fs: run filesystem hooks under fakeroot

Currently, the pre-gen hooks of the various filesystems are run before
we enter the fakeroot.

However, this precludes those hooks from doing actions that require
root, like creating a pseudo-device or the likes.

So, move those pre-gen hooks under fakeroot.

This has currently no side-effect, as they are still called before
everything else in the fakeroot script, even the system-wide chown call.

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

package/dovecot: add optional support for libsodium

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

linux-headers: bump 4.{4, 9, 14, 15}.x series

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

linux: bump default to version 4.15.15

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

package/docker-engine: convert to golang infrastructure

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
[Thomas:
 - adapt to the changes in the golang-package infrastructure
 - keep the logic to support BR2_PACKAGE_DOCKER_ENGINE_STATIC_CLIENT]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/apache: add optional support for lua

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

package/docker-proxy: convert to golang infrastructure

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[Arnout: remove automatically added host-go dependency]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/docker-containerd: convert to golang infrastructure

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/runc: convert to golang infrastructure

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/flannel: convert to golang infrastructure

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

docs/manual: add documentation for the golang infrastructure

This patch adds the documentation for the golang infrastructure.

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/pkg-golang: new package infrastructure

This patch adds a new infrastructure for golang based packages.

Signed-off-by: Angelo Compagnucci <angelo@amarulasolutions.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
[Arnout:
 - Rewrap comments to 80 columns.
 - Create a global definition of GO_TARGET_ENV.
 - <PKG>_GO_ENV is appended to the default env instead of replacing it.
 - Add a note to inner-golang-package that only target is supported.
]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

docker-containerd: remove symlink to $(RUNC_SRCDIR)

The DOCKER_CONTAINERD_CONFIGURE_CMDS creates a symlink to runc's
source directory, but it does not build depend on runc, which means
that the runc package may not have been extracted/built before
docker-containerd.

Therefore, when doing a build with "make docker-containerd", this
symbolic link is broken, but it doesn't prevent docker-containerd from
building.

Therefore, this symlink is not necessary and can be dropped.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

Makefile: create symlink to non-default HOST_DIR

If BR2_HOST_DIR is not the default, it can be difficult to find the
host directory (i.e., HOST_DIR always has to be passed explicitly in
addition to the output directory). For example, the Eclipse plugin
assumes that HOST_DIR=BASE_DIR/host.

Create a symlink from $(BASE_DIR)/host to $(HOST_DIR) if it is not the
default. Also remove it in the clean target.

When BR2_HOST_DIR is the default, HOST_DIR_SYMLINK will be empty so
there will be no additional dependency to dirs and nothing to remove
in clean.

Fixes https://bugs.busybox.net/show_bug.cgi?id=10151

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

Makefile: move mkdir rule to after HOST_DIR is defined

HOST_DIR is defined twice: once to its default value before .config is
included, and once more to BR2_HOST_DIR after .config is included.
However, the rule that defines the mkdir for HOST_DIR comes between
these two, so it will always use the default definition. Therefore,
if a non-default BR2_HOST_DIR is used, there will be no rule to create
that directory, while the dirs target depends on it.

This happens to work at the moment, because in the dirs target,
$(STAGING_DIR) comes before $(HOST_DIR), so $(HOST_DIR) will be created
implicitly. However, this will fail in top-level parallel builds where
both will be created in parallel.

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

python-libusb1: new package

License info from PyPI is wrong: it is LGPL-2.1+, not GPL-2.0. The
package includes a COPYING file with GPL-2.0, but since it is really
LGPL-2.1+ only the COPYING.LESSER file is relevant.

Signed-off-by: Jaap Crezee <jaap@jcz.nl>
[Arnout: regenerate with scanpypi, fix legal info]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

utils/genrandconfig: use --no-check-certificate in wget by default

A number of autobuilder failures are due to the fact that autobuilder
instances use old distributions, with old SSL certificates, and
therefore wget aborts with an error "The certificate of `xyz.org' is
not trusted.".

In order to avoid such failures that are not very interesting in the
context of the autobuilders, we pass --no-check-certificate to
wget. The integrity of the downloaded files is anyway verified by the
hashes, and this is only meant to be used in the context of
testing/CI, not in production.

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

nodejs: security bump to version 8.11.1

Fixes the following security issues:

- Fix for inspector DNS rebinding vulnerability (CVE-2018-7160): A malicious
  website could use a DNS rebinding attack to trick a web browser to bypass
  same-origin-policy checks and allow HTTP connections to localhost or to
  hosts on the local network, potentially to an open inspector port as a
  debugger, therefore gaining full code execution access.  The inspector now
  only allows connections that have a browser Host value of localhost or
  localhost6.

- Fix for 'path' module regular expression denial of service
  (CVE-2018-7158): A regular expression used for parsing POSIX paths could
  be used to cause a denial of service if an attacker were able to have a
  specially crafted path string passed through one of the impacted 'path'
  module functions.

