toolchain: improve musl check to support static toolchains

The check_musl function currently builds a program and verifies if the
program interpreter starts with /lib/ld-musl. While this works fine
for dynamically linked programs, this obviously doesn't work for a
purely static musl toolchain such as [1].

There is no easy way to identify a toolchain as using the musl C
library. For glibc, dynamic linking is always supported, so we look at
the dynamic linker name. For uClibc, there is a distinctive
uClibc_config.h header file. There is no such distinctive feature in
musl.

We end up resorting to looking for the string MUSL_LOCPATH, which is
used by musl locale_map.c source file. This string has been present in
musl since 2014. It certainly isn't a very stable or convincing
solution to identify the C library as being musl, but it's the best we
could find.

Note that we are sure there is a libc.a file, because the
check_unusable_toolchain function checks that there is a such a file.

[1] http://autobuild.buildroot.net/toolchains/tarballs/br-arm-musl-static-2018.05.tar.bz2

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

busybox: don't clobber dangling symlinks

We sometimes create dangling symlinks in the target directory. That is
because we need canonical targets, as relative targets don't work well
with BR2_ROOTFS_MERGED_USR. For example, the vim package installs the
/bin/vi symlink to /usr/bin/vim. This symlink might be dangling when the
build host has no vim installed there.

Patch the busybox install.sh script to avoid clobber of dangling
symlinks.

Fixes:
http://autobuild.buildroot.net/results/796/796107430db6545401d9926e84f19eaf2040b756/

Cc: Adam Duskett <aduskett@gmail.com>
Cc: Carlos Santos <casantos@datacom.com.br>
Cc: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

pv: fix build failures when building under a path containing 'yes'

Depending on the configuration, the cpp output may contain the string 'yes'
in a comment if built under a path containing 'yes', confusing the _AIX
test:

${CROSS}-cpp conftest.h
\# 1 "conftest.h"
\# 1 "<built-in>"
\# 1 "<command-line>"
\# 31 "<command-line>"
\# 1 "/home/peko/source/buildroot/output-yes/host/x86_64-buildroot-linux-gnu/sysroot/usr/include/stdc-predef.h"
\# 32 "<command-line>" 2
\# 1 "conftest.txt"

If misdetected, the configure script adds -lc128 to LIBS, causing the
AC_CHECKS_FUNCS check for stat64 to fail, which in turn causes compilation
errors about redefinition of symbols:

In file included from ./src/include/pv-internal.h:9:0,
                 from src/pv/file.c:5:
./src/include/config.h:76:18: error: redefinition of 'struct stat'
  #  define stat64 stat
                   ^
Fix it by only matching on 'yes' on a line by itself.

As pv doesn't cleanly autoreconf (it doesn't use automake and configure.in
is located in subdir), instead directly patch configure.

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

systemd: add optional dependency on elfutils

systemd can use elfutils when available, so this commit adds the
detection of this library.

Signed-off-by: Keith Mok <ek9852@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

download: fix file:// BR2_PRIMARY_SITE (download cache)

wget is the only downloader currently usable with BR2_PRIMARY_SITE, and that
doesn't work at all for file:// URLs. The symptoms are these:

support/download/dl-wrapper -c '2.4.47' -d '/PATH/build/sw/source/attr' -D '/PATH/build/sw/source' -f 'attr-2.4.47.src.tar.gz' -H 'package/attr//attr.hash' -n 'attr-2.4.47' -N 'attr' -o '/PATH/build/sw/source/attr/attr-2.4.47.src.tar.gz'  -u file\|urlencode+file:///NFS/buildroot_dl_cache/attr -u file\|urlencode+file:///NFS/buildroot_dl_cache -u http+http://download.savannah.gnu.org/releases/attr -u http\|urlencode+http://sources.buildroot.net/attr -u http\|urlencode+http://sources.buildroot.net  --
file:///NFS/buildroot_dl_cache/attr/attr-2.4.47.src.tar.gz: Unsupported scheme `file'.
ERROR: attr-2.4.47.src.tar.gz has wrong sha256 hash:
ERROR: expected: 25772f653ac5b2e3ceeb89df50e4688891e21f723c460636548971652af0a859
ERROR: got     : e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
ERROR: Incomplete download, or man-in-the-middle (MITM) attack

In the case of custom Linux kernel versions, this is fatal, because there isn't
necessarily a hash file to indicate that wget's empty tarball is wrong.

This seems to have been broken by commit c8ef0c03b0b, because:
1. BR2_PRIMARY_SITE always appends "urlencode" (package/pkg-download.mk)
2. Anything with the "|urlencode" suffix in $uri will end up using wget due to
   the backend case wildcarding.
3. The wget backend rejects file:/// URLs ("unsupported scheme"), and we end up
   with an empty .tar.gz file in the downloads directory.

Fix that by shell-extracting the backend name from the left of "|". I'm not
positive if all URLs will have a "|", so this code only looks for a "|" left of
the "+".

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

libssh: fix download location

The current download location fails, and Buildroot falls back to
sources.b.o:

--2018-08-20 23:41:39--  https://red.libssh.org/attachments/download/218/libssh-0.7.5.tar.xz
Resolving red.libssh.org (red.libssh.org)... 78.46.80.163
Connecting to red.libssh.org (red.libssh.org)|78.46.80.163|:443... connected.
The certificate's owner does not match hostname ‘red.libssh.org’
--2018-08-20 23:41:39--  http://sources.buildroot.net/libssh/libssh-0.7.5.tar.xz
Resolving sources.buildroot.net (sources.buildroot.net)... 104.25.211.19, 104.25.210.19, 2400:cb00:2048:1::6819:d313, ...
Connecting to sources.buildroot.net (sources.buildroot.net)|104.25.211.19|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 351632 (343K) [application/x-xz]

This commit fixes the download location:

--2018-08-20 23:43:04--  https://www.libssh.org/files/0.7/libssh-0.7.5.tar.xz
Resolving www.libssh.org (www.libssh.org)... 87.98.168.187, 2001:41d0:2:f80c::4
Connecting to www.libssh.org (www.libssh.org)|87.98.168.187|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 351632 (343K) [application/x-tar]

This patch is extracted from a contribution from Bernd Kuhls who was
also bumping the package at the same time
(http://patchwork.ozlabs.org/patch/959192/).

