toolchain/helper: don't follow symlinks when copying libs to target

In 2a87b64 (toolchain-external: align library locations in target and
staging dir), copying the libraries from the sysroot to the target was
changed to a simple find-based solution.

To be sure that the staging directory was entered to find the libraries,
in case the variable was pointing to a symlink, the -L clause to find
was used.

However, that causes extraneous libraries to be copied over.

For example, a ct-ng toolchain would have this sysroot (e.g for an arm
32-bit toolchain):

    .../sysroot/lib/
    .../sysroot/lib32 -> lib
    .../sysroot/lib64 -> lib
    .../sysroot/usr/lib/
    .../sysroot/usr/lib32 -> lib
    .../sysroot/usr/lib64 -> lib

Which we would carry as-is to our own sysroot.

But then, in target, our skeleton creates the /lib/ and /usr/lib
directories, with the necessary lib32 or lib64 symlink pointing to it.
In this case, a lib32->lib symlink is created, but no lib64 symlink
since this is a 32-bit architecture.

To copy the required libraries from staging into target, we scan the
staging directory for all occurences of the required libraries, and copy
them over to target, keeping the same directory layout relative to the
sysroot.

For example:
    .../sysroot/usr/lib/libfoo.so   -->  .../target/usr/lib/libfoo.so
    .../sysroot/usr/lib32/libbar.so -->  .../target/usr/lib32/libbar.so
    .../sysroot/usr/lib64/libbuz.so -->  .../target/usr/lib64/libbuz.so

So, when we copy over the libraries from our staging to the target
directory, the "find -L .../sysroot -name libblabla.so.*" would find
multiple instances of libblabla, each in the /usr/lib /usr/lib32 and
/usr/lib64 locations (they are all the exact same file, though).

Since we do have the /usr/lib32->lib symlink, all is OK (but there are
two copies going on, which could be avoided). However, since we do not
have the /usr/lib64->lib symlink, the /usr/lib64/ directory is created.

This was very difficult to observe, as no /lib64/ directory is created,
only the /usr/lib64/ one was. To top it off, this only happens with a
merged /usr, which does not seem like not a common case without systemd.

Since the reason to use -L was to be sure to enter our staging
directory, we just need to ensure that the path ends up with a slash, as
was already talked about in this thread:
    http://lists.busybox.net/pipermail/buildroot/2016-April/159737.html

After further discussion, it turns out that the original patch came along
because of the confusion between output/staging (which is a symlink) and
$(STAGING_DIR) which expands to output/host/usr/<tupple>/sysroot (which is
never a symlink), so the symlink handling isn't really needed at all.

[Peter: drop description comment, extend description]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/rpm: needs gcc >= 5.x on SuperH

Compilation triggers an ICE in gcc with gcc 4.9

../db/dist/../lock/lock_deadlock.c: In function '__lock_detect_rpmdb':
../db/dist/../lock/lock_deadlock.c:354:1: internal compiler error: Segmentation fault
 }
 ^

using this defconfig

BR2_sh=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_PACKAGE_BUSYBOX_SHOW_OTHERS=y
BR2_PACKAGE_RPM=y

Compiling rpm with gcc5 works fine using this defconfig:

BR2_sh=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_GCC_VERSION_5_X=y
BR2_PACKAGE_BUSYBOX_SHOW_OTHERS=y
BR2_PACKAGE_RPM=y

This patch adds a dependency to gcc >= 5.x to fix
http://autobuild.buildroot.net/results/e4b/e4b7705e3e148755ae34d498c860a3b9b915e0b0/

[Peter: simpify kconfig, add comment explaining why]
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/gstreamer/gst-ffmpeg: needs gcc >= 5.x on SuperH

Compilation triggers an ICE in gcc with gcc <= 4.9 using this defconfig

BR2_sh=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_PACKAGE_GSTREAMER=y
BR2_PACKAGE_GST_FFMPEG=y
BR2_PACKAGE_GST_FFMPEG_GPL=y

The problem is known upstream, a fix was never committed to gcc <= 4.9:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65151

Compiling gst-ffmpeg with gcc5 works fine using this defconfig:

BR2_sh=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_GCC_VERSION_5_X=y
BR2_PACKAGE_GSTREAMER=y
BR2_PACKAGE_GST_FFMPEG=y
BR2_PACKAGE_GST_FFMPEG_GPL=y

This patch adds a dependency to gcc >= 5.x to fix the problem as
suggested by Thomas:
http://lists.busybox.net/pipermail/buildroot/2016-February/152584.html

Fixes
http://autobuild.buildroot.net/results/939/939da0c7771ddd97c05cedc0a7afc0ad34a21312/

[Peter: fix ML link, simplify kconfig, add comment explaining why]
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/lttng-babeltrace: fix static build

This change adds a patch fixing libuuid detection on static build.

Fixes:
  http://autobuild.buildroot.net/results/43b/43b98ddf9eb44152ed9ac4a98d887af14831d8da/

[Peter: add comment why autoreconf is needed]
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

warpboard: README: Fix a few typos

Signed-off-by: Fabio Estevam <festevam@gmail.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/pulseview: fix build when linking against libatomic is needed

With some toolchains, using atomics requires to explicitly add -latomic
to the linker flags.