- Reject spaces in HTTP Content-Length header values (CVE-2018-7159): The
  Node.js HTTP parser allowed for spaces inside Content-Length header
  values.  Such values now lead to rejected connections in the same way as
  non-numeric values.

While we are at it, also add a hash for the license file.

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

package/zsh: remove busybox show others dependency

Back in 2013, zsh was added to Buildroot [1] with the dependency on
BR2_PACKAGE_BUSYBOX_SHOW_OTHERS. This dependency was keept with the
refactoring [2].
But Busybox doesn't provide a zsh shell in the firt place.
This dependency can be removed.

[1] 3d19b7ef598f3534b46cf0746a81bc437579127c
[2] 1dbd7b991025a8bf39ab5d25c2fd97f47f084bac

Signed-off-by: Romain Naour <romain.naour@smile.fr>
Cc: Phil Eichinger <phil.eichinger@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

squid: add optional dependency on libkrb5

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

libkrb5: new package

Signed-off-by: André Hentschel <nerv@dawncrow.de>
[Thomas:
 - remove LIBKRB5_VERSION_MAJOR
 - minor tweaks to commit title and Config.in help text]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

board: add support for ARC HS Development Kit (HSDK)

Synopsys DesignWare HSDK (which stands for ARC HS
Development Kit) is the latest and greatest development
platform that sports quad-core ARC HS38 in real silicon.

Most noticeable features of the board are:
 * Quad-core ARC HS38 CPU running at 1GHz
 * 4Gb of DDR
 * Built-in Vivante GPU (well supported via open source
   Etnaviv drivers)
 * Built-in Wi-Fi/Bluetooth module (RedPine RS-9113)

And as usual we have:
 * [micro] SD-card slot
 * 2 USB 2.0 ports
 * 1Gbit Ethernet port
 * Built-in Digilent JTAG probe
 * Serial port accessible via micro-USB port

Writing sdcard.img on SDcard creates two partitions:
 * FAT32 with uImage and uboot.env
 * EXT4 with root filesystem

We modify kernel config because in default hsdk kernel config
CONFIG_INITRAMFS_SOURCE parameter is set and when we build
rootfs separately (BR2_TARGET_ROOTFS_INITRAMFS is not set)
error appears. Also we set up CONFIG_ARC_UBOOT_SUPPORT which
enables usage of uboot variables in the boot process.

[Peter: Fix comments and rename defconfig to snps_archs38_hsdk_defconfig,
Add defconfig to DEVELOPERS and fixup board/synopsys entry,
Drop postimage script, rename env file in genimage.cfg and drop size
setting for rootfs partition,
Add "" for CONFIG_INITRAMFS_SOURCE in linux fragment]
Signed-off-by: Evgeniy Didin <didin@synopsys.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: arc-buildroot@synopsys.com
Reviewed-by: Romain Naour <romain.naour@smile.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/x11r7/xserver_xorg-server: add sysv init script

We need ":0.0 vt01" to make sure Xorg finds the correct terminal.

Quoting https://www.x.org/archive/X11R6.8.0/doc/Xorg.1.html

-allowMouseOpenFail
    Allow the server to start up even if the mouse device can't be
    opened or initialised.

We need this option in case BR2_PACKAGE_XDRIVER_XF86_INPUT_MOUSE was
not enabled.

-noreset
    prevents a server reset when the last client connection is closed.

We need this option to prevent auto-close of Xorg.

Tested with a qemu image and Kodi.

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
[Arnout: copy an abbreviate version of the explanation above to the
         script.]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

support/check-uniq-files: support weird locales and filenames

Currently, when a filename contains characters not representable in the
user's locale, we fail hard, especially when the host python is python3.

This is because python2 and python3 handle encoding/decoding strings
differently, with python3 presumable doing the right thing, but it
breaks on some systems, while python2 presumable does the wrong thing,
but it works everywhere. (Just joking, obviously...)

Part of the issue being that the csv reader in python2 is broken with
UTF8.

We fix the issue by ditching the csv reader, and simply read the file in
binary mode, manually partitioning the lines on the first comma.

Then, we use the binary-encoded (really, un-encoded) package names and
filenames as values and keys, respectively.

Finally, for each filename or package we need to print, we try to decode
them with the defaults for the user settings, but catch any decoding
exception and fall back to dumping the raw, binary values. Which codec
is used by default differs between Python version, but in all cases
something sane is printed at least.

Thanks a lot to Arnout for the live help doing this patch. :-)

Reported-by: Jaap Crezee <jaap@jcz.nl>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Jaap Crezee <jaap@jcz.nl>
[Arnout: commit log improvement]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

configs/raspberrypi0w_defconfig: new defconfig for Raspberry Pi Zero W

Up to now, Raspberry Pi Zero W (rpi0w) could be built with the normal
raspberrypi0_defconfig. However, then you don't have support for the
Bluetooth out-of-the-box, which makes using a W a bit pointless.