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

mutt: fix static build with libidn2 and libunistring

When libidn2 is statically build with libunistring support, mutt needs
to add -lunistring to LIBS.
To do that, add a call to PKG_CHECK_MODULES to retrieve this information
from libidn2.pc

Fixes:
 - http://autobuild.buildroot.net/results/177da8f4798f69298db5385957184f1c53cca923

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

zeromq: fix build on m68k_cf

An internal compiler error is raised on m68k_cf at dwarf2cfi.c:2752 in
connect_traces. Error can be fixed by adding -fno-defer-pop, see
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58864

Fixes:
 - http://autobuild.buildroot.net/results/dad241acbe59b1c5a24a0a2f3da6b12a553aec84

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

package/pkg-download: quiet the dl-wrapper call

The download wrapper call is currently always being displayed, even
without V=1, which is a bit annoying. It shows something like this:

thomas@windsurf:~/projets/buildroot (master)$ make tslib-source
>>> tslib 1.16 Downloading
PATH="/home/thomas/projets/buildroot/output/host/bin:/home/thomas/projets/buildroot/output/host/sbin:/usr/local/bin:/usr/bin:/bin:/home/thomas/.rvm/bin:/usr/local/sbin:/usr/sbin:/home/thomas/.rvm/bin:/home/thomas/sys/bin:/home/thomas/.gem/ruby/2.1.0/bin:/home/thomas/.rvm/bin" BR2_DL_DIR=/home/thomas/dl BUILD_DIR=/home/thomas/projets/buildroot/output/build O=/home/thomas/projets/buildroot/output flock /home/thomas/dl/tslib/ support/download/dl-wrapper -c '1.16' -d '/home/thomas/dl/tslib' -D '/home/thomas/dl' -f 'tslib-1.16.tar.xz' -H 'package/tslib//tslib.hash' -n 'tslib-1.16' -N 'tslib' -o '/home/thomas/dl/tslib/tslib-1.16.tar.xz'  -u https+https://github.com/kergoth/tslib/releases/download/1.16 -u http\|urlencode+http://sources.buildroot.net/tslib -u http\|urlencode+http://sources.buildroot.net  --

Let's silence this dl-wrapper call by prepending with $(Q).

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

oracle-mysql: fix hostname path

MySQL detects on the build machine where the hostname program is
located, and uses this value in a number of configuration files and
scripts that are generated and installed in the target:

output/target$ grep -r "bin/hostname" *
etc/inittab:::sysinit:/bin/hostname -F /etc/hostname
usr/share/mysql/mysql.server:  pid_file=$datadir/mysqlmanager-`/usr/bin/hostname`.pid
usr/share/mysql/mysql.server:  server_pid_file=$datadir/`/usr/bin/hostname`.pid
usr/bin/mysql_install_db:hostname=`/usr/bin/hostname`
usr/bin/mysqld_safe:    err_log=$DATADIR/`/usr/bin/hostname`.err
usr/bin/mysqld_safe:  pid_file="$DATADIR/`/usr/bin/hostname`.pid"

However, the hostname on the build machine may not necessarily be at
the same location as the hostname program on the target. Buildroot has
its hostname program (coming from Busybox) in /bin, but some Linux
distributions (such as Fedora) use /usr/bin/hostname, causing the
incorrect hostname paths above.

This commit fixes that by passing the appropriate autoconf cache
variable value.

Signed-off-by: Christopher McCrory <chrismcc@gmail.com>
[Thomas: add commit log]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

liburiparser: bump to version 0.8.6

Version 0.8.6 is a bugfix release including a nasty bug that has
potential to crash applications when parsing certain URIs (like
"//:%aa@", excluding quotes).

For more details please check the change log at

    https://github.com/uriparser/uriparser/blob/uriparser-0.8.6/ChangeLog

Signed-off-by: Carlos Santos <casantos@datacom.com.br>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/libconfuse: security bump to version 3.2.2

Fixes CVE-2018-14447: https://github.com/martinh/libconfuse/issues/109

Signed-off-by: Frank Hunleth <fhunleth@troodon-software.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gr-osmosdr: add host-python-cheetah dependency

Since b7f1b030413a51637323a755cbd985b06026990e host-python-cheetah is no more a
dependency for gnuradio. So this package must be explicitly added for
gr-osmosdr.

Fix : http://autobuild.buildroot.net/results/aa208a69996548c15d937fc7c97d267162a56c48/

Signed-off-by: Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: set version in go linker

This is used when calling the mender client with the
-version option and it says "unknown" if not set in
linker.

Now it displays the following:

    # mender -version
    1.4.0
    runtime: go1.10.2

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: update legal info

Also added license checksums in mender.hash

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: use inventory and identity script from upstream source

These files are part of Mender sources and no point in keeping duplicate
files locally.

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: remove tenant.conf

Tenant Token is a configuration option that has to do with Hosted Mender,
where you you need to set this for the devices to connect to the
correct organization in a multi-tenant system.

The removal of tenant.conf usage (and /var/lib/mender/authtentoken)
was in Mender client version 1.2.0, where it was switched to be an mender.conf
option instead as the example above demonstrates. As the first version that was
integrated in Buildroot was 1.4.0, the inclusion of tenant.conf and the
creation of the symlink is not necessary.

Now it is specified as such in mender.conf:

Example:

/etc/mender/mender.conf
{
    TenantToken: "very long base64 encoded string"
}

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: select BR2_PACKAGE_UBOOT_TOOLS_FWPRINTENV

The Mender client uses fw_printenv/fw_setenv to manipulate the U-boot
environment, e.g to change the boot candidate after a update has been
done.

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: create directory containing Mender state script

Mender state-scripts are essentially "hooks" that can be provided to
influence the update flow.

They should be placed inside /etc/mender/scripts and the directory must
contain a file containing the current state-script format version. It is
currently "2".

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: provide sane values in mender.conf

The current values that are in mender.conf will actually
cause the Mender client to fail to start because of invalid
values.

Provide sane default values that at least allow the Mender client
to parse the configuration options and start running.

The values provided will actually work in a "Demo Environment",
see https://docs.mender.io/getting-started/create-a-test-environment.

Though an entry is required in /etc/hosts to resolve the URL to the
local IP address of the running demo server.

Example:

    echo "192.168.0.10 docker.mender.io s3.docker.mender.io" >> \
         /etc/hosts

Above is required because the demo certificate
(/etc/mender/server.crt) is created for https://docker.mender.io.

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/mender: fix install path for identity and inventory scripts

These scripts are used to generate the device identity and to populate
the the device inventory. The Mender client will call these and at least
the mender-device-identity is required to be present at the correct
location. Inventory scripts are there as an example and not
actually required.

