directories
should already have been created as part of creating STAGING_DIR.
Still, the current situation is clearly wrong, so fix it by moving the
mkdir down to after the assignment to DESTDIR.

While we're at it, also remove a redundant empty line. It's a leftover
from when a lot of variables were declared above.

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

webkitgtk: bump to version 2.16.5

This simply updates to the latest stable release. WebKitGTK+ versions
in the 2.1x series avoid bumping the dependencies in order to allow
distributions to provide updates, therefore no new dependencies are
needed.

Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

CHANGES: update with removal of $(HOST_DIR)/usr

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

check-host-rpath: no longer allow $(HOST_DIR)/usr

Now all packages have been updated to install things in $(HOST_DIR)/lib
instead of $(HOST_DIR)/usr/lib, there should no longer be any reason
to have $(HOST_DIR)/usr/lib in the RPATH, so we don't allow it any more.

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

support/testing: remove references to host/usr

The tools are now installed in host/bin instead of host/usr/bin.

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

libpwquality: Link with zlib when required in static builds

Fixes:

  http://autobuild.buildroot.net/results/cfb3a591391d90762a326d91af3ec502a6161ac5

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

legal-info: add hash for our own license file

This silences the annoying warning that there is no hash file for our
own COPYING file.

Also change the message so that it is more obvious what we're doing.

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

Config.in: add BR2_HOST_GCC_AT_LEAST_7

Signed-off-by: Adrian Perez de Castro <aperez@igalia.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

pciutils: bump to version 3.5.5

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

execline: remove --shebangdir configure option

The host build passes the --shebangdir configure option, the target
build doesn't.  With the removal of $(HOST_DIR)/usr, it is not clear
if this should be /bin or /usr/bin or $(HOST_DIR)/bin. Looking at the
source code, it turns out that this variable is not used at all, and
/usr/bin doesn't appear anywhere in the installed files.

Since it is not clear what this option should be set to, and it
anyway doesn't do anything, remove it entirely.

Cc: Eric Le Bihan <eric.le.bihan.dev@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

policycoreutils: rework host installation

policycoreutils has a pretty peculiar interpretation of DESTDIR and
PREFIX.  PREFIX is not consistently used: some installation paths and
include paths are forced to $(DESTDIR)/usr/... . In other cases,
PREFIX is indeed used. PREFIX defaults to $(DESTDIR)/usr

Try to be a little bit more correct by passing both DESTDIR and PREFIX,
both set to $(HOST_DIR). This is not a complete fix: some things are
still installed in $(HOST_DIR)/usr - but nothing we care about (just
manpages, systemd services, ...). More importantly, however, it still
looks for e.g. D-Bus in $(DESTDIR)/usr/include/dbus-1.0.

Still, it's better than nothing.

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

checkpolicy: rework host installation

checkpolicy has a pretty peculiar interpretation of DESTDIR and PREFIX.
PREFIX simply defaults to $(DESTDIR)/usr, and is used in the rest of
the build system. DESTDIR isn't used any further.

For the host installation, we don't want the usr part, so set PREFIX
instead of DESTDIR.

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

libselinux: rework host installation

libselinux has a pretty peculiar interpretation of DESTDIR and PREFIX.
PREFIX is not consistently used: some installation paths are forced to
$(DESTDIR)/usr/... . In other cases, PREFIX is indeed used. PREFIX
defaults to $(DESTDIR)/usr.

Try to be a little bit more correct by passing both DESTDIR and PREFIX,
both set to $(HOST_DIR). This is not a complete fix: man pages are
still installed in $(HOST_DIR)/usr - but we don't care about that.

Also simplify the symlink creation, like how it's done in libsepol.

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

libsemanage: rework host installation

libsemanage has a pretty peculiar interpretation of DESTDIR and PREFIX.
PREFIX is not consistently used: some installation paths are forced to
$(DESTDIR)/usr/... . In other cases, PREFIX is indeed used. PREFIX
defaults to $(DESTDIR)/usr

Try to be a little bit more correct by passing both DESTDIR and PREFIX,
both set to $(HOST_DIR). This is not a complete fix: man pages are
still installed in $(HOST_DIR)/usr - but we don't care about that.

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

sepolgen: strip /usr/ part from HOST_DIR

sepolgen is a bit weird: DESTDIR acts as a kind of prefix, PYTHONLIBDIR
is relative to it (and a / is added between them).

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Cc: Adam Duskett <aduskett@codeblue.com>
Cc: Clayton Shotwell <clayton.shotwell@rockwellcollins.com>
Cc: Matthew Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

grub2: install in $(HOST_DIR) instead of $(HOST_DIR)/usr

grub2 builds for the target but installs with DESTDIR=$(HOST_DIR). Since
we set prefix to /usr in TARGET_CONF_OPTS, this results in installing
things in $(HOST_DIR)/usr.

To make sure we don't install in $(HOST_DIR)/usr, override --prefix and
--exec-prefix.

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

imx-uxb-loader: use prefix instead of DESTDIR for host installation

prefix defaults to /usr, so setting DESTDIR installs things in
$(HOST_DIR)/usr.

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

dos2unix: use prefix instead of DESTDIR for host installation

prefix defaults to /usr, so setting DESTDIR installs things in
$(HOST_DIR)/usr.

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

raspberrypi-usbboot: remove /usr from DESTDIR-based install commands

We have a patch that adds $(DESTDIR) to the install commands of
raspberrypi-usbboot, but it would still be installed in $(DESTDIR)/usr.
We don't want that, so remove the /usr part in the installation
commands.

Note that upstream has removed the 'install' target entirely, so
there's no point trying to keep the patch in upstreamable shape (i.e.
defaulting DESTDIR to /usr).

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

opkg-utils: use $(HOST_DIR) as PREFIX instead of DESTDIR

... and remove the /usr prefix

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

lzip: use $(HOST_DIR) as prefix instead of DESTDIR

... and remove the /usr prefix

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

libcap: use $(HOST_DIR) as prefix instead of DESTDIR

... and also remove the /usr prefix.

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

genromfs: use $(HOST_DIR) instead of $(HOST_DIR)/usr as prefix

genromfs is special because it uses "PREFIX" in the meaning of DESTDIR
and "prefix" in the meaning of prefix. We were up to know using it
incorrectly for host: PREFIX shouldn't be set and only prefix should
be set.

Add an explanatory comment for this unusual behaviour.

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

libffi: avoid $(HOST_DIR)/usr while moving headers

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

board/csky: remove host/usr reference from readme.txt

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

board/gdb/bfin-bf512: remove host/usr reference from readme.txt

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

lesstif: remove host/usr reference from .mk comment

While we're at it, correct it to $(HOST_DIR).

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

mfgtools: remove host/usr reference from readme.txt

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

lttng-tools: remove host/usr reference from help text

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

grub2: remove host/usr reference from help text

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

manual: remove references to host/usr paths

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

eclipse-register-toolchain: toolchain is no longer installed in $(HOST_DIR)/usr

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

stm32f469-disco: replace $HOST_DIR/usr with $HOST_DIR in flash script

Manual change because the script uses ${OUTPUT_DIR}/host instead of
${HOST_DIR}.

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

stm32f429-disco: replace $HOST_DIR/usr with $HOST_DIR in flash script

Manual change because the script uses ${OUTPUT_DIR}/host instead of
${HOST_DIR}.

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

warpboard: remove $(HOST_DIR)/usr reference from README

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