This change adds a patch to pulseview adding this detection and updating
the LDFLAGS when appropriate.

This patch has be sent upstream:
  http://article.gmane.org/gmane.comp.debugging.sigrok.devel/2097

Fixes:
  http://autobuild.buildroot.org/results/1e3/1e3101261252d5f30fdf842cc99604e4f4c25eef/build-end.log

Cc: Bartosz Golaszewski <bgolaszewski@baylibre.com>
Cc: Romain Naour <romain.naour@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Reviewed-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libdrm: improve atomic handling, fix SPARCv8 build

This commit improves the handling of the "atomic stuff" in the libdrm
package. libdrm can either use the atomic intrinsics (4 byte variant)
when available, or otherwise can use libatomic_ops. Note that the
dependency on atomic operations is not from libdrm itself, but only
from some specific DRM drivers only.

Amongst other things, it fixes the build of the libdrm package on
SPARCv8, therefore fixing:

  http://autobuild.buildroot.org/results/74dd29b5ea146c320fde80a87a2fc910de9b7f60/

This commit does a number of changes that are all related to each
other:

 - Removes the dependency of the Intel DRM driver on
   libatomic_ops. The Intel DRM driver builds perfectly fine without
   libatomic_ops, as long as 4-byte variant __sync operations are
   available, which is always the case on x86 and x86_84 (which are
   the only architectures on which the Intel DRM driver can be
   enabled).

 - Adds an hidden Config.in boolean option
   BR2_PACKAGE_LIBDRM_HAS_ATOMIC that allows DRM driver that need
   atomic operation to know whether atomic support is available
   (either through intrinsics or through libatomic_ops).

 - Adds an hidden BR2_PACKAGE_LIBDRM_ENABLE_ATOMIC Config.in option
   that DRM drivers that need atomic operation should select to ensure
   that the relevant dependencies are selected. It simply selects
   libatomic_ops if 4-byte atomic intrinsics are not available. We
   could let each DRM driver do this, but having an intermediate
   option avoids a bit of duplication.

 - Adds a patch that defines AO_REQUIRE_CAS before including
   <atomic_ops.h>. This is needed because libdrm uses the
   AO_compare_and_swap_full() which is only provided on all
   architectures when AO_REQUIRE_CAS is defined. The exact same fix
   was done in the erlang package in commit
   4a9df2942470241d7a96f326f0e7012aacd36f2e.

 - Adds the dependency on libatomic_ops when the package is enabled,
   and passes the necessary CFLAGS on SPARCv8 to make the thing build
   properly. The same CFLAGS are passed in the nginx package and bdwgc
   package.

Cc: Waldemar Brodkorb <wbx@openadk.org>
Cc: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

core/pkg-kconfig: pass host PKG_CONFIG_PATH env var

This is basically the same change as in 0515fe45661b6d320f8d2071df2
("Makefile: pass host PKG_CONFIG_PATH at "make menuconfig" time"). That
commit made sure to pass host PKG_CONFIG_PATH when invoking Buildroot's
own menuconfig program. This change ensures that the same is true for
third party menuconfig programs (i.e. Linux, uClibc and Busybox).

This unbreaks "make {linux,uclibc}-menuconfig" for host platforms which
rely on PKG_CONFIG_PATH to find .pc files (e.g. NixOS). (When Busybox
updates to a more recent Kconfig snapshot, one that uses pkg-config to
find ncurses, "make busybox-menuconfig" will also start working.)

Tested on Ubuntu and NixOS:

  $ make qemu_arm_versatile_defconfig
  $ make linux-menuconfig
  $ make

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

package/grantlee: disable for nios2 and microblaze

Disable grantlee due to a toolchain issue on nios2 and microblaze architecture [1].
Issue reproduced with binutils 2.26 and GCC 6

Fixes (microblaze):
http://autobuild.buildroot.net/results/091/091bf7df46d6057be44b8ca8653596bd84c38fbc

Fixes (nios2):
http://autobuild.buildroot.net/results/e05/e05280f1cc4f669b7e418161749290723179e987

[1] https://sourceware.org/bugzilla/show_bug.cgi?id=20173

Signed-off-by: Romain Naour <romain.naour@gmail.com>
[Thomas:
 - tweak how the dependency is expressed
 - propagate the dependency to the comments.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

mbedtls: fix companion programs static build

Set LINK_WITH_PTHREAD to explicitly add pthread to the list of libraries.

Fixes:
http://autobuild.buildroot.net/results/e08/e087dfa2a6057b18209cd6adff2760026a9ca1db/
http://autobuild.buildroot.net/results/615/61565f1a88b7d6eea03a355ccf8dd3fe6585c4ab/
http://autobuild.buildroot.net/results/6b2/6b267b9cba37655e7d8e6d78821f96d9a8dd7d2e/

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

package/ipsec-tools: disable for musl toolchains

__P() is used for compatibility with old K&R C compilers. With
ANSI C this macro has no effect.

Unlike for util-linux and ipkg packages where it was easy to remove
each __P() macro, ipsec-tools use it all over the tree and require a
"big" patch to enable musl support.

Since upstream seems not verry active (last release 2014-02-27)

So, disable ipsec-tools with musl based toolchains.

This fixes a compilation error with musl libc because of undeclared
__P.