Therefore, create a separate defconfig for the W. It is a copy of
raspberrypi0_defconfig with the following changes:
 - Add DT overlays (from rpi-firmware) to support Bluetooth.
 - Pass --add-pi3-miniuart-bt-overlay to post-image.sh.

Since there is now a separate raspberrypi0w_defconfig, the W support
can be removed from raspberrypi0_defconfig.

Signed-off-by: Yves Deweerdt <yves.deweerdt.linux@gmail.com>
[Arnout:
 - Bump to same kernel version as raspberrypi0.
 - Remove redundant comment.
 - Remove redundant default BR2_PACKAGE_RPI_FIRMWARE_INSTALL_DTB_OVERLAYS=y.
 - Improve commit log.
 - Refresh .gitlab-ci.yml.
]
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>

package/openocd: fix fallout after no-C++ fixups

Commit 4cd1ab158 (core: alternate solution to disable C++) made use of a
non-existent 'no' binary when C++ is not available in the toolchain.

However, some packages, like jimtcl as bundled in opeocd, really want to
find the binary that $CXX contains.

Revert openocd to use 'false' instead of 'no'.

Fixes:
    http://autobuild.buildroot.org/results/cbd/cbd5ab97fb0659968ff628461130627cf1745955/

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/jimtcl: fix fallout after no-C++ fixups

Commit 4cd1ab158 (core: alternate solution to disable C++) made use of a
non-existent 'no' binary when C++ is not available in the toolchain.

However, some packages, like jimtcl, really want to find the binary that
$CXX contains.

Revert jimtcl to use 'false' instead of 'no'.

Fixes:
    http://autobuild.buildroot.org/results/54f/54f3df03551fbdf293d33dc1e3f08005faa15321/

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

picocom: bump to version 3.1

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

woff2: fix static builds

Include a patch to make CMake correctly find the Brotli libraries when
they have been built as static libraries.

Fixes:
- http://autobuild.buildroot.net/results/f1c4b5aeb12af7b7a3e8ae01c219004ecd9befd6/
- http://autobuild.buildroot.net/results/74d20ff38766466623cc4a9eb18afcda831bc20b/

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

brotli: fix building of static libraries

Include a patch to make CMake honor the BUILD_SHARED_LIBS which
Buildroot sets to choose the kind of libraries to build.

Fixes:
- http://autobuild.buildroot.net/results/f1c4b5aeb12af7b7a3e8ae01c219004ecd9befd6/
- http://autobuild.buildroot.net/results/74d20ff38766466623cc4a9eb18afcda831bc20b/

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

support/config-fragments/autobuild: fix SSP in br-nios2-glibc

Commit c8680956819fae8776d7bd6d1f0e67a7b6436672 ("toolchain: fix
detection of SSP support") fixed the SSP check so that it does the
correct thing for nios2 toolchains. While this commit fixed the
description of the Sourcery NIOSII toolchain, it didn't fix the
description for the autobuilders of the br-nios2-glibc toolchain,
causing some build failures. This commit adjusts br-nios2-glibc.config
to indicate that the toolchain doesn't have SSP support.

Fixes:

  http://autobuild.buildroot.net/results/6c44e328b7bffd8474d29d5bdf1ea109ec15f4ad/

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

ncmpc: bump to version 0.29

Since version 0.28 ncmpc only supports the meson build.
Therefore, adapt to package infrastructure according to the user manual.

Add a hash for the license file.

Add a nmpc entry for myself in DEVELOPERS

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

docker-compose: bump version to 1.20.1

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

python-docker: bump version to 3.1.4

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

zstd: fix legal info

zstd is dual licensed under BSD-3-Clause or GPL-2.0 as per README.md
and source files license header.

Cc: Andrey Smirnov <andrew.smirnov@gmail.com>
Signed-off-by: Rahul Bedarkar <rahulbedarkar89@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

utils/genrandconfig: adjust to core-dependencies removal

In commit ea9669fffae7d807ccb3810eb6c662cbb9e8f96f ("core: kill
DEPENDENCIES_HOST_PREREQ"), the core-dependencies make target was
removed, and is now named just "dependencies".

This broke the utils/genrandconfig script, and this commit intends to
fix that.

Since this script is part of the Buildroot tree, it is provided
together with Buildroot, so we don't need to support the legacy
core-dependencies target. Someone checking out an older Buildroot will
have a working setup, with support/dependencies exposing
core-dependencies and utils/genrandconfig using core-dependencies. The
only broken situation will be between
ea9669fffae7d807ccb3810eb6c662cbb9e8f96f and this commit, but that's
not a sufficient reason to add some backward compatibility code.