Example output from identity script:

    $ ./mender-device-identity
    mac=de:ad:ca:fe:00:01
    cpuid=1112233

Example output from inventory script:

    $ ./mender-inventory-network
    mac_br-fbfdad18c33c=02:42:7e:74:96:85
    network_interfaces=br-fbfdad18c33c
    ipv4_br-fbfdad18c33c=172.21.0.1/16
    mac_enp0s25=de:ad:be:ef:bb:05
    network_interfaces=enp0s25
    ipv4_enp0s25=123.22.0.197/16
    ipv4_enp0s25=10.20.20.105/16
    ipv6_enp0s25=fe80::2aad:beff:feef:bb05/64

Inventory and device identity data is presented on the Mender server
front-end.

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

boot/uboot: needs host-{flex,bison}

Recent U-Boot no longer ship the flex/bison generated kconfig parser, as
of commit e91610da7c8a9fe42f3e5a75f06c3d1a0cb5f815 (kconfig: re-sync
with Linux 4.17-rc4).

So, add the conditional kconfig dependencies, as we just did for the
kernel.

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: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

linux: kconfig needs the toolchain

Starting with linux-4.18, the kconfig from the kernel can call
to the compiler to test its capabilities; see:

    https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/scripts/Kconfig.include

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Jan Kundrát <jan.kundrat@cesnet.cz>
Tested-by: Jan Kundrát <jan.kundrat@cesnet.cz>
Reviewed-by: Jan Kundrát <jan.kundrat@cesnet.cz>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

linux: host-{flex, bison} only needed for DTS

host-{flex,bison} are only needed to generate the dtc parser, so we
don't need them if the kernel does not have support for device tree.

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: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

linux: kconfig may need host-{flex, bison} to build the configurators

Rely on the system provided ones if avalable, and only resort to use our
owns if the sytem does not provide them.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Jan Kundrát <jan.kundrat@cesnet.cz>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

support/dependencies: check for system-provided bison and flex

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: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/bison: fix build on host with glibc-2.28

Similarly to c48f8a6462 (package/m4: fix build on host with glibc-2.28),
backport the two fixes fromn gnulib upstream, that allows building
host-bison on systems using glibc 2.28.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reported-by: c32 on IRC
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

Update for 2018.08-rc2

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

qt5base: fix double-conversion for or1k

Fixes [1]:

  ../3rdparty/double-conversion/include/double-conversion/utils.h:81:2: error: #error Target architecture was not detected as supported by Double-Conversion.
  #error Target architecture was not detected as supported by Double-Conversion.

[1] http://autobuild.buildroot.net/results/a3535cdf5e91df011a59a4b9f60d69195f5efdcb

Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core: drop useless assignments to BISON and FLEX

They were added back in 5432f26f0 (Adding Central config.cache options),
supposedly to be able to cache the result of configure tests, but they
were never, ever referenced anywhere in our code... Besides, we dropped
the idea of getting a configure cache long ago now (it does not work)...

They are causing spurious error messages on some distros (e.g. Fedora)
which use GNU's which (whatever package that comes from), while it is
silent on other distros (e.g. Ubuntu) which use debianutils' which.

Drop them.

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

python-pyqt5: add Qt 5.11 compatibility patch

This commit adds a patch to python-pyqt5 to make it build properly
against Qt 5.11.

PyQt5 is using a dual-licensing model, and the commercial company
behind it (RiverBank) only provides release tarballs, and no public
Git repository, so we cannot see the individual changes they make. By
diffing the PyQt5 5.10 and 5.11 releases, we could see that they opted
for dropping entirely support for the waitForEvents() method, rather
than keeping it for Qt < 5.11. We take the same approach in the below
patch, since this is anyway what will happen when we will bump to
PyQt5 5.11.

The patch is not Git-formatted, because there is no upstream Git
repository for this project.

Fixes:

  http://autobuild.buildroot.net/results/1f1e92374fe71a1d4343243db5f530c33db06698/

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

wireless_tools: Fix site URL does not work

Replace broken http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux
by https://hewlettpackard.github.io/wireless-tools.

[Peter: also adjust URL in Config.in]
Signed-off-by: Vadim Kochan <vadim.kochan@petcube.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/nodejs: security bump version to 8.11.4

Release notes:
https://nodejs.org/en/blog/vulnerability/august-2018-security-releases/

Fixes CVE-2018-12115, also CVEs were fixed in included OpenSSL code
which do not use for the target build.

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

package/ipsec-tools: add security patch to fix CVE-2016-10396

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

linux-headers: bump 4.{4, 9, 14, 17}.x series

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

linux: bump default to version 4.17.17

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

package/bind: security bump to version 9.11.4-P1

Fixes CVE-2018-5740: https://ftp.isc.org/isc/bind9/9.11.4-P1/CHANGES

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

core/pkg-kconfig: allow dependencies before configurators

Some users of kconfig need some packages to be built before their
kconfig infra be used.

For example, the linux kernel, starting with 4.16, needs flex and bison
to generate the parser code. Furthermore, starting with 4.18, it will
also need the cross-compiler before parsing the kconfig stuff, because
that calls the compiler to check its features.

Currently, this is broken, even the flex/bison ones, even though they
are listed, because there is no way to define dependencie that are
guaranteed before the (visual) configurators. For example:

    $ make distclean
    $ make menuconfig
      --> enable the linux kernel, choose a defconfig, save, exit
    $ make linux-menuconfig
    [...]
      HOSTCC  scripts/basic/fixdep
      HOSTCC  scripts/kconfig/conf.o
      YACC    scripts/kconfig/zconf.tab.c
    /bin/sh: bison: command not found
      LEX     scripts/kconfig/zconf.lex.c
    scripts/Makefile.lib:196: recipe for target 'scripts/kconfig/zconf.tab.c' failed
    make[3]: *** [scripts/kconfig/zconf.tab.c] Error 127
    make[3]: *** Waiting for unfinished jobs....
    /bin/sh: flex: command not found
    scripts/Makefile.lib:188: recipe for target 'scripts/kconfig/zconf.lex.c' failed
    make[3]: *** [scripts/kconfig/zconf.lex.c] Error 127
    Makefile:528: recipe for target 'rpc_defconfig' failed
    make[2]: *** [rpc_defconfig] Error 2
    linux/linux.mk:511: recipe for target '/home/ymorin/dev/buildroot/buildroot/output/build/linux-4.17.11/.config' failed
    make[1]: *** [/home/ymorin/dev/buildroot/buildroot/output/build/linux-4.17.11/.config] Error 2
    Makefile:79: recipe for target '_all' failed
    make: *** [_all] Error 2

So, we introduce a new type of dependencies for kconfig-based packages,
that are guaranteed to be built and installed before the (visual)
configurators are called.

