support/testing: simplify logging by keeping the log file open

We currently call infra.smart_open() to open log files each time we
need to write to them.

Opening the file once in the constructor of Builder and Emulator and
writing to it whenever needed is simpler and slightly more efficient.

Remove smart_open and instead create a new open_log_file() function
which just opens the logfile. Also let it compute the filename, in
order to simplify even further the Builder and Emulator code.

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/testing/run-tests: help: put the one-letter form before the long form

This is what the manpages usually do, and what Python does with the
automatically-added -h/--help parameter:

Before the change:
  $ ./support/testing/run-tests
    [...]
  optional arguments:
    -h, --help            show this help message and exit
    --list, -l            list of available test cases
    --all, -a             execute all test cases

After the change:
  $ ./support/testing/run-tests
    [...]
  optional arguments:
    -h, --help            show this help message and exit
    -l, --list            list of available test cases
    -a, --all             execute all test cases

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/testing: document get_elf_prog_interpreter

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/testing: rename check_broken_links to has_broken_links

has_broken_links makes it self-explanatory that this is a predicate
function, and that the return value tells whether there _are_ broken
links, not the opposite.

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

gnutls: bump version to 3.5.12

For details, see the release notes:

https://lists.gnupg.org/pipermail/gnutls-devel/2017-May/008427.html
https://lists.gnupg.org/pipermail/gnutls-devel/2017-April/008393.html

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

fbgrab: explicitly build fbgrab target to workaround gzip issue

Fixes #9871

gzip reads default command line options from the environment variable GZIP.
The fbgrab Makefile internally also uses a GZIP make variable to know what
command to use to compress the manpage. Unfortunaly make will export the
value of this make variable to the environment if GZIP is already present in
the enviroment, confusing gzip (as 'gzip' isn't a valid command line argument).

This can either be triggered by users having GZIP set in their environment
(E.G.  for custom options), or by enabling BR2_REPRODUCIBLE, where we use
this feature to force the -n option (to not store name/timestamp) to gzip.

We don't really need to compress the manpage as it isn't installed anyway,
so work around the issue by only building the fbgrab application.

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

ntp: add patch to fix -fPIC issue

This commit adds a patch that fixes the build on AArch64 and MIPS of the
ntp package, which was caused by some parts of ntp being built without
-fPIC.

Fixes:

  [aarch64] http://autobuild.buildroot.net/results/866b1d28595efd8b6becf83d0a64b596538d58b0
  [mips]    http://autobuild.buildroot.net/results/c2a945855172970736a8ffea9c564f029a023344

Thanks to Romain Naour for the initial analysis.

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

libqmi: move canonicalize_file_name()

Previously was defined in qmi-util.h to fix musl
compatibility.  It was missed that this is a shared
header which causes other dependent package builds
to fail, so the static definition was moved into the
implementation where it was used.

Upstream bug report has been updated and this resolves
the following autobuilder related failures.

 http://autobuild.buildroot.net/results/fe9b6b1b1399a4e8aafc6a326c81ec97c2480025
 http://autobuild.buildroot.net/results/40eef797b4d8d53fc6e10f2048316d63549caf6d

test-pkg testing was performed against 10 random toolchains

++ Config snippet ++
BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_EUDEV=y
BR2_PACKAGE_MODEM_MANAGER_LIBQMI=y
BR2_PACKAGE_MODEM_MANAGER=y
BR2_PACKAGE_LIBQMI=y

++ Tested Toolchains ++ (Verified I hit all libc variants)
br-arm-cortex-a9-musl
br-arm-full-static
br-m68k-68040-full    (uclibc)
br-mips32r6-el-hf-glibc
br-mips64-n64-full
br-mips64r6-el-hf-glibc
br-mipsel-o32-full
br-sparc64-glibc
linaro-arm
sourcery-arm

Signed-off-by: Matthew Weber <matthew.weber@rockwellcollins.com>
Tested-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/pulseview: fix a build issue

Backport an upstream patch fixing the build issue related to boost
math definitions.

Fixes:
http://autobuild.buildroot.net/results/e9f/e9f3f175e203529c44ecf92a34b82a0b3a473e34/

Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

rabbitmq-c: openssl/popt sub-options don't work in static linking

rabbitmq-c currently fails to build in a number of static linking
situations, due to two issues:

 - CMake FindOpenSSL module is buggy. Even though it uses pkg-config,
   it doesn't use the information returned by pkg-config, and
   therefore doesn't know about second order libraries that need be
   part of the link for static linking to succeed. Due to this, -lz is
   not passed, and therefore rabbitmq-c fails when linking against
   libssl/libcrypto. This issue has been reported to upstream CMake at
   https://gitlab.kitware.com/cmake/cmake/issues/16885.

 - popt might use libintl, but CMake doesn't know about that. For
   autotools based packages, we typically work around this by passing
   LIBS=, but CMake apparently has no equivalent to LIBS=.

To workaround this, we only use the OpenSSL and Popt optional
dependencies in dynamic linking situations.

Fixes:

  http://autobuild.buildroot.net/results/798dbe5e5fd0463bb2066cb115656795144c327f/

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

Update for 2017.05-rc2

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

qt5declarative: fix examples compile without OpenGL support

Fixes [1]:

   main.cpp:(.text._ZN11QQmlPrivate10createIntoI6FbItemEEvPv[_ZN11QQmlPrivate10createIntoI6FbItemEEvPv]+0x18): undefined reference to `QQuickFramebufferObject::QQuickFramebufferObject(QQuickItem*)'
   .obj/main.o: In function `QQmlPrivate::QQmlElement<FbItem>::~QQmlElement()':
   main.cpp:(.text._ZN11QQmlPrivate11QQmlElementI6FbItemED2Ev[_ZN11QQmlPrivate11QQmlElementI6FbItemED5Ev]+0x5c): undefined reference to `vtable for QQuickFramebufferObject'
   .obj/main.o: In function `QQmlPrivate::QQmlElement<FbItem>::~QQmlElement()':
   main.cpp:(.text._ZN11QQmlPrivate11QQmlElementI6FbItemED0Ev[_ZN11QQmlPrivate11QQmlElementI6FbItemED0Ev]+0x64): undefined reference to `vtable for QQuickFramebufferObject'
   .obj/main.o:(.data.rel.ro._ZTVN11QQmlPrivate11QQmlElementI6FbItemEE[_ZTVN11QQmlPrivate11QQmlElementI6FbItemEE]+0x48): undefined reference to `QQuickFramebufferObject::isTextureProvider() const'
   .obj/main.o:(.data.rel.ro._ZTVN11QQmlPrivate11QQmlElementI6FbItemEE[_ZTVN11QQmlPrivate11QQmlElementI6FbItemEE]+0x4c): undefined reference to `QQuickFramebufferObject::textureProvider() const'
   .obj/main.o:(.data.rel.ro._ZTVN11QQmlPrivate11QQmlElementI6FbItemEE[_ZTVN11QQmlPrivate11QQmlElementI6FbItemEE]+0xb4): undefined reference to `QQuickFramebufferObject::geometryChanged(QRectF const&, QRectF const&)'
   .obj/main.o:(.data.rel.ro._ZTVN11QQmlPrivate11QQmlElementI6FbItemEE[_ZTVN11QQmlPrivate11QQmlElementI6FbItemEE]+0xb8): undefined reference to `QQuickFramebufferObject::updatePaintNode(QSGNode*, QQuickItem::UpdatePaintNodeData*)'
   .obj/main.o:(.data.rel.ro._ZTVN11QQmlPrivate11QQmlElementI6FbItemEE[_ZTVN11QQmlPrivate11QQmlElementI6FbItemEE]+0xbc): undefined reference to `QQuickFramebufferObject::releaseResources()'
   .obj/moc_fbitem.o: In function `FbItem::qt_metacast(char const*)':
   moc_fbitem.cpp:(.text+0x70): undefined reference to `QQuickFramebufferObject::qt_metacast(char const*)'
   .obj/moc_fbitem.o: In function `FbItem::qt_metacall(QMetaObject::Call, int, void**)':
   moc_fbitem.cpp:(.text+0x80): undefined reference to `QQuickFramebufferObject::qt_metacall(QMetaObject::Call, int, void**)'
   .obj/moc_fbitem.o: In function `FbItem::~FbItem()':
   moc_fbitem.cpp:(.text._ZN6FbItemD2Ev[_ZN6FbItemD5Ev]+0x38): undefined reference to `vtable for QQuickFramebufferObject'
   .obj/moc_fbitem.o: In function `FbItem::~FbItem()':
   moc_fbitem.cpp:(.text._ZN6FbItemD0Ev[_ZN6FbItemD0Ev]+0x40): undefined reference to `vtable for QQuickFramebufferObject'
   .obj/moc_fbitem.o:(.data.rel.ro+0x8): undefined reference to `typeinfo for QQuickFramebufferObject'
   .obj/moc_fbitem.o:(.data.rel.ro+0x58): undefined reference to `QQuickFramebufferObject::isTextureProvider() const'
   .obj/moc_fbitem.o:(.data.rel.ro+0x5c): undefined reference to `QQuickFramebufferObject::textureProvider() const'
   .obj/moc_fbitem.o:(.data.rel.ro+0xc4): undefined reference to `QQuickFramebufferObject::geometryChanged(QRectF const&, QRectF const&)'
   .obj/moc_fbitem.o:(.data.rel.ro+0xc8): undefined reference to `QQuickFramebufferObject::updatePaintNode(QSGNode*, QQuickItem::UpdatePaintNodeData*)'
   .obj/moc_fbitem.o:(.data.rel.ro+0xcc): undefined reference to `QQuickFramebufferObject::releaseResources()'
   .obj/moc_fbitem.o:(.data.rel.ro+0xf0): undefined reference to `QQuickFramebufferObject::staticMetaObject'

[1] http://autobuild.buildroot.net/results/64a/64a198397736db12b73c1f693dbe1c47d73b53da

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

protobuf: link with libatomic when needed

The protobuf library uses atomic intrinsics, so we need to link
against libatomic.

Fixes the build of protobuf on Sparc:

  http://autobuild.buildroot.net/results/f3d76eaebd529a61bce849e355182c60f233ed06/

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

gcc: drop unused legacy arc-2016.09 patch

This was forgotten when the arc toolchain version was bumped.

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

package/boost: disable boost-locale for static only build with icu

Boost fails to build with the following error:

error: Tried to build the target twice, with property sets having
error: these incompatible properties:
error:
error:     -  <runtime-link>static <warnings>all
error:     -  <runtime-link>shared <warnings>on

when the following conditions are met:

 - BR2_STATIC_LIBS=y
 - BR2_PACKAGE_ICU=y
 - BR2_PACKAGE_BOOST_LOCALE=y
 - Another BR2_PACKAGE_BOOST_xyz option is enabled, which enables a
   feature not provided just by header files, but that requires
   building a library.

In such a situation, Boost absolutely wants to build the libboost
libraries as shared libraries. Not having boost-locale, or not having
icu is sufficient to avoid the issue.

So, as a simple work-around, we prevent from building boost-locale
when icu and static linking are used.

Fixes:

  http://autobuild.buildroot.net/results/c8f7aa85f5791d8ae8cf4b9085788adc5152286f/

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Yegor Yefremov <yegorslists@googlemail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
[Thomas:
 - only disable boost-locale when icu is enabled
 - improve commit log]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

fxload: disable on Microblaze

gcc on Microblaze is affected by PR63261, which causes a build failure
when optimization *and* debugging symbols are enabled. Since anyway
fxload is unlikely to be useful on Microblaze, let's disable this
package on this architecture.

Fixes:

  http://autobuild.buildroot.net/results/24640a042d84f45339246c0a18e10905494b2199/

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

rtmpdump: security bump to current HEAD

Fixes:

- CVE-2015-8271: The AMF3CD_AddProp function in amf.c in RTMPDump 2.4 allows
  remote RTMP Media servers to execute arbitrary code
  https://www.talosintelligence.com/reports/TALOS-2016-0067/

- CVE-2015-8272: RTMPDump 2.4 allows remote attackers to trigger a denial of
  service (NULL pointer dereference and process crash).
  https://www.talosintelligence.com/reports/TALOS-2016-0068/

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

git: security bump to version 2.12.3

Fixes CVE-2017-8386 - Git Shell Bypass By Abusing Less

For more details, see:
https://insinuator.net/2017/05/git-shell-bypass-by-abusing-less-cve-2017-8386/
http://www.mail-archive.com/git@vger.kernel.org/msg120982.html

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

rpcbind: add upstream security fix for CVE-2017-8779

CVE-2017-8779: rpcbind through 0.2.4, LIBTIRPC through 1.0.1 and 1.0.2-rc
through 1.0.2-rc3, and NTIRPC through 1.4.3 do not consider the maximum RPC
data size during memory allocation for XDR strings, which allows remote
attackers to cause a denial of service (memory consumption with no
subsequent free) via a crafted UDP packet to port 111, aka rpcbomb.