Fixes:
http://autobuild.buildroot.net/results/42242e3f4485b9e77a916e6fe480c83f70e024e4

While at it, reorder "depends on" and "select" lines in Config.in

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

package/iptraf-ng: really add _GNU_SOURCE in CFLAGS

When iptraf-ng is build with musl, it needs _GNU_SOURCE in CFLAGS to define
the content of "struct tcphdr".

iptraf-ng.mk try to add _GNU_SOURCE in CFLAGS but it's not taken into account.

Add it using IPTRAF_NG_CONF_ENV instead of IPTRAF_NG_MAKE_ENV.

Fixes:
http://autobuild.buildroot.net/results/a1b/a1b18f2e3d075d349c19536a7c5553f24b75a323

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

configs: zynq: revive BR2_LINUX_KERNEL_UIMAGE

Commit 5c67cb1d0445 ("linux: use zImage by default on ARM") changed
the default kernel image, but missed to update Zynq defconfigs.

U-Boot on Zynq boards still loads uImage, so BR2_LINUX_KERNEL_UIMAGE
should be defined to generate uImage.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/mplayer: fix aarch64 compilation

Added code to define HAVE_ARMV8 when needed.

Fixes
http://autobuild.buildroot.net/results/5f8/5f85c32eb89aac48ae8da892d9800bd13274cd3e/

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

system: add help entry to "none" init system

It can be a little bit misleading to have no init system...

Add a comment that states the user has to provide his own init system,
either via a package or a rootfs overlay.

It is expected that such a user will know what to provide, so we don't
really need to specify that it should be /init or /sbin/init or any
arbitrary executable pointed to by the kernel command line "init=..."
or anything else...

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

package/mplayer: add x86-specific configure options

The mplayer configure script tries to detect the capabilities of the CPU
used by probing the host CPU. This leads to compilation failures if the
target CPU has lesser features, like missing mmx support for
BR2_x86_i686=y:

Checking for CPU vendor ... GenuineIntel (6:58:9)
Checking for CPU type ...  Intel(R) Core(TM) i7-3770S CPU @ 3.10GHz
Checking for kernel support of sse ... yes
Checking for kernel support of sse2 ... yes
Checking for kernel support of sse3 ... yes
Checking for kernel support of ssse3 ... yes
Checking for kernel support of sse4_1 ... yes
Checking for kernel support of sse4_2 ... yes
Checking for kernel support of avx ... yes

For this patch I copied most of ffmpeg configure options for x86 CPUs
because mplayer contains its own copy of ffmpeg.

Fixes
http://autobuild.buildroot.net/results/c5a/c5a722607ec9797c317b63b0fd3235608a340c98/

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

package/openpowerlink: avoid kernel header issue with musl

The Virtual Ethernet driver doesn't build when the musl libc is used on the
system. As stated in the musl wiki [1], the userspace and kernel headers are
mixed leading to a "clash" with the definitions provided by musl.

Remove netinet/if_ether.h userspace header and replace ETHER_ADDR_LEN by
ETH_ALEN [2] and ETHERMTU by ETH_DATA_LEN [3] in veth-linuxuser.c.

Fixes:
http://autobuild.buildroot.org/results/2ca/2ca04bb046263e479e7597867b56469893d3c11d/

Upsteam status: pending
https://github.com/OpenAutomationTechnologies/openPOWERLINK_V2/pull/120

[Rebase on v2.2.2]
[1] http://wiki.musl-libc.org/wiki/FAQ#Q:_why_am_i_getting_.22error:_redefinition_of_struct_ethhdr.2Ftcphdr.2Fetc.22_.3F
[2] https://git.musl-libc.org/cgit/musl/tree/include/net/ethernet.h?h=v1.1.14#n35
[3] https://git.musl-libc.org/cgit/musl/tree/include/net/ethernet.h?h=v1.1.14#n48

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

tinyalsa: fix musl build

Add upstream patch adding missing header include.

Fixes:
http://autobuild.buildroot.net/results/042/04259ddbdf24afc507b0f21e1f5ba8738ec492b1/
http://autobuild.buildroot.net/results/371/371a1eab885d1a6b5b06bdab6c2280b20593eed4/
http://autobuild.buildroot.net/results/377/37743b3e73fcb0df9fd50113e082eac712c25e35/

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

Update for 2016.05-rc3

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

php: security bump to version 5.6.22

Fixes:
Core: Integer Overflow in php_html_entities
Core: Integer underflow / arbitrary null write in fread/gzread
GD: imagescale out-of-bounds read
Intl: get_icu_value_internal out-of-bounds read

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

ffmpeg: disable for mips64r2 and mips64r6

MIPS architecture detection is not accurate and is always detected as
mips64 even if we are using mips64r2 or mips64r6. Due to that, ffmpeg's
build system will pass the -mips64 flag which will conflict with the
-march option that our toolchain wrapper uses, and it will fail to build
showing errors like this one:

error: '-mips64' conflicts with the other architecture options, which
specify a mips64r2 processor

This problem has been already fixed upstream, but we would need to
backport 17 patches plus some changes in the ffmpeg.mk file. This is too
much, so better to just disable ffmpeg for mips64r2 and mips64r6 for the
upcoming Buildroot release.

This commit can be reverted in the next ffmpeg's version bump.