Since those dependencies are phony targets and therefore always out of
date, a normal dependency would cause the .config target to be rebuilt
on each invocation of make. So we use an order-only pre-requisite, like
is done for the patch dependency.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Cc: Jan Kundrát <jan.kundrat@cesnet.cz>
Tested-by: Jan Kundrát <jan.kundrat@cesnet.cz>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

linux: explain why we need host-{flex, bison}

The commit that added the dependency on host-{bison,flex} did
so because the pre-generated kconfig parser source files were
removed from the kernel tree, in linux-4.16.

But then, in linux-4.17, the pre-generated dtc parser source
files were in turn removed as well.

So, document the two reasons why they are needed, so we don't
accidentally remove them when we (soon) introduce the kconfig
dependencies.

(Also fix the first assignment to LINUX_DEPENDENCIES to be a
simple assignement, not an append-assignment.)

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: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

boost: enable back log on powerpc with uclibc

boost-log builds fine with powerpc on uclibc nowadays so enable it back.
By removing this dependency, build failure on azmq is also fixed as this
package is currently selecting boost-log without fulfilling this
dependency

Fixes:
 - http://autobuild.buildroot.net/results/9c373d0b5a1a59e2271d71c480d55a90a67b84cb

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

boost: context needs thread without C++11 mutex

Extract from output/build/boost-1.67.0/libs/context/build/Jamfile.v2:

explicit cxx11_hdr_mutex_check ;
local cxx11_mutex = [ check-target-builds
      cxx11_hdr_mutex_check "C++11 mutex"
    :
    : <library>/boost/thread//boost_thread
  ] ;

So select boost_thread if gcc <= 4.7

Fixes:
 - http://autobuild.buildroot.net/results/fb046c04fe18bec973d120e4ab33971f32ba5769

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

package/mariadb: security bump to version 10.1.35

Fixes CVE-2018-3064, CVE-2018-3063, CVE-2018-3058 & CVE-2018-3066:
https://mariadb.com/kb/en/library/mariadb-10135-release-notes/

Added all hashes provided by upstream.

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

libselinux: use correct name and content for patch 0004

Commit 6288409642d8368104f916bd264d2cb042942dfa ("libselinux: add
patch to fix build with gcc < 4.7") introduced a patch, but its file
name was incorrect, so it was never applied. In addition, the patch
was generated against the Git repository of SELinux, which includes
all projects, and therefore it doesn't apply to the libselinux source
code extracted from the tarball: the "libselinux/" component path
needs to be removed from the patch.

This commit fixes both problems, which should finally and really fix:

  http://autobuild.buildroot.net/results/c3272566bb808e43bb77ec59cfe596f7e0fe9a64/

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

boost: thread needs atomic if GCC hasn't lock-free

When gcc has not always lock-free atomic ints:
 - lockfree boost::atomic_flag : no

boost thread needs boost atomic:
output/host/usr/bin/nios2-linux-readelf -d output/staging/usr/lib/libboost_thread.so

Dynamic section at offset 0x2cee0 contains 32 entries:
  Tag        Type                         Name/Value
 0x00000001 (NEEDED)                     Shared library: [libboost_system.so.1.67.0]
 0x00000001 (NEEDED)                     Shared library: [libboost_atomic.so.1.67.0]

Fixes:
 - http://autobuild.buildroot.net/results/5a7db292f1365f27e32695527701d5b827f60092
 - http://autobuild.buildroot.net/results/413dff87f5329d3c5180167a8711cdedea5dec67
 - http://autobuild.buildroot.net/results/a7eb4cbcdbd9412c344f45336dec58c82e84dab9

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

toolchain: add BR2_TOOLCHAIN_SUPPORTS_ALWAYS_LOCKFREE_ATOMIC_INTS hidden option

Add BR2_TOOLCHAIN_SUPPORTS_ALWAYS_LOCKFREE_ATOMIC_INTS variable and
use it in BR2_TOOLCHAIN_HAS_GCC_BUG_64735.

This new variable will be used to select boost atomic when lock-free
atomic ints are not available

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

package/samba4: Fix uClibc build on 64bit platforms by including stdint.h

Patch needed to be updated for samba-4.8.4

Fixes
http://autobuild.buildroot.net/results/140/1404a594583ff192d70086ef590f924472465c89/

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

package/php: security bump to version 7.2.9

Version 7.2.8 fixed CVE-2018-12882, CVE-2018-14883 & CVE-2018-14851:
http://www.php.net/ChangeLog-7.php#7.2.8

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

package/x11r7/xdriver_xf86-video-ati: fix glamor dependency

Glamor support needs egl, fix dependency after
https://git.buildroot.net/buildroot/commit/?id=5b4bcbdafbe8e7b42d4e085a0524d70665cdbaf5

Fixes
http://autobuild.buildroot.net/results/6ac/6acd5ad820a47fc3442d3e3a5d8d72b396fb6e41/

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

ruby: security bump to version 2.4.4

Fixes the following security issues:

CVE-2017-17405: Command injection vulnerability in Net::FTP (2.4.3):
https://www.ruby-lang.org/en/news/2017/12/14/net-ftp-command-injection-cve-2017-17405/

CVE-2017-17742: HTTP response splitting in WEBrick (2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/http-response-splitting-in-webrick-cve-2017-17742/

CVE-2018-6914: Unintentional file and directory creation with directory
traversal in tempfile and tmpdir (2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/unintentional-file-and-directory-creation-with-directory-traversal-cve-2018-6914/

CVE-2018-8777: DoS by large request in WEBrick (2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/large-request-dos-in-webrick-cve-2018-8777/

CVE-2018-8778: Buffer under-read in String#unpack (2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/buffer-under-read-unpack-cve-2018-8778/

CVE-2018-8779: Unintentional socket creation by poisoned NUL byte in
UNIXServer and UNIXSocket (2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/poisoned-nul-byte-unixsocket-cve-2018-8779/

CVE-2018-8780: Unintentional directory traversal by poisoned NUL byte in Dir
(2.4.4):
https://www.ruby-lang.org/en/news/2018/03/28/poisoned-nul-byte-dir-cve-2018-8780/

Multiple vulnerabilities in RubyGems (2.4.4):
https://www.ruby-lang.org/en/news/2018/02/17/multiple-vulnerabilities-in-rubygems/

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

eigen: Fix pkg-config file prefix error

eigen generates a pkg-config file which has a broken prefix
(@CMAKE_INSTALL_PREFIX@).  This broken prefix causes an incorrect path when
other packages call pkg-config --cflags eigen.