For more details, see:
https://guidovranken.wordpress.com/2017/05/03/rpcbomb-remote-rpcbind-denial-of-service-patches/

Backport upstream fix to version 0.2.3 and unconditionally include syslog.h
to fix a build issue when RPCBIND_DEBUG is disabled (which it is in
Buildroot).

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

package/x11vnc: fix build issue with --without-ssl

Fixes:
http://autobuild.buildroot.net/results/873/873ed8f2ade1d969abdff15b7b6d63e04819af9a

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

php: bump version to 7.1.5

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

package/faketime: fix build with gcc >= 6

Upstream added -Wno-nonnull-compare to disable this error [1].

Fixes:
libfaketime.c:513:7: error : nonnull argument « buf » compared to NULL [-Werror=nonnull-compare]
    if (buf != NULL)
       ^
[1] https://github.com/wolfcw/libfaketime/commit/47e958b753fc15098a2b7d0e9ef26b83ee255874

[Peter: add upstream git commit id to patch]
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

ffmpeg: bump version to 3.3.1

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

gcc: arc-2017.03-rc2: Fix "unrecognized supposed constant" error

This fixes the following problem:
------------------------------------>8--------------------------------
arc-linux-gcc -c -Os -fPIC iso9660.i
iso9660.c: In function 'strip_trail':
iso9660.c:155:1: error: unrecognized supposed constant
 }
 ^
(unspec:SI [
        (symbol_ref:SI ("*.LANCHOR1") [flags 0x182])
    ] ARC_UNSPEC_GOTOFFPC)
iso9660.c:155:1: internal compiler error: in arc_legitimate_constant_p, at config/arc/arc.c:6028
------------------------------------>8--------------------------------

Found by Buildroot autobuilder [1].

The fix [2] is in arc-2017.03 development branch of ARC GCC and once it
becomes a part the next release of ARC tools this should be removed
from Buildroot.

[1] http://autobuild.buildroot.net/results/c90/c909e8c397ab972b6aa4d370572cad4fae284d00/build-end.log
[2] https://github.com/foss-for-synopsys-dwc-arc-processors/gcc/commit/139fed9d29ab935b3bc5159c0bdf7b8b8a39442d

Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

linux: bump default version to 4.11.1

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

linux-headers: bump 4.{4, 9, 10, 11}.x series

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

libv4l: add patch to link qv4l2 tool with librt

The qv4l2 tool in libv4l uses clock_gettime(), so it should link
against librt to build properly with old versions of glibc. Therefore,
we add a patch to libv4l to fix this issue. Autoreconfiguring libv4l
is now necessary since the patch touches Makefile.am.

Fixes:

  http://autobuild.buildroot.net/results/5dff0ec19205e02f6ee373d34cb79f39ac25b609/

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

package/bluez_utils: select BR2_PACKAGE_CHECK

Commit [1] added check package to bluez_utils dependencies without
selecting it at Kconfig level.

Fixes:
Makefile:535: *** check is in the dependency chain of bluez_utils that has added it to its _DEPENDENCIES variable without selecting it or depending on it from Config.in.

[1] 8554be9f82301282161cf6310bacdb6067fefb16

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

package/opencv3: disable explicitly lapack support

lapack support is enabled by default but the opencv3.mk doesn't handle
this dependency yet.

Fixes:
http://autobuild.buildroot.net/results/c8a/c8a25af38c405594722b44aef34e5e4c549639db

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

package/bluez_utils: fix test build issues with musl

Add one missing header and avoid encrypt redefinition.

Fixes:
http://autobuild.buildroot.net/results/06c/06c930d9c5299b79500d018ac3fb2861ce834c7c/

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

package/bluez_utils: add missing check dependency for test program

Tests are enabled since this has been introduced [1], so keep
them for now.

bluez_utils needs check tool and check if it's available [2].

[1] 06c3e2159686a886e52b2522a47e60c300cfb7f7
[2] https://git.kernel.org/pub/scm/bluetooth/bluez.git/tree/configure.ac?h=4.101#n51

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

package/bluez_utils: add missing host-pkgconf dependency

The configure script use pkg-config.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Yegor Yefremov <yegorslists@googlemail.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/binutils: finalize the bump ARC tools to arc-2017.03-rc2

The commit bumping ARC tools to arc-2017.03-rc2 [1] forgot to
update the ARC tools version in binutils.mk

[1] 043737282010f83e8ec20618a034131bc7e8cae1

Fixes:
http://autobuild.buildroot.net/results/f9c/f9c40610209fc22ac8c0db6bd57bd3b11bbe6d9c

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

.gitlab-ci.yml: regenerate after snps_archs38_zebu_defconfig rename

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

openblas: SPARC support is for Sparc v9

Makefile.sparc in the openblas source code passes -mcpu=v9, i.e it
assumes a SPARCv9 CPU. Therefore, this commit adjusts openblas
Config.in file so that SPARC is only used for BR2_sparc_v9.

Fixes:

  http://autobuild.buildroot.net/results/805087e87b8bb7d11adb49d9eca3959a37aca3a2/

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

libcdio: disable on ARC, triggers a toolchain bug

libcdio doesn't build on ARC, with the following compiler error:

iso9660.c:155:1: error: unrecognized supposed constant
[...]
iso9660.c:155:1: internal compiler error: in arc_legitimate_constant_p, at config/arc/arc.c:6028
Please submit a full bug report,

Since this has been happening for a while and is quite noisy in the
autobuilders results, we disable this package on ARC.

Fixes:

  http://autobuild.buildroot.net/results/e837737abaea851bea428a35ea9d3395a6565c54/

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

package/mke2img: check nb_blocks is specified

Since we do not have autocalculation anymore, the user must specify the
exact number of blocks.

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

radvd: fix build with 4.11 kernel headers

Since the introduction of 4.11 kernel headers in Buildroot, radvd has
failed to build with all toolchains using this kernel header version.