Fixes:

  http://autobuild.buildroot.net/results/7fd/7fd8187c0110cdcac622e667f4a81d2db84f11ef/

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Reviewed-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

ffmpeg: add BR2_PACKAGE_FFMPEG_ARCH_SUPPORTS

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Reviewed-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/go: add double quotes around TARGET_CC/TARGET_CXX

Using double quotes around TARGET_CC/TARGET_CXX is mandatory, since
they are composed of several words when ccache support is enabled.

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

package/go: Set file timestamp

Set all file timestamps to prevent the go compiler from rebuilding any
built in packages when programs are built.

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/flannel: Add BR2_TOOLCHAIN_HAS_THREADS

flannel uses the cgo package, so needs a toolchain with thread
support.

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/go: Add HOST_GO_CGO_ENABLED

The go compiler's cgo support uses threads.  If BR2_TOOLCHAIN_HAS_THREADS is
set, build in cgo support for any go programs that may need it.  Note that
any target package needing cgo support must include
'depends on BR2_TOOLCHAIN_HAS_THREADS' in its config file.

Fixes build errors like these:

  error: #warning requested reentrant code

  http://autobuild.buildroot.net/results/42a8d07101d8d954511d1c884ecb66e8d861899e

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/flannel: Use HOST_GO_TARGET_ENV

Use the newly added HOST_GO_TARGET_ENV variable to pickup the
correct go environment for package builds.

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/go: Add HOST_GO_TARGET_ENV

For the convenience of package makefiles define the new
make variables HOST_GO_TOOLDIR and HOST_GO_TARGET_ENV.

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/go: Build host tools with host CC

The go build system doesn't have the notion of cross compiling, but just the
notion of architecture.  When the host and target architectures are different
it expects to be given a target cross compiler in CC_FOR_TARGET.  When the
architectures are the same it will use CC_FOR_TARGET for both host and target
compilation.  To work around this limitation build and install a set of
compiler and tool binaries built with CC_FOR_TARGET set to the host compiler.
Also, the go build system is not compatible with ccache, so use
HOSTCC_NOCCACHE.  See https://github.com/golang/go/issues/11685.

Fixes build errors like these:

  host/usr/bin/go: No such file or directory

  http://autobuild.buildroot.net/results/6664189a6f3a815978e8d0a1d7ef408ca47e2874/

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

eudev: link to real homepage

The Gentoo wiki page is much more informative than the download directory.
This is the official homepage according to top level README.md.

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

lirc-tools: add patch to fix parallel build issue

The lirc-tools package fails to build once in a while in the
autobuilders. Some quick analysis of the problematic Makefile.am has
revealed one issue. However, since the issue is difficult to
reproduce, we could only check that the new solution continue to work,
and we're not 100% sure it fixes the entire problem: only the
autobuilders can say, over time.

Supposedly fixes:

  http://autobuild.buildroot.org/results/eb47d57de8182d25b1dacbf0ac3726ed20063d04/

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

jamvm: add patch to fix musl build

Add a patch available from JamVM's bug tracker to fix the build with
the musl C library. The build was verified with the musl and uClibc C
libraries.

Fixes:

  http://autobuild.buildroot.org/results/8292973e9f6f2971d090f02f24d11a31709254cf/

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

package/ustr: disable on musl

libsemanage is the only package depending on ustr, both packages do not
build using a musl-based toolchain, suggested by Thomas:
http://article.gmane.org/gmane.comp.lib.uclibc.buildroot/149138

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

package/libsemanage: disable on musl

Suggested by Thomas:
http://article.gmane.org/gmane.comp.lib.uclibc.buildroot/149138
"getpwent_r() is a glibc-specific extension, so it will most likely not
 be implemented by musl."

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

package/tinc: optionally include linux/if_tun.h to fix musl build

Fixes
http://autobuild.buildroot.net/results/5b1/5b1d0c8c1bd75f7e228c313cb21b1638301568ae/

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

package/tinc: needs libdl for OpenSSL support

Fixes
http://autobuild.buildroot.net/results/d5b/d5b2f905d8da79cebda1408ffceac6d4c99f9e7b/

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

package/ltris: fix static linking with libmad

Fixes
http://autobuild.buildroot.net/results/b06/b0671af6be81550221f5abf98c2c797787590199/

[Peter: add host-pkgconf to _DEPENDENCIES]
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

hidapi: fix the patch renaming a file

The patch
0001-hidtest-dont-use-a-C-source-file-since-it-s-pure-C.patch in the
hidapi package needs to rename a file from .cpp to .c to avoid a
dependency on C++. This renaming currently uses the Git-way of
describing renames in patches. While this is interpreted properly by
recent enough versions of the 'patch' tool, it is ignored and not
understood by older versions of 'patch'. Due to this, with these older
versions of 'patch', the file is not renamed, and it causes a build
failure.

We fix this by not using the Git-like way of describing rename, but
rather using the old-style way of doing renames. It makes the patch
longer, but compatible with older versions of 'patch'.

Fixes:

  http://autobuild.buildroot.org/results/d7509d9fdf8f86332a023358a740975f535eafef/

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

cups: fix static linking problem

In static linking configurations, cups fails to build due to the lack of
Scrt1.o from uClibc toolchains. This Scrt1.o is only needed for PIE
binaries. Since we don't really care about PIE binaries in the context
of Buildroot, this commit solves the problem by patching cups to not
generate a PIE binary.