This patch fixes the prefix in the generated eigen pc file, so projects
which depend on this pc file can now correctly find the eigen include
directory at build time.

Also correct the Cflags output to use the runtime prefix instead of the
build time STAGING_DIR, like we do elsewhere.

[Peter: drop backslashes, tweak commit message]
Signed-off-by: Matt Flax <flatmax@flatmax.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

python-django: security bump to version 1.11.15

Bump to the latest release of the 1.11.x LTS series as 1.10.x is no longer
supported upstream:

https://www.djangoproject.com/download/

Fixes the following security issues:

- CVE-2017-12794: Possible XSS in traceback section of technical 500 debug
  page (1.11.5)

- CVE-2018-6188: Information leakage in AuthenticationForm (1.11.10)

- CVE-2018-7536: Denial-of-service possibility in urlize and urlizetrunc
  template filters (1.11.11)

- CVE-2018-7537: Denial-of-service possibility in truncatechars_html and
  truncatewords_html template filters (1.11.11)

- CVE-2018-14574: Open redirect possibility in CommonMiddleware (1.11.15)

Also add a hash for the license file.

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

package/m4: fix build on host with glibc-2.28

glibc-2.28 did quite some lifting in their headers, which breaks the
way some packages were detecting glibc, like gnulib.

However, packages do bundle gnulib (it was meant to be bundled),
and so does m4.

Since m4 hasn't seen the slightest commit since 2017-01-09, it is
bundling an old gnulib version, that predates glibc-2.28, and thus
breaks. It also means that upstream hasn't already fixed the issue.

Furthermore, as upstream is using a git submodule for gnulib, and
that the paths are not the same in the release tarball (in lib/)
and in the git tree (in gnulib/), we can't do a plain backport.

So, we selectively backport the two patches from gnulib upstream,
restricted to only the files that happen to be used in m4.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Reported-by: c32 on IRC
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

dahdi-tools: unify dependencies comments

The threads dependency comment is currently shown even though the
toolchain supports threads, only because kernel build is disabled.

Merge the kernel and threads comments. This is similar to what we have
in other packages that need the kernel.

Cc: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

dbus: bump to version 1.2.10

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

cryptsetup: fix build with pre C11 toolchains

Add a patch removing a redefined typedef. Redefinition of typedef, even
to the same type, used to be illegal before C11.

Fixes:
http://autobuild.buildroot.net/results/93b/93b300a62f2ddbad66eab08e25fc3225969f224b/
http://autobuild.buildroot.net/results/6c6/6c6092d633400498ee5b8090733a949a9632e473/
http://autobuild.buildroot.net/results/f46/f46ef6123b5fa92753ff534b4ef7bea3f53ac388/

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

libfuse: security bump to version 2.9.8

Fixes CVE-2018-10906 - In fuse before versions 2.9.8 and 3.x before 3.2.5,
fusermount is vulnerable to a restriction bypass when SELinux is active.
This allows non-root users to mount a FUSE file system with the
'allow_other' mount option regardless of whether 'user_allow_other' is set
in the fuse configuration.  An attacker may use this flaw to mount a FUSE
file system, accessible by other users, and trick them into accessing files
on that file system, possibly causing Denial of Service or other unspecified
effects.

And additionally:

- libfuse no longer segfaults when fuse_interrupted() is called outside the
  event loop.

- The fusermount binary has been hardened in several ways to reduce
  potential attack surface.  Most importantly, mountpoints and mount options
  must now match a hard-coded whitelist.  It is expected that this whitelist
  covers all regular use-cases.

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

orangepi_zero_defconfig: bump linux to 4.17.15 to include latest fixes

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

orangepi_zero_defconfig: bump u-boot to 2018.07 to fix build issue

u-boot 2018.01 now fails to build with the following error:

  CC      arch/arm/lib/asm-offsets.s
In file included from /builds/buildroot.org/buildroot/output/host/include/libfdt.h:54:0,
                 from /builds/buildroot.org/buildroot/output/build/uboot-2018.01/scripts/dtc/libfdt/fdt.c:54:
/builds/buildroot.org/buildroot/output/host/include/libfdt_env.h:82:24: error: redefinition of 'fdt16_to_cpu'
 static inline uint16_t fdt16_to_cpu(fdt16_t x)
                        ^~~~~~~~~~~~
In file included from /builds/buildroot.org/buildroot/output/build/uboot-2018.01/scripts/dtc/libfdt/fdt.c:51:0:
/builds/buildroot.org/buildroot/output/build/uboot-2018.01/scripts/dtc/libfdt/libfdt_env.h:81:24: note: previous definition of 'fdt16_to_cpu' was here
 static inline uint16_t fdt16_to_cpu(fdt16_t x)

https://gitlab.com/buildroot.org/buildroot/-/jobs/88314891

Fix it by bumping the u-boot version to 2018.07.

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

orangepi_pc_defconfig: bump linux to 4.17.15 to include latest fixes

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

orangepi_pc_defconfig: bump u-boot to 2018.07 to fix build issue

u-boot 2018.05 now fails to build with the following error:

  HOSTCC  scripts/dtc/flattree.o
In file included from /builds/buildroot.org/buildroot/output/host/include/libfdt.h:54:0,
                 from /builds/buildroot.org/buildroot/output/build/uboot-2018.05/scripts/dtc/libfdt/fdt.c:54:
/builds/buildroot.org/buildroot/output/host/include/libfdt_env.h:82:24: error: redefinition of 'fdt16_to_cpu'
 static inline uint16_t fdt16_to_cpu(fdt16_t x)
                        ^~~~~~~~~~~~
In file included from /builds/buildroot.org/buildroot/output/build/uboot-2018.05/scripts/dtc/libfdt/fdt.c:51:0:
/builds/buildroot.org/buildroot/output/build/uboot-2018.05/scripts/dtc/libfdt/libfdt_env.h:81:24: note: previous definition of 'fdt16_to_cpu' was here
 static inline uint16_t fdt16_to_cpu(fdt16_t x)
                        ^~~~~~~~~~~~

https://gitlab.com/buildroot.org/buildroot/-/jobs/88314886

Fix it by bumping the u-boot version to 2018.07.