The issue comes from the fact that radvd includes both <net/if_arp.h>
and <linux/if_arp.h> if they are available. Until 4.11,
<linux/if_arp.h> was in fact not included, because the
AC_CHECK_HEADERS() test concluded this header was unsuitable. This has
been fixed in the upstream kernel by commit
2618be7dccf8739b89e1906b64bd8d551af351e6 ("uapi: fix linux/if.h
userspace compilation errors"). So now, the radvd configure script
considers both <net/if_arp.h> and <linux/if_arp.h> as suitable
headers, and includes both of them, leading to the duplicate
definition of various types.

Since it's redundant to include both <net/if_arp.h> and
<linux/if_arp.h>, we simply force radvd to believe that
<linux/if_arp.h> is not available by passing the appropriate autoconf
cache variable. This gets us back to the previous situation, where
<linux/if_arp.h> was never used.

This has been tested with a uClibc toolchain using 4.11 kernel
headers, and verified to work on glibc and musl with older kernel
headers as well.

Fixes:

  http://autobuild.buildroot.net/results/949a75d96299394e4ac957746fa23a4b52f31b43/

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

core/download: fix source check for git method

Running foo-source-check on packages retrieved from git fails.

This is because there is no associated stamp file, so we do not have a
rule-assignment that sets PKG for foo-source-check.

But it does not make sense to have a stamp file at all, because
source-check is not supposed to change anything: the status after is
exactly the same as before; nothing is downlaoded, so there is no
progress (whatsoever) to memorise.

Fix that by just defining PKG in the source-check rule definition.

Fixes #9796.

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

package/odroid-scripts: relax dependencies to include armeb

odroid-scripts are just scripts, so they should also work on a armeb
board.

Besides, odroid-mali selects it, but it only depends on arm-eabihf,
which means armeb is not ruled out.

Fixes unmet dependencies (with BR2_armeb=y):
    warning: (BR2_PACKAGE_ODROID_MALI) selects BR2_PACKAGE_ODROID_SCRIPTS
    which has unmet direct dependencies(BR2_aarch64 || BR2_arm)

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

toolchain: Bump ARC tools to arc-2017.03-rc2

This commit bumps ARC toolchain to arc-2017.03-rc2

Please note that it is a release candidate and it might contain some
breakages, please don't use it for production builds.

Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/gpsd: do not interact with systemd on install

When building on a systemd-based host for a systemd-based target, gpsd's
buildsystem will try to tell systemd to reload the its configuration and
recreate the entire dependency tree of services.

Of course, this can not work when run as a user:
    >>> gpsd 3.16 Installing to target
    ==== AUTHENTICATING FOR org.freedesktop.systemd1.reload-daemon ===
    Authentication is required to reload the systemd state.
    Authenticating as: root
    Password:

In fact, this does not even mean anything in cross-compilation.

Fortunately, upstream has already fixed their sin, so we backport it
(tweaked to apply onto 3.16).

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

package/uclibc-ng-test: disable tst-cond22 with glibc based toolchain

uclibc-ng-test is broken with glibc 2.25 due to pthread_cond_t changes
introduced by [1].
Waldemar has disabled this test (tst-cond22) with glibc based toolchain
[2] but the current uclibc-ng-test package doesn't contain this patch.

So here it is.

Fixes:
[nios2] http://autobuild.buildroot.net/results/d61c8d169fc070271114a515159650f7a0f333c0
[arm] http://autobuild.buildroot.net/results/22295711046c519df44d5d976e0bfcae1f006e5a

[1] https://sourceware.org/git/?p=glibc.git;a=commit;h=ed19993b5b0d05d62cc883571519a67dae481a14
[2] https://gogs.waldemar-brodkorb.de/linux/uclibc-ng-test/commit/787e584f18afc967d91cfe1a8443df2657579ecb

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

package/vlc: security bump to version 2.2.5.1

This release contains "various security improvements in demuxers and
decoders".
Release notes: http://www.videolan.org/vlc/releases/2.2.5.html

Added all hashes provided by upstream.

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

package/opencv3: regenerate patch without renames

Patch using renames are not properly handled by patch < 2.7, which
some users (including autobuilder instances) might be using. Such
renames are silently ignored by old patch versions, caused strange
build failures due to the missed renames.

This patch fixes this by regenerating the patch without renames.

Fixes:

  http://autobuild.buildroot.net/results/347d8fd2e286b3e4e5e18743e64d862bdb66dbb9/

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

package/opencv3: fix build for x86_64 target

This change backports a fix from upstream fixing missing symbols due to
wrong compiler optimization flags.

On Buildroot side, also remove deprecated options passed on the
configure command.

Fixes:
  http://autobuild.buildroot.net/results/af950c6172a578c4eeeb9440b6990ceb48263fa6/
  http://autobuild.buildroot.net/results/e8832302cd4f30cec076f67965f6aacbaca5b955/

Cc: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

kvm-unit-tests: fix build on x86-64 with specific gcc versions

kvm-unit-tests uses the following code on x86/x86-64:

static inline u64 scale_delta(u64 delta, u64 mul_frac)
{
        u64 product, unused;

        __asm__ (
                "mul %3"
                : "=d" (product), "=a" (unused) : "1" (delta), "rm" ((u64)mul_frac) );

        return product;
}

The "mul" instruction does not have a suffix that indicates the width of
the data being multiplied. When the data is passed in a register, there
is no need to specify the width, but with some gcc versions, the data is
passed as a memory reference, and therefore the assembler does not know
the width of the data to be multiplied. It causes the following build
failure:

x86/hyperv_clock.c: Assembler messages:
x86/hyperv_clock.c:21: Error: no instruction mnemonic suffix given and no register operands; can't size instruction

Since the data being multiplied is 64 bit, we explicitly specify the
instruction as being "mulq".

Fixes:

  http://autobuild.buildroot.net/results/a4a65d01f049db83a93de92660f228dd18532625/

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

package/espeak: add comment to ignore 'unmet dependencies'

On master 86b4eeccc4, espeak can generate spurious 'unmet dependencies'
messages:

    $ make KCONFIG_SEED=0x7A85DEE0 randconfig
    warning: (BR2_PACKAGE_ESPEAK_AUDIO_BACKEND_ALSA) selects
    BR2_PACKAGE_PORTAUDIO_CXX which has unmet directdependencies
    (BR2_PACKAGE_PORTAUDIO && BR2_INSTALL_LIBSTDCPP)

However, the dpendency chain *is* correct. There is something that
causes the kconfig parser to get really confused...

Add a comment statng the issue is spurious, so noone tries to fix it.