Fixes:

  http://autobuild.buildroot.net/results/445a401da2f63a6c43d7c166516287db6cc977ab/

Cc: Olivier Schonken <olivier.schonken@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

boot/uboot: fix missing host-openssl for i.MX28 target

Building an U-Boot image for the i.MX23 or i.MX28 target requires to run the
bootloaders 'mxsimage' tool on the host. As mxsimage needs unconditionally
OpenSSL, building U-Boot for those targets fails if it is not available on
the host:

tools/mxsimage.c:18:25: fatal error: openssl/evp.h: No such file or directory
  #include <openssl/evp.h>

Add the required dependency 'host-openssl' to all the different U-Boot image
types used to build a bootloader image for an i.MX23/i.MX28 target.

Also pass HOST_CFLAGS and HOST_LDFLAGS to the U-Boot build process so the right
-I/-L options will be used to find OpenSSL.

Ported from the Armadeus project:
https://sourceforge.net/p/armadeus/mailman/message/33595402/

Signed-off-by: Julien Boibessot <julien.boibessot@armadeus.com>
[Jörg: port to recent Buildroot version]
Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mpg123: unbreak static linking with alsa/portaudio

Fixes:
http://autobuild.buildroot.org/results/b2b/b2bc143fbd0b34e75a44af41ab6899dd9fa3c21a/
http://autobuild.buildroot.org/results/60d/60d56b0a75d209f2cfeff0727c2f900abc89d263/
http://autobuild.buildroot.org/results/6a8/6a8729ce69055821fdf0b91adcea2ea201d40189/

And many more.

A recent upstream build system change caused LIBS to be ignored at build
time (but still used a configure time), causing our static linking
workarounds to no longer work.

Fix it by adding a patch to cause it to no longer ignore LIBS.

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

time: mark as not available for static-only builds

Due to some code imported from gnulib, the time package doesn't build
in static-only configurations. Indeed the gnulib code redefines the
error() and error_at_line() functions, which are also provided by the
C library. Since fixing the gnulib code is really difficult, let's
just disable this package for static-only builds.

Fixes:

  http://autobuild.buildroot.org/results/a0e64faba69fa86755c693f575fb258a77e4e9d1/

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

oprofile: not available on Microblaze

Commit c45979c732cb610ad5d54e23dd7d4d49e519d45b marked OProfile as not
available on the Xtensa architecture, due to the lack of memory
barrier operations. This commit does the same for the Microblaze
architecture, for the same reason, which allows to fix the following
autobuilder failure:

  http://autobuild.buildroot.org/results/9a872ddc906e9d552d30762e849a1b537b4e5095/

It is worth mentioning that most likely Xtensa and Microblaze are
architectures implementing a strongly-ordered memory model, in which
case we could define the memory barriers as no-ops. But until someone
who actually cares about OProfile on Xtensa and Microblaze shows up,
it's probably better to disable the package on those architectures.

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

support/scripts: fix graph-build-time help text

The graph-build-time help text currently looks like this:

usage: graph-build-time [-h] [--type GRAPH_TYPE] [--order GRAPH_ORDER]
                        [--alternate-colors] [--input OUTPUT] --output OUTPUT

Obviously, naming the parameter for --input as OUTPUT is not a very
good idea, so this commit fixes that to name it "INPUT", as expected.

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

package/mesa3d: Execute MESA3D_REMOVE_OPENGL_PC also for DRI configs lacking OpenGL

Mesa3d provides libgl only with DRI drivers and X.org enabled. Since

https://git.busybox.net/buildroot/commit/package/mesa3d/mesa3d.mk?id=f1894ec95728806e09405d26663e0ea371afaeab

DRI drivers can be enabled without X.org support, but mesa3d still does
not provide OpenGL support so we also need to remove the pkgconf files
in this case. A POST_INSTALL_*_HOOK is already present but needs to be
executed not only when DRI support was disabled but also with DRI en-
and X.org disabled.

Fixes
http://autobuild.buildroot.net/results/342/3420ad6aa8e1e8151bf11b43be2c97cbe1dbb400/

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

libxslt: security bump to version 1.1.29

CVE-2015-7995 - Fix for type confusion in preprocessing attributes

Also drop upstream patch.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/skeleton: fix message about non-merged custom skeleton

Use of a merged /usr is not restricted to systemd anymore, thus the
current error message is misleading, as it only speaks about systemd.

Fix the message by just ditching the reference to systemd altogether.

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

erlang-p1-stringprep: backport upstream commit to fix musl build

Amusingly, a fix made upstream to fix the build on Windows also fixes
the build with the musl C library.

Fixes:

  http://autobuild.buildroot.net/results/d7b0cacc5d99d8dc91fdeaf770bb5d2b1b1975b6/

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

libgpgme: fix musl build with argp-standalone

The libgpgme configure script does not detect the error_t type under musl,
because musl does not define error_t in errno.h. As a result config.h adds a
local definition of error_t. When argp-standalone is present this local
definition collides with the error_t typedef in argp.h. Assume we always have
error_t when argp-standalone is present.