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

opus-tools: update license

As per COPYING file, opusinfo is licensed under GPL version 2.

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

opus-tools: add hash for license file

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

opusfile: add hash for license file

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

opus: add hash for license file

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

libnss: bump version to 3.35

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

libnspr: bump version to 4.19

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

core: alternate solution to disable C++

Some packages that use libtool really need some love to be able to
disable C++ support.

This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX
is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP
will want a C++ preprocessor that works on valid input *and* fail on
invalid input.

So, providing 'false' as the C++ compiler will then require that we do
have a working C++ preprocessor. Which is totally counter-productive
since we do not have a C++ compiler to start with...

bd39d11d2e (core/infra: fix build on toolchain without C++) was a
previous attempt at fixing this, by using the host's C++ preprocessor.

However, that is very incorrect (that's my code, I can say so!) because
the set of defines will most probably be different for the host and the
target, thus causing all sorts of trouble. For example, on ARM we'd have
to include different headers for soft-float vs hard-float, which is
decided based on a macro, which is not defined for x86, and thus may
redirect to the wrong (and missing) header.

Instead, we notice that libtool uses the magic value 'no' to decide that
a C++ compiler is not available, in which case it skips the call to
AC_PROG_CXXCPP.

Given that 'no' is not provided by any package in Debian and
derivatives, as well as in Fedora, we can assume that no system will
have an executable called 'no'. Hence, we use that as a magic value to
disable C++ detection altogether.

Fixes: #10846 (again)
Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Damien Riegel <damien.riegel@savoirfairelinux.com>
Cc: Peter Seiderer <ps.report@gmx.net>
Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Tested-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/dovecot-pigeonhole: bump version to 0.5.1

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

package/dovecot: bump version to 2.3.1

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

libunistring: bump to version 0.9.9

Drop --{enable,disable}-threads; threading support is detected
automatically.

Add GPLv3 license file.

Add license files hash.

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

busybox: bump version to 1.28.2

Drop 0003-tar-unzip-postpone-creation-of-symlinks-with-suspici.patch now upstream.

>From the release notes:

Bug fix release.  1.28.2 has fixes for tcpsvd (fixed fallout from
opt_complementary removal), udhcpd (do not ignore SIGTERM), tar and unzip
(reverted to previous, more permissive symlink handling), ssl_client (fixed
option parsing).

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

libsigsegv: bump to version 2.12

Add license file hash.

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

bash: bump to version 4.4.18

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

configs/atmel: bump to linux4sam_5.8

Bump at91sam9x5ek, atmel_sama5d27_som1_ek, atmel_sama5d2_xplained,
atmel_sama5d3_xplained, and atmel_sama5d4_xplained all variants to
linux4sam_5.8. The 3 foundation components have their tags changed
(AT91Bootstrap, U-Boot, Linux kernel).

at91bootstrap 3.8.10 is required to support gcc7.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
Cc: Nicolas Ferre <nicolas.ferre@microchip.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libopenssl: security bump to version 1.0.2o

Fixes the following security issues:

Constructed ASN.1 types with a recursive definition could exceed the stack
(CVE-2018-0739)

Constructed ASN.1 types with a recursive definition (such as can be found in
PKCS7) could eventually exceed the stack given malicious input with
excessive recursion.  This could result in a Denial Of Service attack.
There are no such structures used within SSL/TLS that come from untrusted
sources so this is considered safe.

Incorrect CRYPTO_memcmp on HP-UX PA-RISC (CVE-2018-0733)

Because of an implementation bug the PA-RISC CRYPTO_memcmp function is
effectively reduced to only comparing the least significant bit of each
byte.  This allows an attacker to forge messages that would be considered as
authenticated in an amount of tries lower than that guaranteed by the
security claims of the scheme.  The module can only be compiled by the HP-UX
assembler, so that only HP-UX PA-RISC targets are affected.

rsaz_1024_mul_avx2 overflow bug on x86_64 (CVE-2017-3738)

This issue has been reported in a previous OpenSSL security advisory and a
fix was provided for OpenSSL 1.0.2.  Due to the low severity no fix was
released at that time for OpenSSL 1.1.0.  The fix is now available in
OpenSSL 1.1.0h.

There is an overflow bug in the AVX2 Montgomery multiplication procedure
used in exponentiation with 1024-bit moduli.  No EC algorithms are affected.
Analysis suggests that attacks against RSA and DSA as a result of this
defect would be very difficult to perform and are not believed likely.
Attacks against DH1024 are considered just feasible, because most of the
work necessary to deduce information about a private key may be performed
offline.  The amount of resources required for such an attack would be
significant.  However, for an attack on TLS to be meaningful, the server
would have to share the DH1024 private key among multiple clients, which is
no longer an option since CVE-2016-0701.