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

lua: use target LDFLAGS

This adds '-static' to the link command when BR2_STATIC_LIBS=y, making the lua
binary really static.

Cc: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Acked-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/clamav: renumber patch

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

package/efl: fix unmet dependencies

EFL's elput selects libinput but forgot to propagate its dependencies.
Which requires they be propagated further to a second sub-option, then
to a third one.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Romain Naour <romain.naour@openwide.fr>
[Thomas: fix typoes.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

strongswan: fix crypt() redefinition

Add patch from upstream pending pull request (#72) to fix crypt() definition
conflict.

Fixes:
http://autobuild.buildroot.net/results/612/6120add5ad6e7f5d575fc2a2442cd7fe5302eb46/
http://autobuild.buildroot.net/results/ba1/ba1298e71ef28857654ae8d4593d09e4fe8cdda0/
http://autobuild.buildroot.net/results/6b9/6b928e2a6a35bfb21c792fca083bc4cbdaef9304/

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

package/linux-headers: bump 3.12.x and 4.{4, 9, 10}.x series

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

package/openvpn: security bump to 2.4.2

Fixes CVE-2017-7478 & CVE-2017-7479:
https://community.openvpn.net/openvpn/wiki/QuarkslabAndCryptographyEngineerAudits

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

qwt: fix build when BR2_STATIC_LIBS=y

By default, qwt build a shared library, but when BR2_STATIC_LIBS is set
compilation failed with errors like :
__uClibc_main.c:(.text+0x164): undefined reference to `__fini_array_end'
__uClibc_main.c:(.text+0x168): undefined reference to `__fini_array_start'
__uClibc_main.c:(.text+0x16c): undefined reference to `__fini_array_start'

This patch disable QwtDll to build a static library when BR2_STATIC_LIBS=y.

fix:
http://autobuild.buildroot.net/results/739/739406bb8073d1861933872a47802954d9767634/

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

postgresql: bump version to 9.6.3 (security)

Three security vulnerabilities have been closed by this release:

CVE-2017-7484: selectivity estimators bypass SELECT privilege checks
CVE-2017-7485: libpq ignores PGREQUIRESSL environment variable
CVE-2017-7486: pg_user_mappings view discloses foreign server passwords

More details: https://www.postgresql.org/about/news/1746/

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

sudo: fix static linking with openldap having openssl support

If we are building sudo statically and openldap was linked with openssl,
then when we link sudo with openldap we need to specify the openssl
libs, otherwise it will fail with "undefined reference" errors.

Fixes:
  http://autobuild.buildroot.net/results/ebb/ebbb4c3138b5023a0c8bd938db1932a25ba5b6fb/
  http://autobuild.buildroot.net/results/58e/58ee4fddea85f0c79b26582b04a573258e27eb47/
  http://autobuild.buildroot.net/results/051/05151c4ef7f0b3702a6cd5e6df4888d826a37431/
  http://autobuild.buildroot.net/results/b1d/b1dfae81985daee106700191fcb82387833c2e3f/
  http://autobuild.buildroot.net/results/5b4/5b42e8505856156389d480d0da19f6982cf120d8/
  http://autobuild.buildroot.net/results/30a/30a76190428f902f03bcf54ba9f1f4f4377c6fc9/
  http://autobuild.buildroot.net/results/e8c/e8c6bdc314b9a8d5de303633a3abd7b7e44beb2e/
  http://autobuild.buildroot.net/results/483/4830c69cc6a62080e1516f0d9009c2ba619c23c1/

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

efibootmgr: backport upstream patch to fix build with gcc 6.x

efibootmgr was built with -fshort-wchar, which causes gcc to use
2-byte wchar_t instead of the normal 4-byte wchar_t. But this doesn't
work anymore with gcc 6.x, as it causes a build failure:

/home/test/autobuild/run/instance-3/output/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-gnueabihf/6.3.0/../../../../arm-buildroot-linux-gnueabihf/bin/ld: warning: efibootmgr.o uses 2-byte wchar_t yet the output is to use 4-byte wchar_t; use of wchar_t values across objects may fail
/home/test/autobuild/run/instance-3/output/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-gnueabihf/6.3.0/../../../../arm-buildroot-linux-gnueabihf/bin/ld: warning: efi.o uses 2-byte wchar_t yet the output is to use 4-byte wchar_t; use of wchar_t values across objects may fail
/home/test/autobuild/run/instance-3/output/host/opt/ext-toolchain/bin/../lib/gcc/arm-buildroot-linux-gnueabihf/6.3.0/../../../../arm-buildroot-linux-gnueabihf/bin/ld: warning: unparse_path.o uses 2-byte wchar_t yet the output is to use 4-byte wchar_t; use of wchar_t values across objects may fail

Upstream solution was do simply drop -fshort-wchar:

  https://github.com/rhinstaller/efibootmgr/commit/3466fd05c8c6f1052e0426d64eed40f8a88fd78f

So this commit simply backports this upstream patch.

Fixes:

  http://autobuild.buildroot.net/results/87c3ff2427331eb9ab31c9eb685fe80954a0a98f/

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

support/testing: remove unused variable

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

support/testing: show messages in a more readable format

Currently messages from run-tests are quite unpleasant:

[br-tests/TestPostScripts/2017-05-09 15:51:57] Building
[br-tests/TestPostScripts/2017-05-09 15:52:23] Building done
[br-tests/TestPostScripts/2017-05-09 15:52:23] Cleaning up
.[br-tests/TestNoTimezone/2017-05-09 15:52:23] Starting
[br-tests/TestNoTimezone/2017-05-09 15:52:23] Building
[br-tests/TestNoTimezone/2017-05-09 15:53:17] Building done
[br-tests/TestNoTimezone/2017-05-09 15:53:22] Cleaning up
.[br-tests/TestGlibcNonDefaultLimitedTimezone/2017-05-09 15:53:22] Starting
[br-tests/TestGlibcNonDefaultLimitedTimezone/2017-05-09 15:53:22] Building
[br-tests/TestGlibcNonDefaultLimitedTimezone/2017-05-09 15:54:33] Building done
[br-tests/TestGlibcNonDefaultLimitedTimezone/2017-05-09 15:54:37] Cleaning up
[...]