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

linux: replace shell for loop with make foreach loop

This commit replaces the loop copying out-of-tree DTS into the kernel
tree by a make foreach loop instead of a shell for loop. This allows
to error out if one of the DTS file cannot be copied (for example if
it doesn't exist).

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

sheevaplug: add a basic readme

Describing how to update the board.

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

sheevaplug_defconfig: bump linux to 4.14.63 to include latest fixes

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

sheevaplug_defconfig: bump u-boot to 2018.07 to fix build issue

u-boot 2016.05 no longer builds with the default gcc/binutils versions:

https://gitlab.com/buildroot.org/buildroot/-/jobs/88314946

  LD      u-boot
fs/built-in.o: In function `read_symbol':
/builds/buildroot.org/buildroot/output/build/uboot-2016.05/fs/jffs2/mini_inflate.c:129: undefined reference to `pull_bit'
fs/built-in.o: In function `decompress_huffman':
/builds/buildroot.org/buildroot/output/build/uboot-2016.05/fs/jffs2/mini_inflate.c:159: undefined reference to `pull_bits'
/builds/buildroot.org/buildroot/output/build/uboot-2016.05/fs/jffs2/mini_inflate.c:170: undefined reference to `pull_bits'
..
/builds/buildroot.org/buildroot/output/host/bin/arm-buildroot-linux-uclibcgnueabi-ld.bfd: BFD (GNU Binutils) 2.29.1 assertion fail elf32-arm.c:9509
Makefile:1192: recipe for target 'u-boot' failed
make[1]: *** [u-boot] Error 1
make[1]: Leaving directory '/builds/buildroot.org/buildroot/output/build/uboot-2016.05'

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

package/libopenssl: security bump to version 1.0.2p

Fixes CVE-2018-0732 & CVE-2018-0737:
https://www.openssl.org/news/vulnerabilities.html

Added upstream sha1 hash.

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

linux: bump default to version 4.17.15

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

linux-headers: bump 4.{4, 9, 14, 17}.x series

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

boost: put back chrono select for coroutine

coroutine module does use chrono directly:

./libs/coroutine/performance/asymmetric/segmented/Jamfile.v2: <library>/boost/chrono//boost_chrono
./libs/coroutine/performance/asymmetric/Jamfile.v2: <library>/boost/chrono//boost_chrono
./libs/coroutine/performance/symmetric/segmented/Jamfile.v2: <library>/boost/chrono//boost_chrono
./libs/coroutine/performance/symmetric/Jamfile.v2: <library>/boost/chrono//boost_chrono

So put back select of chrono for coroutine

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

libselinux: add patch to fix build with gcc < 4.7

This commit adds a patch from Hollis Blanchard on libselinux to fix
build on host machines that have gcc < 4.7.

Fixes:

  http://autobuild.buildroot.net/results/a82bb0c0b22ff24263ad7a7d165b21c0df7b3b1d/

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

bzip2: change upstream site

The bzip.org website is down. Use the Buildroot backup download site.
Remove the website link as there is no clear alternative upstream at
this point.

  https://lwn.net/Articles/762264/

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

boost: thread needs chrono

chrono is selected by thread (see libs/thread/build/Jamfile.v2):

rule usage-requirements ( properties * )
{
[...]
result += <library>/boost/chrono//boost_chrono ;
}

So add this select for BR2_PACKAGE_BOOST_THREAD and remove it from
BR2_PACKAGE_BOOST_COROUTINE, BR2_PACKAGE_BOOST_LOG,
BR2_PACKAGE_TYPE_ERASURE and BR2_PACKAGE_BOOST_WAVE

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

boost: locale needs thread with icu

output/host/usr/bin/nios2-linux-readelf -d output/staging/usr/lib/libboost_locale.so

Dynamic section at offset 0x125ec0 contains 36 entries:
  Tag        Type                         Name/Value
 0x00000001 (NEEDED)                     Shared library: [libboost_chrono.so.1.67.0]
 0x00000001 (NEEDED)                     Shared library: [libboost_thread.so.1.67.0]
 0x00000001 (NEEDED)                     Shared library: [libboost_system.so.1.67.0]
 0x00000001 (NEEDED)                     Shared library: [librt.so.1]
 0x00000001 (NEEDED)                     Shared library: [libicudata.so.60]
 0x00000001 (NEEDED)                     Shared library: [libicui18n.so.60]
 0x00000001 (NEEDED)                     Shared library: [libicuuc.so.60]

Fixes:
 - http://autobuild.buildroot.net/results/57838f7cd84f37b66ab7007deaea847af8f54b72

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

domoticz: fix appversion.default

Retrieve correct appversion.default (9700) from upstream

Without this patch, the wrong version is displayed in the web ui and
when the user checks for an update, domoticz wrongly says that a new
version is available
This issue was reported by an email from Eyal Eshed <eeshed@coldroll.ca>

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

package/samba4: security bump to version 4.8.4

Release notes: https://www.samba.org/samba/history/samba-4.8.4.html

Fixes

o  CVE-2018-1139  (Weak authentication protocol allowed.)
o  CVE-2018-1140  (Denial of Service Attack on DNS and LDAP server.)
o  CVE-2018-10858 (Insufficient input validation on client directory
                   listing in libsmbclient.)
o  CVE-2018-10918 (Denial of Service Attack on AD DC DRSUAPI server.)
o  CVE-2018-10919 (Confidential attribute disclosure from the AD LDAP
                   server.)

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

package/pkg-golang: post-pone evaluation of variables

As explained in pkg-generic.mk, all variable references inside the
inner-xxx-package should use $$(...). Otherwise, they are evaluated
too early, and will not contain the expected value. In the content of
the pkg-golang infrastructure, the <pkg>_SRC_DOMAIN, <pkg>_SRC_VENDOR
and <pkg>_SRC_SOFTWARE variables were not properly escaping their
reference to the $$($(2)_SITE) variable.

This was not visible until now, as only target Go packages were
supported, where $(2)_SITE was always defined prior to this macro
being expanded. With the upcoming support of host Go packages, we need
to fix this, as $(2)_SITE may be defined later, inherited from
$(3)_SITE.