This only affects processors that support the AVX2 but not ADX extensions
like Intel Haswell (4th generation).

For more details, see https://www.openssl.org/news/secadv/20180327.txt

The copyright year changed in LICENSE, so adjust the hash to match.

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

e2fsprogs: bump to version 1.44.1

From the release notes (https://lwn.net/Articles/750103/):

E2fsprogs 1.44.0 introduced a regression introduced which caused e2fsck
to fail to support HTree directories on big-endian systems.  Fix how we
read block numbers for internal htree nodes.

Removed a potential memory leak from fsck.

E2image now correctly creates e2image files for bigalloc file systems.

Dumpe2fs and debugfs now correctly support e2image files for file
systems that have the meta_bg option enabled.

E2fsck and debugfs now correctly handle delete inodes (including
processing the orphaned inode list in the case of e2fsck) for bigalloc
file systems.  (Addresses Google Bug: #73795618)

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

sngrep: fix libgcrypt handling

Fixes:
http://autobuild.buildroot.net/results/f1c6494133806b9fc26ae3ce9e9c6a22fa2eda6f/

Commit 6205b75873c (sngrep: gnutls support also needs libgcrypt) ensured
that --with-gnutls is only used when both gnutls and libgcrypt are enabled,
but it didn't ensure libgcrypt gets built before sngrep or told the
configure script where to find libgcrypt-config, breaking the build.

Fix both issues.

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

gawk: bump to version 4.2.1

Enable static build. gawk builds statically just fine. The shared
library check has been removed in 4.1.1.

Remove the symlink force hunk from the no-versioned patch. The only user
of LN is in the part that this patch disables.

Add license file hash.

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

package/php: security bump to version 7.2.4

Fixes https://bugs.php.net/bug.php?id=75605, no CVE-ID yet.

Removed patch 0008, applied upstream. Re-numbered patch 0009.

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

time: bump to version 1.9

Fix seed variable name for the wait3 system call AC_TRY_RUN test.

Remove manual installation; not needed anymore.

Cc: Will Newton <will.newton@gmail.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

cjson: bump to version 1.7.5

Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/{mesa3d, mesa3d-headers}: bump version to 18.0.0

Rebased 0002-Fix-endianess-detection-with-musl-based-toolchains.patch

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Tested-by: Joseph Kogut <joseph.kogut@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux-headers: bump 4.{1, 4, 9, 14, 15}.x series

Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux: bump default to version 4.15.14

Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

modem-manager: remove wrong gnutls dependency comment

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Acked-by: Petr Vorel <petr.vorel@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

xerces: add upstream security fix

CVE-2017-12627: dereference of a NULL pointer while processing the path
to the DTD.

xerces 3.2.1 includes this patch. But this version also added
AC_RUN_IFELSE to its configure script, making cross compilation harder.

Switching to cmake is also problematic since the minimum required cmake
version is 3.2.0. The host dependencies check currently allows minimum
cmake version 3.1.

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

ympd: add optional openssl dependency

Fixes:
 - http://autobuild.buildroot.net/results/d9b1375f5f8ecde8fef027344a5d5b288f7cee7c

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

libcdio: bump to version 2.0.0

Add license hash

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

lshw: use data files provided by hwdata

lshw is seldomly released, so its data files become easily
outdated. Instead, this commit makes use of the data files provided by
hwdata. This is easily possible because lshw looks for the files in
several directories, including /usr/share/hwdata, where the hwdata
collection is installed.

We remove the entire /usr/share/lshw directory, where the not very
up-to-date data files where installed. Four files were installed
there: pci.ids, usb.ids (which are now provided by hwdata), manuf.txt
and oui.txt (which are not used at run time).

Signed-off-by: Carlos Santos <casantos@datacom.ind.br>
[Thomas:
 - rework commit log
 - replace patch by a simple removal of the /usr/share/lshw directory
 - add "runtime" comment in Config.in for the BR2_PACKAGE_LSHW
   dependency]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/avahi: Add sysusers.d snippet

Signed-off-by: Chris Lesiak <chris.lesiak@licor.com>
Reviewed-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/avahi: Set the path to D-Bus system.d directory

Since D-Bus 1.9.18, the recommended location for the system and
session busses configuration files is /usr/share instead of /etc. From
the D-Bus NEWS file:

D-Bus 1.9.18 (2015-07-21)
==

The “Pirate Elite” release.