Change them in a more readable way by removing the date and using a
columnar style:

15:12:22 TestPostScripts                          Starting
15:12:25 TestPostScripts                          Building
15:12:48 TestPostScripts                          Building done
15:12:48 TestPostScripts                          Cleaning up
.15:12:48 TestNoTimezone                           Starting
15:12:54 TestNoTimezone                           Building
15:13:44 TestNoTimezone                           Building done
15:13:49 TestNoTimezone                           Cleaning up
.15:13:49 TestGlibcNonDefaultLimitedTimezone       Starting
15:14:00 TestGlibcNonDefaultLimitedTimezone       Building
15:14:56 TestGlibcNonDefaultLimitedTimezone       Building done
15:15:01 TestGlibcNonDefaultLimitedTimezone       Cleaning up
[...]

Note the '.' and other characters presented by nose2 are still
printed. They are not affected by this change.

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/ltp-testsuite: refine architecture dependency

As stated in commit [1], sparc toolchains doesn't have
any of __sync_*() family of functions implementation.

When __sync_add_and_fetch() is missing, ltp fallback to a local
implementation of tst_atomic_add_return() specific for each
supported architecture.
But there is none for sparc.

So add a dependency on BR2_TOOLCHAIN_HAS_SYNC_4 except for
architectures where a specific implementation is provided
in ltp-testsuite.

Fixes:
http://autobuild.buildroot.net/results/d7c/d7c3b145a64ed3916b89ddb4090050f3b9205e37

[1] 6856e417da4f3aa77e2a814db2a89429af072f7d

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

package/ltp-testsuite: introduce BR2_PACKAGE_LTP_TESTSUITE_ARCH_SUPPORTS

ltp-testsuite needs __sync*() built-ins for 4-byte data, except on a few
architectures for which a specific implementation is provided in
ltp-testsuite source code.

To prepare for the introduction of this dependency, add
BR2_PACKAGE_LTP_TESTSUITE_ARCH_SUPPORTS.

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas: slightly adjust how the ARCH_SUPPORTS option is defined.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/ltp-testsuite: remove ldd command test with static only build

ldd command build system try to build a shared library unconditionally:

arc-linux-gcc -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -matomic -Os -static \
-I[...]/sysroot/usr/include/tirpc   -g -O2 -fno-strict-aliasing -pipe \
-Wall -W -Wold-style-definition -shared -o lddfile1.obj.so lddfile1.o

Fixes:
http://autobuild.buildroot.net/results/2ec/2eccf9f517ab15d8d459b06195423fdfe3fba9fa

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

package/expedite: bump to the current master

The current master is ahead of one commit, this commit include
the fix for the build issue reported upstream [1].
Remove the patch 0001 which is no longer necessary.

[1] https://phab.enlightenment.org/T5440

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

mpir: fix build on MIPS R6 CPUs

Commit ef9eebf65f554d9aacfd420600c8ba17787151a8 tried to fix this
problem in a wrong way. It was fixed only for generic R6 CPUs, so the
problem still persist if you choose any other R6 CPU, like I6400 for
instance.

Fix it properly by using the right kconfig options.

Fixes:
  http://autobuild.buildroot.net/results/f33/f33cb1e0b49fe8aee68fb4b66ea4c304b71a6f8f/

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

DEVELOPERS: add a number of packages to Matt Weber

This commit adds Matt Weber as a developer for the libqmi, nginx-upload,
cgroupfs-mount, validjson and SELinux-related packages.

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

daemon: fix parallel build

The 'ready' target creates a symlink in the libslack/ subdirectory to make
up the path to header files.  Unfortunately, the target building sources
files that need that symlink, does not depend on the 'ready' target.  This
might break highly parallel builds.  As a workaround make the 'ready' target
before making 'all'.

Should fix:
http://autobuild.buildroot.net/results/ed2/ed21ac166f2151aa69a7790a17ff05f05afa512d/
http://autobuild.buildroot.net/results/333/3338539743c8b0399c6b0fcbbb7c28b58bf3f387/

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

mpir: fix build on powerpc64le

The optimized powerpc64 code is not compatible with powerpc64le, so use
the generic MPIR code on powerpc64le.

Fixes:

   http://autobuild.buildroot.net/results/17e523be894b60d7a9b2a3303f02c3fb34816828/

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

mpir: fix build on mips32r6/mips64r6

The optimized mips32 and mips64 assembly code is not compatible with
mips32r6/mips64r6, so use the generic code of MPIR on such platforms.

Fixes:

 - http://autobuild.buildroot.net/results/db2542a2ec3029eda81738a7ae5faeb9ea4aedee/
   mips32r6 toolchain

 - http://autobuild.buildroot.net/results/50b94e6d0dbc33664c6bfe06f83b211083aa07f7/
   mips64r6 toolchain

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mpir: fix build on ARM Thumb-only platforms

The optimized ARM assembly code is not suitable for Thumb-only
platforms, so use the generic code in such situations.

Fixes:

  http://autobuild.buildroot.net/results/4d7f2a1326f75e11693bdd51ad274079b0dd0ce4/
  (ARM Cortex-M4 toolchain)

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

mpir: fix build on ARM with old binutils versions

Old binutils versions are affected by a bug when the assembly code has
whitespace between [ ] signs, causing a build failure of the mpir
package with ARM toolchains using such old binutils versions.

This commit fixes that by adding a patch that removes those whitespaces.

Fixes:

 - http://autobuild.buildroot.net/results/ec67ffdef74b9a61d0491e71f8cb2f8f0b6daa50/
   armv7-ctng-linux-gnueabihf Crosstool-NG toolchain

 - http://autobuild.buildroot.net/results/2e4a27af60065db51aca9e69701bc191d24b2209/
   armv5-ctng-linux-gnueabi Crosstool-NG toolchain

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

xfsprogs: note musl incompatibility reason

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

mpv: fix X11 logic after bump to 0.25.0

Between 0.24.0 and 0.25.0, mpv has changed its logic around the
detection/usage of X11 (in upstream commit
https://github.com/mpv-player/mpv/commit/927afa311d4a89b022305508acbf4e7325fe3928). Instead
of having fine grained option for x11, xext, xinerama, xscreensaver and
xrandr, mpv now requires all five libraries to be available in order to
enable X.org support.

Therefore, this commit adjusts the mpv packaging so that:

 1. All necessary X11 libraries are selected if BR2_PACKAGE_XORG7 is
    enabled. It would indeed be too cumbersome for the user to know that
    all of x11, xext, xinerama, xscreensaver and xrandr are necessary in
    order to have X.org support.

 2. mpv.mk is simplified by enabling X.org support when
    BR2_PACKAGE_XORG7 is enabled.