Fixes:
http://autobuild.buildroot.net/results/656/6566a1c77983404d90920997bfd15bd98c7ddde7/
http://autobuild.buildroot.net/results/a7b/a7b92ad68ff6e13d20ff44e955dc5748b2efd6a4/
http://autobuild.buildroot.net/results/0db/0db1b2aa0969d6cbae57b1400b7b766a5988abf9/

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

strongswan: needs __atomic operations

strongswan uses the __atomic_*() intrinsics, so we make it depend on
BR2_TOOLCHAIN_HAS_ATOMIC and link with libatomic when available. This
allows to fix the build on SPARC, therefore fixing:

  http://autobuild.buildroot.org/results/7e090237801874fb889c76f84863dc4c2ca20450/

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

webkitgtk: security bump to version 2.12.3

Fixes:
CVE-2016-1856 - allows remote attackers to execute arbitrary code or
cause a denial of service (memory corruption) via a crafted web site
CVE-2016-1857 - allows remote attackers to execute arbitrary code or
cause a denial of service (memory corruption) via a crafted web site

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/lxc: add patch to fix musl build

Add a patch to fix a build issue with the musl C library.

struct in6_addr is both defined in the C library header <netinet/in.h> and the
Linux kernel header <linux/in6.h>.

lxc_user_nic.c includes both <netinet/in.h> and <linux/if_bridge.h>. The later
one includes <linux/in6.h>.

This breaks build with the musl libc:
  error: redefinition of ‘struct in6_addr’

As lxc_user_nic.c does not use any references from <linux/if_bridge.h> it is
safe to remove this header.

Fixes:
http://autobuild.buildroot.net/results/f32/f321823be6b477be7dc55393e563a3a61794265d/
http://autobuild.buildroot.net/results/8ab/8ab2dc8e0634fbfb582db94cdf94cee5712711f9/
http://autobuild.buildroot.net/results/e27/e27d637b1bb7278ae30d86a72a2d88432d805d2c/
.. and more.

Upstream status: Pending
https://github.com/lxc/lxc/pull/1029

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

xserver_xorg-server: add explicit handling for xwayland

It's normally autodetected, which can lead to unexpected/unhandled
results.
configure checks for libdrm, libepoxy and wayland, however a proper
libxcomposite check is missing thus it can lead to build failure under
some odd conditions.
There's no autobuilder failure to quote here, however there's a mailing
list post that exemplifies this:
http://lists.busybox.net/pipermail/buildroot/2016-May/161793.html

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

flann: disable on microblaze

All gcc versions fail to build flann on the Microblaze architecture, due
to gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69401, so let's
disable this package on this architecture.

Fixes:

  http://autobuild.buildroot.net/results/3f44a1f30a88dbe4a3a83055267b472b58769e15/

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

flann: add hash file

flann is fetched from Github, so we can add a hash file for this
package.

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

flann: bump version to fix build with gcc 6

The current version of flann in Buildroot fails to build with gcc 6.x,
with the following failure:

  error: call of overloaded 'abs(flann::KDTreeIndex<flann::L2<float> >::ElementType)' is ambiguous

By bumping the upstream version one commit further, we get an upstream
fix that fixes the build problem.

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

toolchain-external: correct hash value for Linaro AArch64 toolchain source

The aarch64 Linaro toolchain source hash is not correct, probably due
to a copy/paste error. The new hash has been verified by downloading
the tarball, validating the signature, and computing the hash.

Signed-off-by: Clayton Shotwell <clayton.shotwell@rockwellcollins.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-treq: add missing BR2_INSTALL_LIBSTDCPP dependency

python-treq selects python-pyopenssl, which depends on C++ support, but
this dependency was not propagated to python-treq, causing the following
kconfig warning:

warning: (BR2_PACKAGE_PYTHON_TREQ) selects BR2_PACKAGE_PYTHON_PYOPENSSL which has unmet direct dependencies ((BR2_PACKAGE_PYTHON || BR2_PACKAGE_PYTHON3) && BR2_INSTALL_LIBSTDCPP)

This commit fixes that by adding the appropriate dependency.

Cc: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas: add Config.in comment.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

python-service-identity: add missing BR2_INSTALL_LIBSTDCPP dependency

python-service-identity selects python-pyopenssl and python-pyasn, both
of which depend on C++ support, but this dependency was not propagated
to python-service-identity causing the following kconfig warning:

warning: (BR2_PACKAGE_PYTHON_SERVICE_IDENTITY) selects BR2_PACKAGE_PYTHON_PYOPENSSL which has unmet direct dependencies ((BR2_PACKAGE_PYTHON || BR2_PACKAGE_PYTHON3) && BR2_INSTALL_LIBSTDCPP)
warning: (BR2_PACKAGE_PYTHON_SERVICE_IDENTITY) selects BR2_PACKAGE_PYTHON_PYASN which has unmet direct dependencies ((BR2_PACKAGE_PYTHON || BR2_PACKAGE_PYTHON3) && BR2_INSTALL_LIBSTDCPP)
warning: (BR2_PACKAGE_PYTHON_SERVICE_IDENTITY) selects BR2_PACKAGE_PYTHON_PYASN_MODULES which has unmet direct dependencies ((BR2_PACKAGE_PYTHON || BR2_PACKAGE_PYTHON3) && BR2_INSTALL_LIBSTDCPP)