Configuration changes:

• The basic setup for the well-known system and session buses is now done
   in read-only files in ${datadir}, moving a step closer to systems
   that can operate with an empty /etc directory. In increasing order
   of precedence:

   · ${datadir}/dbus-1/s*.conf now perform the basic setup such as setting
     the default message policies.
   · ${sysconfdir}/dbus-1/s*.conf are now optional. By default
     dbus still installs a trivial version of each, for documentation
     purposes; putting configuration directives in these files is
     deprecated.
   · ${datadir}/dbus-1/s*.d/ are now available for third-party software
     to install "drop-in" configuration snippets (any packages
     using those directories should explicitly depend on at least this
     version of dbus).
   · ${sysconfdir}/dbus-1/s*.d/ are also still available for sysadmins
     or third-party software to install "drop-in" configuration snippets
   · ${sysconfdir}/dbus-1/s*-local.conf are still available for sysadmins'
     overrides

   ${datadir} is normally /usr/share, ${sysconfdir} is normally /etc,
   and "s*" refers to either system or session as appropriate.

Therefore, this commit adjusts the Avahi package to install the D-Bus
related files to /usr/share/dbus-1/system.d.

Signed-off-by: Chris Lesiak <chris.lesiak@licor.com>
Reviewed-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

configs/arcturus_ucls1012a: new defconfig

The uCLS1012A-SOM product family (ucls1012a) is an Arcturus Networks Inc.
64bit ARM Cortex-A53 class System on Module powered by a NXP QorIQ LS1012A
Low Power Communication Processor.  This 314 pin MXM3.0/SMARC module card
contains DDR3, QSPI NOR Flash, eMMC NAND Flash, optional Audio CODEC and is
available with 1 or 2 Gig-Ethernet PHYs.

Signed-off-by: Oleksandr Zhadan <oleks@arcturusnetworks.com>
Signed-off-by: Michael Durrant <mdurrant@arcturusnetworks.com>
Changes v1 -> v2:
    - updated kernel and u-boot source repositories to github

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

Config.in: Document BR2_CCACHE_DIR override

This variable, like BR2_DL_DIR, is designed to be overridable from the
environment.  Unlike BR2_DL_DIR, it is not documented as such in the
Config.in help text.  Do so now.

Signed-off-by: Trent Piepho <tpiepho@impinj.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

rpi-bt-firmware: remove compatibility symlink

Commit 361d1b969ecdba1a8a62613626f90fc025b94660 defines FIRMWARE_DIR for
bluez5_utils. The compatibility symlink for the firmware is no longer
needed.

Signed-off-by: Ryan Coe <bluemrp9@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

Config.in: add -Ofast option

-Ofast (introduced in GCC 4.6) It combines the existing optimization level -O3
with options that can affect standards compliance but result in better optimized
code. For example, -Ofast enables -ffast-math.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/systemd: add upstream build fix

Fixes:

  http://autobuild.buildroot.org/results/4c439ee000354f90b4e59ee4006530f77263db47/

Signed-off-by: Stefan Becker <chemobejk@gmail.com>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

prosody: replace tweaked configuration by patches

Upgrading a tweaked configuration is painful, so stay with the
upstream configuration which is already installed, and just apply a
few patches.

Everybody could do its own configuration in its root filesystem
overlay.

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
[Thomas: squash patches.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

prosody: bump to version 0.10.0

this version is not restricted to Lua 5.1

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

qemu: remove support for some PowerPC processors in host qemu

The 620, 630, and 970 are not supported at this time by qemu.

Signed-off-by: Adam Duskett <aduskett@gmail.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

libgpg-error: bump to version 1.28

Add upstream patch fixing arm/arm64 targets build failure.

Use the smaller bz2 compressed tarball.

Switch _SITE to https for better security and corporate firewall
compatibility.

Add license files hash.

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

pkg-cmake: enable toolchain install staging phase

Really fix commit 6729050f3a49dba42d2d881d31dbefd6b82b8dad

That commit added a script to "toolchain" post install staging hooks.
This will only be executed if TOOLCHAIN_INSTALL_STAGING is set to YES.

Signed-off-by: Stefan Becker <chemobejk@gmail.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

busybox: disable PAM in the config if linux-pam is not selected

Currently there is only logic to enable PAM when linux-pam is selected.
However, busybox will fail to build with PAM enabled if the linux-pam
package has not been built before. So we should forcibly disable PAM in
busybox in that case.

Normally this is not an issue since our default busybox config doesn't
have PAM enabled. However, if you enable linux-pam, then save the
busybox config to a custom configuration file, then disable linux-pam
again, and then do a "make clean; make", the build will fail. A more
practical situation where this can occur is when the same custom
busybox config is used in a Buildroot config with and without
linux-pam.

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

package/eudev: don't forcibly link with -lrt

This was carried over from udev back when it were an separate package,
but was not even needed by eudev 1.3 when we introduced it back in 2014
(937a95449ec: eudev: new package).

Drop it now.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Bernd Kuhls <bernd.kuhls@t-online.de>
Cc: Eric Le Bihan <eric.le.bihan.dev@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

brotli: disable unneeded check for C++ compiler

This adds a local patch, also submitted upstream, which modifies
CMakeLists.txt to avoid the check for the unneeded C++ compiler.