Fixes:

  http://autobuild.buildroot.net/results/b470cfba451ce9ba648e3a2f3ae08429db8bdee9/
  (--disable-xss was used to disable xscreensaver support, and this
  option no longer exists)

[Peter: reword X11 comment in .mk]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

Update for 2017.05-rc1

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

CHANGES: Update for 2017.05-rc1

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

qt5base: fix examples compile with disabled gui module

Fixes [1]:

  Project ERROR: Could not find feature opengl.

[1] http://autobuild.buildroot.net/results/040/040ab283dbbec623fe5ff4eaece754d9701fafc5

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

Revert "toolchain-external: CodeSourcery NiosII 2015.11 affected by PR19405"

An autobuilder exception has been added to avoid testing qt gui module
with the CodeSourcery NiosII toolchain. This allow to remove the
BR2_TOOLCHAIN_HAS_BINUTILS_BUG_19405 symbol.

This reverts commit 5dce3c05b5ea38539f6ee6efc24c13db936260dc.

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

support/testing: add toolchain tests

This commit adds an initial toolchain test case, testing the ARM
CodeSourcery toolchain, just checking that the proper sysroot is used,
and that a minimal Linux system boots fine under Qemu.

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

support/testing: add package tests

This commit adds some basic tests for two Buildroot packages: python and
dropbear. These tests are by no mean meant to be exhaustive, but mainly
to serve as initial examples for other tests.

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

support/testing: add fs tests

This commit adds a number of test cases for various filesystem formats:
ext2/3/4, iso9660, jffs2, squashfs, ubi/ubifs and yaffs2. All of them
except yaffs2 are runtime tested. The iso9660 set of test cases is
particularly rich, testing the proper operation of the iso9660 support
with all of grub, grub2 and isolinux.

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

support/testing: add core tests

This commit adds a few Buildroot "core" tests, testing functionalities
such as:

 - post-build and post-image scripts
 - root filesystem overlays
 - timezone support

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

support/testing: core testing infrastructure

This commit adds the core of a new testing infrastructure that allows to
perform runtime testing of Buildroot generated systems. This
infrastructure uses the Python unittest logic as its foundation.

This core infrastructure commit includes the following aspects:

 - A base test class, called BRTest, defined in
   support/testing/infra/basetest.py. This base test class inherited
   from the Python provided unittest.TestCase, and must be subclassed by
   all Buildroot test cases.

   Its main purpose is to provide the Python unittest setUp() and
   tearDown() methods. In our case, setUp() takes care of building the
   Buildroot system described in the test case, and instantiate the
   Emulator object in case runtime testing is needed. The tearDown()
   method simply cleans things up (stop the emulator, remove the output
   directory).

 - A Builder class, defined in support/testing/infra/builder.py, simply
   responsible for building the Buildroot system in each test case.

 - An Emulator class, defined in support/testing/infra/emulator.py,
   responsible for running the generated system under Qemu, allowing
   each test case to run arbitrary commands inside the emulated system.

 - A run-tests script, which is the entry point to start the tests.

Even though I wrote the original version of this small infrastructure, a
huge amount of rework and improvement has been done by Maxime
Hadjinlian, and squashed into this patch. So many thanks to Maxime for
cleaning up and improving my Python code!

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

rpi-firmware: extend vcdbg architecture dependency comment

The main reason why vcdbg is only available on ARM is the fact that this is
a prebuilt ARM binary rather than the rpi-userland dependency, so adjust the
comment to clarify.

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

package/cmake: propagate libuv dependencies

In c2c06a6de02,  cmake version was bumped and a new dependency on libuv
was added, but it did not propagate all the dependencies of libuv,
missing sync_4 (probably because sync_4 was added after the patch was
sent on the list, but before it was applied to the tree).

Fix that, which fixes unmet dependencies.

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

package/rpi-firmware: fix unmet dependencies

Currently, vcdbg is only supported in 32-bit mode. Furthermore, vcdbg
needs rpi-userland, which we currently only support in 32-bit mode.

Add a dependency on BR2_arm.

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

connman: update musl incompatibility reason

connman no longer triggers a conflict of userspace and kernel headers since
commit 69ced96355 (musl: workaround kernel if_ether.h header conflict). The
real blocker for musl build (besides a few missing includes here and there)
is the missing res_ninit() implementation. Update the comment accordingly.

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

package/sunxi-mali: remove sunxi-mali libMali for r2p4 Mali kernel modules

sunxi-mali userspace drivers (libMali.so) for r2p4 Mali kernel modules
are linked against libUMP.so.2 but libump package in Buildroot only
provide libUMP.so.3.

In upstream commit 1c5063f43cdc9de341c0d63b2e3921cab86c7742 [1], library
versioning was added. For the r3p* libraries, libUMP.so.3 is used (and
provided by Buildroot libump package), but for the r2p* libraries,
libUMP.so.2 is used (and not provided by Buildroot libump package).

Due to this any program or library trying to link with -lGLESv2 or -lEGL
will fail with the following error:

arm-none-linux-gnueabi/bin/ld: warning: libUMP.so.2, needed by output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libMali.so, not found (try using -rpath or -rpath-link)

output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_close'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_mapped_pointer_get'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_secure_id_get'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_mapped_pointer_release'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libMali.so: undefined reference to `ump_reference_add'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libMali.so: undefined reference to `ump_size_get'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_reference_release'
output/host/usr/arm-buildroot-linux-gnueabi/sysroot/armv4t/usr/lib/libEGL.so: undefined reference to `ump_open'
collect2: error: ld returned 1 exit status

Since nothing provides libUMP.so.2, this commit removes the
BR2_PACKAGE_SUNXI_MALI_R2P4 option and adds the BR2_ARM_EABIHF
dependency directly to sunxi-mali package.

Fixes:

  http://autobuild.buildroot.net/results/8d0/8d0b78798abf0c4ca124952d0d0455da6f8fa14f/

[1] https://github.com/linux-sunxi/sunxi-mali-proprietary/commit/1c5063f43cdc9de341c0d63b2e3921cab86c7742
[2] http://lists.busybox.net/pipermail/buildroot/2017-February/183500.html

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Francois Perrad <francois.perrad@gadz.org>
[Thomas: fix conflicts when applying on master, tweak commit log.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

triggerhappy: bumped to 0.5.0 to fix systemd unit

This bumps the version of triggerhappy to 0.5.0
The systemd unit supplied by buildroot assumed that filename
globbing was supported by thd. But this was just recently
added. So with this version bump usage of "/dev/input/event*"
is now possible. Systemd Unit is adjusted accordingly.

eventtable.h patch was removed as this was patched upstream.