Cc: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Yegor Yefremov <yegorslists@googlemail.com>
[Thomas: add Config.in comment.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

midori: add missing BR2_TOOLCHAIN_USES_GLIBC dependency

midori selects webkitgtk, and webkitgtk needs a glibc toolchain, but
midori did not propagate this dependency, causing the following kconfig
warning:

warning: (BR2_PACKAGE_MIDORI) selects BR2_PACKAGE_WEBKITGTK which has unmet direct dependencies (BR2_INSTALL_LIBSTDCPP && BR2_TOOLCHAIN_GCC_AT_LEAST_4_9 && BR2_TOOLCHAIN_USES_GLIBC && BR2_PACKAGE_LIBGTK3 && BR2_PACKAGE_WEBKITGTK_ARCH_SUPPORTS)

Interestingly, the Config.in comment of the midori package already
mentionned the (e)glibc dependency, but the BR2_TOOLCHAIN_USES_GLIBC
dependency was not expressed (either on the comment or on the main
package option itself).

Cc: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

qt5serialbus: depend on toolchain headers >= 3.6

qt5serialbus requires CAN FD support, that was introduced
in kernel 3.6. Add appropriate toolchain headers dependency.

Fixes:
http://autobuild.buildroot.net/results/617/6176a24799b5d80312be954089efbbb498fa1571/

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

aircrack-ng: mention script dependencies in the help text

aircrack-ng has some scripts (airmon-ng, airmon-zc) that use external
tools for their functionality. We don't select the corresponding
packages because some of the other aircrack-ng tools don't need these
at all.

Still, the user should be informed of this. So update the help text to
refer to all packages used by the scripts:
- ethtool
- iw
- rfkill
- util-linux for lspci and lsusb (script uses options not available in
  busybox)
- wireless_tools for iwconfig and iwpriv

See also bug #8936

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

package/iptraf-ng: Needs -D_GNU_SOURCE to fix musl build

Compiling iptraf-ng using musl is broken:

src/ipfrag.c:193:52: error: 'struct tcphdr' has no member named 'source'
    ftmp->s_port = ntohs(((struct tcphdr *) tpacket)->source);

Musl guards these fields in the netinet/tcp.h header file with
_GNU_SOURCE, for details see https://dev.openwrt.org/ticket/19889

Patch ported from
http://git.alpinelinux.org/cgit/aports/tree/main/iptraf-ng/APKBUILD#n28

Fixes
http://autobuild.buildroot.net/results/4ce/4cedd4dbaafca76e839f45cc0a4bb8c2c181920d/
http://autobuild.buildroot.net/results/72b/72b019ce5bb52b8b87d92cce675708f7b9f8a96e/
http://autobuild.buildroot.net/results/f67/f67f1c2a627a6d535dd5c00e2ee131b5c5a99a0a/

Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Tested-by: Jörg Krause <joerg.krause@embedded.rocks>
[Thomas: one assignment only for IPTRAF_NG_CONF_ENV.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

xorriso: fix musl build

Add a patch that adds the missing header for ssize_t.

Fixes:
http://autobuild.buildroot.net/results/820/82014874f5b97de8504d8409b286889b582a7454/

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

valgrind: disable for MIPS soft-float

As stated here [1], recent changes on the MIPS binutils sources have
made it necessary for GCC to pass the -msoft-float to the assembler. Due
to that, valgrind fails to build for MIPS soft-float when using a
version of binutils >= 2.25 because its using some hard-float
instructions.

However, we cannot just disable it for the soft-float and binutils >=
2.25 combination since external toolchains don't provide information
about the binutils version they use. So, instead, we simply disable it
for soft-float.

1: https://gcc.gnu.org/ml/gcc-patches/2014-08/msg00905.html

Fixes:
  http://autobuild.buildroot.net/results/5f5/5f576c7f8d56058a19ed0e7ff4b1ec620bcafb65/

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

libepoxy: cosmetic fixes

It's --enable-glx rather than --enable-egl, both are automatic though.
Change the autoreconf comment to reflect that it's required for all
patches.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

docs: Fixed syntax error in makedev-syntax example.

In the makedev syntax documentation, one of the examples lacks the "mode"
value.  This patch fixes that example.

Signed-off-by: Alan Yaniger <alan@tkos.co.il>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

toolchain: improve SSP logic

Don't enable SSP support on external toolchains just because they use
glibc or musl. Instead of that, make the external toolchains explictily
declare if they support SSP or not. And also add a check to detect SSP
support when using custom external toolchains.

For internal toolchains we always enable SSP support for glibc and musl.

Fixes:

  http://autobuild.buildroot.net/results/ac7c9b3ad2e52abfe6b79a80045e4218eeb87175/

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
[Thomas:
 - remove uClibc-specific SSP check, since there is now a generic
   check being done.
 - send potential compilation errors caused by the SSP check to
   oblivion, in order to avoid causing confusion for the user.
 - add autobuilder reference.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libxml2: security bump to version 2.9.4

Fixes a bunch of security issues including:

  CVE-2016-1762: Heap-based buffer overread in xmlNextChar

  CVE-2016-1834: heap-buffer-overflow in xmlStrncat

  CVE-2016-3705: Missing increments of recursion depth counter to XML parser

A few more security fixes are listed in the release announcement at
https://mail.gnome.org/archives/xml/2016-May/msg00023.html.