Fixes:
- http://autobuild.buildroot.net/results/68e05c634e62ba4b6f6eb8dc95eeb16faa58fd60

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

libfcgi: Use SPDX license identifier

The fcgi license is covered by SPDX, the identifier however is not
obvious.

For details, see https://spdx.org/licenses/OML.html

[Peter: add spdx.org link]
Signed-off-by: Alexander Dahl <post@lespocky.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Makefile: fix build break in sdk target

After commit 6729050f3a49dba42d2d881d31dbefd6b82b8dad nothing creates
$(HOST_DIR)/share/buildroot anymore, causing sdk to fail with:

 /bin/bash: .../output/host/share/buildroot/sdk-location: No such file or directory

Add creation of that directory to the "sdk" build steps itself.

Signed-off-by: Stefan Becker <chemobejk@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-pymodbus: new package

Signed-off-by: Falco Hyfing <hyfinglists@gmail.com>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

safeclib: fix build without wchar and host-doxygen

Add a patch (sent upstream) to add a --disable-doc option

Fixes:
 - http://autobuild.buildroot.net/results/e597843f2bb528da01d1e617929c818613b56cec

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

toolchain: fix detection of SSP support

GCC issues a warning message if -fstack-protector is passed but SSP is
not available, so in order to force the compilation to fail we must also
pass -Werror.

All external toolchains were verified. The only one whose configuration
incorrectly selected BR2_TOOLCHAIN_HAS_SSP was CodeSourcery NIOSII.

Fixes:
  http://autobuild.buildroot.net/results/ce8fe8ac9cf0db01ae15d476ea714ff176965cfb
  http://autobuild.buildroot.net/results/09ce8f05e28c0219f499ce55130e896cae0c8b45

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

setools: update to add sedta and seinfoflow

Update to add sedta and seinfoflow to setools

Signed-off-by: Jared Bents <jared.bents@rockwellcollins.com>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

python-networkx: new package

Update to add python-networkx package to buildroot. While 2.1
is the latest version, the current version of setools does not
work with versions past 1.11.

Signed-off-by: Jared Bents <jared.bents@rockwellcollins.com>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

python-decorator: update to add host package

Update to add host package for python decorator

Signed-off-by: Jared Bents <jared.bents@rockwellcollins.com>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

sam-ba: fix sam-ba symbolic link

Following the removal of $(HOST_DIR)/usr, the symbolic link from
$(HOST_DIR)/bin/sam-ba/ to $(HOST_DIR)/opt/sam-ba/sam-ba was broken,
so we fix this.

In addition, sam-ba being prebuilt, it comes in two separates
binaries, one for x86 and the other for x86_64, so we take this into
account as well.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
[Thomas:
 - add spaces around = signs
 - rework commit log.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/libss7: propagate dependency from dahdi-tools

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

eudev: prevent udev init script nonexistent directory error

The following error occures in the udev init script because the kernel config
may optionally not include uevent_helper.

/etc/init.d/S10udev: line 47: can't create /proc/sys/kernel/hotplug: nonexistent directory

Work around this by not trying to access the destination if it's not available.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

strace: bump to version 4.21

Download site moved to strace.io. No upstream provided hashes.

Drop the musl specific autoconf seed values. These values are now
detected correctly.

Add --enable-mpers=check to fix configure on aarch64 that does not
support binaries for the compat personality.

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

core: kill DEPENDENCIES_HOST_PREREQ

Now that DEPENDENCIES_HOST_PREREQ is no longer used anywhere, we can
kill it.

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

package/pkg-generic: handle host-fakedate as a regular dependency

This commit moves the host-fakedate dependency handling from
DEPENDENCIES_HOST_PREREQ to a proper regular dependency handled by the
package infrastructure.

host-fakedate is added as dependency to all packages, except
host-skeleton, because we depend on it.

In addition, we make sure that host-fakedate does not grow a
dependency on host-{tar,xz,lzip,ccache} to avoid circular
dependencies. host-fakedate does not need any extraction tool and does
not need to build C/C++ code (the source code is just a shell script
available in Buildroot).

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

package/pkg-generic: handle host-ccache as a regular dependency

This moves the host-ccache dependency handling from
DEPENDENCIES_HOST_PREREQ to a proper package dependency. When
BR2_CCACHE=y, we add host-ccache as a regular dependency of all
packages except:

 - The extractor packages host-tar, host-xz and host-lzip

 - host-ccache itself

 - host-skeleton, because all packages depend on it

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: Matt Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/pkg-generic: handle host-lzip as an extract dependency

This moves the host-lzip dependency handling from
DEPENDENCY_HOST_PREREQ to an extract dependency.