This patch fixes bug #9836.

Signed-off-by: Michael Heinemann <posted@heine.so>

libseccomp: enable on ARM

libseccomp can now be used on ARM, so allow it to be selected.

Fixes bug #9806.

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

libseccomp: introduce BR2_PACKAGE_LIBSECCOMP_ARCH_SUPPORTS

Instead of duplicating the architecture dependency between the main
option and the Config.in comment, add a
BR2_PACKAGE_LIBSECCOMP_ARCH_SUPPORTS hidden option.

This is done in preparation to enabling libseccomp on ARM.

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

package/protobuf: needs gcc >= 4.5

Even with an upstream patch from protobuf v3.3 [1], the build fail with another issue:

In file included from google/protobuf/dynamic_message.cc:80:
./google/protobuf/map_field.h: In member function 'void google::protobuf::internal::MapField<Key, T, key_wire_type, value_wire_type, default_enum_value>::Swap(google::protobuf::internal::MapFieldLite<Key, T, kKeyFieldType, kValueFieldType, default_enum_value>*)':
./google/protobuf/map_field.h:139: error: object missing in reference to 'google::protobuf::internal::MapFieldBase::repeated_field_'
./google/protobuf/map_field_inl.h:342: error: from this location
./google/protobuf/map_field.h:150: error: object missing in reference to 'google::protobuf::internal::MapFieldBase::state_'
./google/protobuf/map_field_inl.h:344: error: from this location

Add a dependency on gcc >= 4.5.

[1] https://github.com/google/protobuf/commit/a83ac8663fb8042a881bc60b12a8bd3a0c03a3ff

Fixes:
http://autobuild.buildroot.org/results/77d/77dbb6bbbc0ea9e9bcdd22b10011ef9728c20d54
http://autobuild.buildroot.org/results/21f/21f5e1ea4f37e1d174604d6da78c0e916c89f1e3
http://autobuild.buildroot.org/results/24e/24e880086c87d40b5d79a90d805acc75b33d484c

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

package/ltp-testsuite: needs threads NPTL

Fixes:
http://autobuild.buildroot.net/results/3e0/3e0bbf41e339e62422463773bf07945f048a9501

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

configs/armadeus_apf9328_defconfig: remove RPC option

The BR2_TOOLCHAIN_BUILDROOT_INET_RPC has been removed in commit
98f7de8796b282e72d42d3fc4d82d8eec43c6053, following the removal from
upstream uClibc-ng of the RPC support.

However, armadeus_apf9328_defconfig was selecting this option, causing a
build failure due to the selection of a legacy option. We simply remove
it as anyway defconfigs should not needlessly enable toolchain options.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/builds/15762159

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

configs/galileo_defconfig: remove legacy binutils version selection

Back when galileo_defconfig was added, an explicit binutils version was
used, because only binutils >= 2.25 had support for the
-momit-lock-prefix option, necessary to workaround a CPU issue.

Support for binutils 2.25 has now been removed from Buildroot, and
therefore the option no longer exists, causing a build failure. The
oldest binutils version that can be selected in Buildroot is binutils
2.26, which has support for -momit-lock-prefix. Therefore, we simply
remove the explicit binutils version selection.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/builds/15762201

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

configs/snps_archs38_zebu: rename zebu to haps

As in upstream Linux zebu_hs_smp_defconfig was renamed to
haps_hs_smp_defconfig we update Linux kernel defconfig name
respectively.

Also it fixes build failure, pointed us by Arnout:
https://gitlab.com/buildroot.org/buildroot/builds/14419271

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

board/stmicroelectronics/stm32f469-disco: remove OpenOCD patches

Since the bump of OpenOCD from 0.9.0 to 0.10.0 in commit
3b6c74d708252a32fe172be944951396d419ba88, stm32f469_disco_defconfig
fails to build:

 - 0001-add-config.patch was merged in upstream OpenOCD (commit
   26045588f462476cee356d15a24944d6c26befb4).

 - 0002-flash-nor.patch was merged in upstream OpenOCD (commit
   3e219648c80a27f0519003f3b088cbb846e255d8)

Therefore, this commit removes the two patches that are no longer
necessary.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/builds/15762279

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

configs/pc_x86_64_efi_defconfig: increase ext2 filesystem size

Since commit c6bca8cef0 removed autocalculation of the ext2 filesystem
size, the default size is now set to 60MB. However, this is too small
for pc_x86_64_efi_defconfig. Indeed, the ext2 filesystem contains the
kernel (4MB), the wireless modules (4MB), all firmware for wireless
modules (40MB), and the wifi userspace (9MB) and the udev hwdb (5MB)
which brings the total to 70MB.

Increase the filesystem size to 120000K, which is a nice and round
number and leaves enough space for overhead on a 128MB flash drive.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/builds/15762234

This commit is identical to 9c393ad2fdd2bb6a0955d91bc3d72741b9487b78
from Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>, except it
is done for pc_x86_64_efi_defconfig.

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

configs/beaglebone_qt5_defconfig: fix kernel build

In commit 4b2440b7dcc68369548c0f2be6984fe645fc6bb0 ("beaglebone: add DT
for BeagleBone Green"), both beaglebone_defconfig and
beaglebone_qt5_defconfig were updated to add "am335x-bonegreen" to the
list of Device Tree files to be produced.

However, beaglebone_qt5_defconfig uses an older kernel version than
beaglebone_defconfig, in which am335x-bonegreen doesn't exist.

Therefore, revert the change on beaglebone_qt5_defconfig.

Fixes:

  https://gitlab.com/buildroot.org/buildroot/builds/15762182

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

.gitlab-ci.yml: sync with latest defconfig additions

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

package/libubox: fix lua module install path

The LUAPATH variable is only used when installing the lua module, and
cmake knows very well how to do out-of-tree isntalls, so there is no
reason to include the staging path in LUAPATH, which will cause our
post-install sanity checks to kick in and whine:

    libubox: installs files in /home/lhk/workspace/orangepi/host/usr/a
    rm-buildroot-linux-gnueabihf/sysroot//home/lhk/workspace/orangepi

As for libuci, just pass the runtime LUAPATH.

Fixes #9856

Reported-by: linhuikui@gmail.com
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>