Signed-off-by: Mirza Krak <mirza.krak@northern.tech>
[Thomas: rework commit log.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

qt5serialbus: examples need qt5 widgets module

Fixes [1]:

  make[2]: Entering directory '.../build/qt5serialbus-5.11.1/examples'
  Some of the required modules (qtHaveModule(widgets)) are not available.
  Skipped.
  [...]
  cp -dpfr .../host/mips64el-buildroot-linux-gnu/sysroot/usr/lib/qt/examples/serialbus .../target/usr/lib/qt/examples/
  cp: cannot stat '.../host/mips64el-buildroot-linux-gnu/sysroot/usr/lib/qt/examples/serialbus': No such file or directory

[1] http://autobuild.buildroot.net/results/147809b5f8758af935bee48c0fc83fd86a8509e9

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

package/lvm2: Fix runtime crash when using uclibc

When using uclibc libdevmapper.so was calling dm_task_get_info_base()
function recursively, leading to segmentation fault. This was
happening because uclibc linker loader just takes first existing
'dm_task_get_info' (which is 'dm_task_get_info_base') symbol in elf
binary, instead of default version.

Add upstreamable lvm2 patch [1], which introduces
--enable-symvers[=STYLE] switch. Use that switch to disable symbol
versions, as we do not plan to support binaries compiled against
old libdevmapper library.

Fixes bug #10781.

[1] https://www.redhat.com/archives/dm-devel/2018-July/msg00187.html

Signed-off-by: Marcin Niestroj <m.niestroj@grinn-global.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

lighttpd: Add systemd tmpfile to fix lighttpd systemd boot

Problem starting lighttpd application with systemd.

/usr/sbin/lighttpd -D -f /etc/lighttpd/lighttpd.conf
2018-06-22 11:21:34: (server.c.733) opening errorlog '/var/log/lighttpd-error.log' failed: Permission denied
2018-06-22 11:21:34: (server.c.1420) Opening errorlog failed. Going down.

Lighttpd can not write the 'lighttpd-access.log' and 'lighttpd-error.log' files
to the directory '/var/log/'.

When using systemd the directory '/var/log' does not allow the user www-data to
write.

To correct the problem, we add /usr/lib/tmpfiles.d/lighttpd.conf.
This file create the 'lighttpd-access.log' and 'lighttpd-error.log' files  with
the permission

Signed-off-by: Laurent Hartanerot <laurent.hartanerot@atos.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

package/aircrack-ng: powerpc arch requires altivec

The powerpc support in this package currently requires an arch with
altivec accelerator support. This is a bug and the the powerpc arch
should have a non-simd option like ARM/x86.

Upstream issue: https://github.com/aircrack-ng/aircrack-ng/issues/1941

Fixes
http://autobuild.buildroot.net/results/87e82a5e8d0b1c1ff10ec3e59d25bcd56b329075

Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

qt5base: fix double-conversion compile for aarch64_be

Fixes [1]:

  ../3rdparty/double-conversion/include/double-conversion/utils.h:81:2: error: #error Target architecture was not detected as supported by Double-Conversion.
  #error Target architecture was not detected as supported by Double-Conversion.

[1] http://autobuild.buildroot.net/results/1fe2be0f26e5b92db57a5cfb5646dd253b731a5c

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

configs/arm_juno: bump ATF to v1.3

ATF in version 1.2 fails to build with:

./build/juno/release/bl1/context_mgmt.o: In function `cm_prepare_el3_exit':
context_mgmt.c:(.text.cm_prepare_el3_exit+0x54): undefined reference to `cm_set_next_context'
context_mgmt.c:(.text.cm_prepare_el3_exit+0x54): relocation truncated to fit: R_AARCH64_JUMP26 against undefined symbol `cm_set_next_context'

This has been fixed in ATF v1.3. Even though there are even newer
versions of ATF available, we take a conservative approach, and bump
to the first version that has the build issue fixed.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/-/jobs/88314771

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

board/technologic/ts7680: fix bogus size in genimage.cfg

512B is not a correct size to express "512 bytes", and causes a
genimage failure:

  ERROR: Invalid size suffix 'B' in '512B'

To express "512 bytes", using just "512" is sufficient. With this
commit, genimage works fine, and we indeed have a 512 bytes unused
partition:

$ fdisk -l output/images/sdcard.img
Disk output/images/sdcard.img: 60 MiB, 62915584 bytes, 122882 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0x00000000

Device                    Boot Start    End Sectors  Size Id Type
output/images/sdcard.img1          1      1       1  512B  0 Empty
output/images/sdcard.img2          2 524289  524288  256M 83 Linux

Fixes:

  https://gitlab.com/buildroot.org/buildroot/-/jobs/88314963

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

configs: add missing dependencies of Linux on host-openssl

Four defconfigs have Linux kernel configuration that need OpenSSL on
the host, but forgot to express this dependency, causing build
failures. This commit adjusts those defconfigs, fixing the following
defconfig builds:

   imx6ulpico_defconfig
   https://gitlab.com/buildroot.org/buildroot/-/jobs/88314839

   imx7dpico_defconfig
   https://gitlab.com/buildroot.org/buildroot/-/jobs/88314841

   mx51evk_defconfig
   https://gitlab.com/buildroot.org/buildroot/-/jobs/88314847

   orangepi_lite_defconfig
   https://gitlab.com/buildroot.org/buildroot/-/jobs/88314883

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

configs/snps_archs38_vdk: use post-build script instead of overlay

In commit 2cdfa6c84951b25e805d329cb7bae833d333538f ("synopsys/axs10x:
Update /etc/inittab by post-build"), two Synopsys platforms were
changed to use a post-build script adding an extra getty in the
inittab instead of a rootfs overlay containing a custom
inittab. However, in this commit, configs/snps_archs38_vdk_defconfig
was not changed, even though it was using the same rootfs overlay.

This commit therefore adjusts configs/snps_archs38_vdk_defconfig to
also use the newly introduced post-build script, fixing the build of
configs/snps_archs38_vdk_defconfig.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/-/jobs/88314952

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

configs/raspberrypi2: increase ext4 filesystem size

The default size of the ext4 filesystem is no longer sufficient to
hold all the kernel modules built by the RPi Linux kernel
configuration. Let's increase to 120 MB, like we did for
raspberrypi3_defconfig.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/-/jobs/88314938

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

ncmpc: fix build with some old toolchains

With some "old" toolchains (glibc, uclibc in version 4.9.4, 5.3, 5.4,
5.5 ...), the following error is raised by the compiler:

../src/screen.cxx:60:29:   required from here
/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/ext/new_allocator.h:120:4:
error: no matching function for call to 'std::pair<const screen_functions* const, std::unique_ptr<Page> >::pair(const screen_functions*, Page*)'

[...]