To achieve that, check-host-lzip.mk fills in the
BR2_LZIP_HOST_DEPENDENCY with host-lzip if building a host-lzip is
needed. The name BR2_LZIP_HOST_DEPENDENCY has been chosen because it
matches the name BR2_CMAKE_HOST_DEPENDENCY already used in
check-host-cmake.mk.

The BR2_LZIP_HOST_DEPENDENCY is added to all packages, except:

 - host-lzip, because we would otherwise depend on ourself.

 - host-tar, because lzip itself is delivered as a tarball, so we need
   to have host-lzip depend on host-tar, and not host-tar depend on
   host-lzip

 - host-skeleton, because we need to have host-lzip depend on
   host-skeleton, and not the opposite.

We also mutually exclude host-lzip and host-xz from dependending on
each other, to avoid a circular dependency.

In addition, we modify lzip.mk to explicitly build host-lzip without
ccache. We generally took the approach of building host-ccache *after*
all the extractors have been built.

[Peter: fix s/host-tar/host-lzip/ typo, fix s/xz/lzip/ typo]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/pkg-generic: handle host-xz as an extract dependency

This moves the host-xz dependency handling from
DEPENDENCY_HOST_PREREQ to an extract dependency.

To achieve that, check-host-xz.mk fills in the
BR2_XZ_HOST_DEPENDENCY with host-xz if building a host-xz is
needed. The name BR2_XZ_HOST_DEPENDENCY has been chosen because it
matches the name BR2_CMAKE_HOST_DEPENDENCY already used in
check-host-cmake.mk.

The BR2_XZ_HOST_DEPENDENCY is added to all packages, except:

 - host-xz, because we would otherwise depend on ourself.

 - host-tar, because xz itself is delivered as a tarball, so we need
   to have host-xz depend on host-tar, and not host-tar depend on
   host-xz

 - host-skeleton, because we need to have host-xz depend on
   host-skeleton, and not the opposite.

In addition, we modify xz.mk to explicitly build host-xz without
ccache. We generally took the approach of building host-ccache *after*
all the extractors have been built.

[Peter: fix s/host-tar/host-xz/ typo]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/pkg-generic: handle host-tar as an extract dependency

This moves the host-tar dependency handling from
DEPENDENCY_HOST_PREREQ to an extract dependency.

To achieve that, check-host-tar.mk fills in the
BR2_TAR_HOST_DEPENDENCY variable with host-tar if building a host-tar
is needed. The name BR2_TAR_HOST_DEPENDENCY has been chosen because it
matches the name BR2_CMAKE_HOST_DEPENDENCY already used in
check-host-cmake.mk.

The BR2_TAR_HOST_DEPENDENCY is added to all packages, except host-tar
itself (obviously) and host-skeleton, because we depend on
host-skeleton to install host-tar properly in HOST_DIR.

In addition, we modify tar.mk to explicitly build host-tar without
ccache: since ccache source code is available as a tarball, ccache
will obviously depend on host-tar if the system tar is insufficient.

Finally, to make things really clean, we also add
$(BR2_TAR_HOST_DEPENDENCY) to the dependencies of the tar filesystem
format, since it requires tar, so we'd better make sure we have a
suitable tar.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: Matt Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/pkg-generic: add the concept of extract dependency

Extract dependencies are dependencies that must be ready before the
extract step of a package, i.e for tools that are needed to extract
packages themselves. Current examples of such tools are host-tar,
host-lzip and host-xz.

They are currently handled through DEPENDENCIES_HOST_PREREQ. However,
this mechanism has a number of drawbacks:

 - First and foremost, because host-tar/host-lzip/host-xz are not
   listed in the dependencies of packages, the package infrastructure
   does not know it should rsync them in the context of per-package
   SDK.

 - Second, there is no dependency handling *between* them. I.e, we
   have no mechanism that says host-tar should be built before
   host-lzip, while it is in fact the case: if you need to build
   host-lzip, you need to extract a tarball, so you may need host-tar
   if your system tarball is not capable enough.

For those reasons, it makes sense to add explicit support for "extract
dependencies" in the package infrastructure, through the
<pkg>_EXTRACT_DEPENDENCIES variable. It is unlikely this variable will
ever be used by a package .mk file, but it will be used internally by
the package infrastructure.

[Peter: fix typo in manual]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Tested-by: Matt Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

pkg-cmake: install CMake files as part of a package

Currently, the toolchainfile.cmake and Buildroot.cmake files are
installed outside of any package, just triggered by the toolchain
target.

As part of the per-package SDK effort, we are trying to avoid anything
that installs to the global $(HOST_DIR), and this is one of the
remaining files installed in $(HOST_DIR) outside of any package. We
fix this by installing such files as part of the toolchain package
post-install staging hooks.

Yes, a post-install staging hook to install things to $(HOST_DIR) is a
bit weird, but the toolchain infrastructure is made of target packages
only, and they all install a lot of stuff to $(HOST_DIR) already.

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