/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/bits/stl_pair.h:112:26:
note: candidate: constexpr std::pair<_T1, _T2>::pair(const _T1&, const _T2&) [with _T1 = const screen_functions* const; _T2 = std::unique_ptr<Page>]
       _GLIBCXX_CONSTEXPR pair(const _T1& __a, const _T2& __b)
                          ^
/usr/lfs/v0/rc-buildroot-test/scripts/instance-1/output/host/opt/ext-toolchain/mips-linux-gnu/include/c++/5.3.0/bits/stl_pair.h:112:26:
note:   no known conversion for argument 2 from 'Page*' to 'const std::unique_ptr<Page>&'

This is due to the fact that init function of screen_functions
structure returns Page* but PageMap wants a std::unique_ptr<Page>

To fix this, cast raw pointer into a unique_ptr with an explicit cast

Fixes:
 - http://autobuild.buildroot.net/results/d8a7339d8bdd5cdc6bd1716585d4bcf15a2e8015

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

mesa3d-headers: fix logic to generate the dri.pc file

As noted by Arnout in [1], the logic in mesa3d-headers.mk generates a
bogus dri.pc file, which looks like this:

prefix=/usr
exec_prefix=/usr
libdir=/lib
includedir=/include
dridriverdir=/dri

Indeed, the ${...} are expanded as shell variables when the sed
command is executed, while the intention is that those ${...} should
go in the .pc file. By escaping those using $${...}, we get the
expected .pc file:

prefix=/usr
exec_prefix=/usr
libdir=${exec_prefix}/lib
includedir=${prefix}/include
dridriverdir=${libdir}/dri

This was detected by the not yet committed check-package improvement
from Ricardo that detects bogus ${...} usage to reference make
variables.

[1] http://lists.busybox.net/pipermail/buildroot/2018-July/225402.html

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

package/vlc: needs speexdsp for optional speex support

Fixes configure warning:

checking for SPEEX... yes
checking for SPEEXDSP... no
configure: WARNING: Package speexdsp was not found in the pkg-config search path.
Perhaps you should add the directory containing `speexdsp.pc'
to the PKG_CONFIG_PATH environment variable
Package 'speexdsp', required by 'world', not found.

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

package/vlc: Remove warning from Config.in

Four years ago this warning was added to this package
https://git.buildroot.net/buildroot/commit/package/vlc?id=e33019aaa0409e35b725dddffe09e3e2a42959d1
due to "autobuilder failures for vlc on many platforms".

The last autobuilder failure occured May 1st, 2018 and was fixed by
https://git.buildroot.net/buildroot/commit/package/vlc?id=9f115bfc2ac798c737d0c2a8ec8c96670e814a49

Therefore this warning is not needed anymore.

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

mtd: backport upstream bug fix

A regression in the mtd package prevents non-consecutive ubi volume
IDs from working properly. This change backported from upstream
reverts the commit [0]. The issue was originally described on the mtd
mailing list:

http://lists.infradead.org/pipermail/linux-mtd/2018-June/081562.html

[0] http://git.infradead.org/mtd-utils.git/commitdiff/dede98ffb706676309488d7cc660f569548d5930

Signed-off-by: David Owens <david.owens@rockwellcollins.com>
Reviewed-by: Matt Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

busybox: bump version to 1.29.2

Includes fixes for fdisk (compat fixes, allow 2TB+ sizes), gzip
(FEATURE_GZIP_LEVELS was producing badly-compressed .gz), hexedit (segfault
fix).

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

uboot-tools: force host CFLAGS

U-Boot now adds -std=gnu11 when building its build utilities, like
fixdep. This option is only supported since gcc version 4.7.

  https://gcc.gnu.org/gcc-4.7/changes.html

Force usage of Buildroot HOST_CFLAGS, like we do already for
host-uboot-tools, to avoid the -std=gnu11 option.

Fixes:
http://autobuild.buildroot.net/results/ea0/ea09b614a3c10d494939d9551c0c3bfca4626ece/
http://autobuild.buildroot.net/results/f8e/f8efd9af4d86c51fe4f5afe44db9abf9adc5718f/
http://autobuild.buildroot.net/results/c0a/c0ab35c3cb46f84863fe20c0ee92ecc2379578b6/

Cc: Matt Weber <matthew.weber@rockwellcollins.com>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

support/testing: add test for file capabilities

Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

makedevs: allow leading whitespace for capabilities

Currently makedevs silently ignores extended attributes with leading
whitespace, for example those added to a <PACKAGE>_PERMISSIONS following
the recommended style from check-package.

Makedevs already ignores leading whitespace for normal entries (file
permission changes and device files creation). Do the same for extended
attributes.

Fixes: #11191.
Reported-by: Jean-pierre Cartal <jpcartal@free.fr>
Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

ncurses: disable terminfo database install for host-ncurses

Since commit b35ad5d0b45e (ncurses: make host-ncurses use host terminfo), we
are now pointing host-ncurses to the host terminfo (typically) located in
/usr/share/terminfo.

With this change we are reusing the existing host terminfo database, so
there is no point in trying to install our own on top.  The user running
buildroot typically will have no write access to /usr/share/terminfo, but
tic in that case falls back to writing the database to $HOME/.terminfo.
Neither of which are desirable.

In case $HOME/.terminfo also isn't writable, tic fails, breaking the install
step for host-ncurses:

** Building terminfo database, please wait...
Running sh ./shlib tic to install /usr/share/terminfo ...

        You may see messages regarding extended capabilities, e.g., AX.
        These are extended terminal capabilities which are compiled
        using
                tic -x
        If you have ncurses 4.2 applications, you should read the INSTALL
        document, and install the terminfo without the -x option.

"terminfo.tmp", line 21272, terminal 'v3220': /home/peko/.terminfo: permission denied (errno 30)

To fix all of this, simply disable the terminfo database install.

Suggested-by: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Acked-by: Hollis Blanchard <hollis_blanchard@mentor.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>

chrony: backport upstream patch to fix blocking on getrandom() at startup with recent kernels

chrony calls getrandom() at startup if available, so it needs a workaround
for the blocking behaviour on recent (4.14.39+), similar to what was done
for util-linux in commit c4d86707cd6 (util-linux: add two upstream patches
to fix blocking on getrandom() with recent kernels).

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

package/waylandpp: fix build error with gcc 4.4/4.9

Fixes
http://autobuild.buildroot.net/results/51a/51afcd6d3c3c4839d8216b7b0dba35c8e38211f4/

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

linux: bump default to version 4.17.14

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, 17}.x series

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

package/wpa_supplicant: add upstream security fix

Fixes CVE-2018-14526:
http://w1.fi/security/2018-1/unauthenticated-eapol-key-decryption.txt

Added license hash.

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