Also fixes:
http://autobuild.buildroot.net/results/6db/6db405a097b192876c0b1b8d59051d614563c617/
http://autobuild.buildroot.net/results/62a/62addf4abd2a0df8222a81a83c16b2b9a61c9481/
http://autobuild.buildroot.net/results/204/20402690ad05d10d456a219da5252a38badf1da0/

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

ruby: fix build for nommu

Fixes:
http://autobuild.buildroot.net/results/a8b/a8b5fe95bbc1b3cc23db828fbc80d815c88f18c2/

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

libcurl: bump version to 7.49.0

Fixes CVE-2016-3739, https://curl.haxx.se/docs/adv_20160518.html.

Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
[Thomas: add reference to the CVE being fixed, pointed by Gustavo.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/openpgm: Rework headers includes to fix build with musl libc

Fixes
http://autobuild.buildroot.net/results/420/420367cc4640c893f48336d71fa04eaeff55c8e6/

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

package/openpowerlink: fix build issue with debugging symbols

Update the top level build patch to fix a build issue with debugging symbols.

This only affect the demo applications build.

Fixes:
http://autobuild.buildroot.net/results/da4/da445b65cb136d71577f04e3a17fdb2ef6302a9b

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

liquid-dsp: wrapping isnan in T_ABS to help compilation w/ certain gcc

Backporting an upstreamed patch in order to fix a build failure like
this one:

In file included from src/math/src/polyc.c:43:0:
src/math/src/poly.findroots.c: In function
'polyc_findroots_bairstow_recursion':
src/math/src/poly.findroots.c:305:9: error: non-floating-point argument
in call to function '__builtin_isnan'
         if (isnan(du) || isnan(dv)) {

Upstream commit URL:

  https://github.com/jgaeddert/liquid-dsp/commit/3055eb3da9d0a202c1a975f7db0c8370a09a30bc

Fixes:

  http://autobuild.buildroot.net/results/01d/01d7d4f34b256bcdf30b16180c015f146bd50e63/

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

package/hplip: fix static linking with cups

Fixes
http://autobuild.buildroot.net/results/c52/c5277c42344d194e004dcd0a5af71bc6110da97e/

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

expat: add fix for CVE-2016-0718

Fixes:
CVE-2016-0718 - The Expat XML parser mishandles certain kinds of
malformed input documents, resulting in buffer overflows during
processing and error reporting. The overflows can manifest as a
segmentation fault or as memory corruption during a parse operation. The
bugs allow for a denial of service attack in many applications by an
unauthenticated attacker, and could conceivably result in remote code
execution.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

connman: remove xtables build fix patch

This was added in 0dece985 to deal with kernel headers 4.5 -> 4.5.4
breakage. Unfortunately the fix in 4.5.5 and 4.6+ doesn't deal correctly
with this causing more build breakage, so we'd rather "break" for the
small range of 4.5.x versions broken than for the rest of time.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

linux-headers: bump 3.14.x and 4.{4, 5}.x series

Kernel headers 4.5 -> 4.5.4 broke netfilter uapi headers leading to
build failures for cases where glibc's net/if.h was included before
linux/if.h
This is fixed since 4.5.5 via linux kernel commit 4a91cb61 and fixes:
http://autobuild.buildroot.net/results/736/7362431f62ebb1b436ac41ea3ef85228c763e6f3/

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Tested-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

package/qlibc: fix musl build by removing usage of internal glibc header sys/cdefs.h

Fixes
http://autobuild.buildroot.net/results/15c/15ce460214875c2e56d5d26372778b43cd7325ad/
http://autobuild.buildroot.net/results/790/79098fb27ef2d5b09df777cd7365a61d6de55c52/
http://autobuild.buildroot.net/results/8d3/8d3dad2dcf5b3c7fa94cc759386759b1ac9e3287/
http://autobuild.buildroot.net/results/9d8/9d8f6259c62e15cc528ec99d6375cc02c2bc7489/
http://autobuild.buildroot.net/results/a05/a0540ed23bd2d48cd742aa1b0228719b13e162b6/

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

package/stress-ng: mark as not available on nios2

For technical details see
https://git.busybox.net/buildroot/commit/package/ffmpeg/Config.in?id=8b58ec016973bee87a4fc1cc99a32b206bdee136

Fixes
http://autobuild.buildroot.net/results/bb3/bb3b748a14465a6168917bfbed90b2ff84fbdedf/
http://autobuild.buildroot.net/results/0ee/0ee3132f4ee6b5f21851f93516e0d0c419b36b1b/

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

package/postgresql: fix snprintf() detection

https://git.busybox.net/buildroot/commit/package/postgresql/postgresql.mk?id=3d3f70b567718ae308b5704ddd7d107c4849309e
forced postgresql not to use its own snprintf() implementation which
needs libm. This fix got broken after this upstream commit

http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=ce486056ecd28050f367894a2b5aad3656d37511#patch1

changed the variable name from
  pgac_cv_snprintf_long_long_int_format
to
  pgac_cv_snprintf_long_long_int_modifier

leading to problems in the rsyslog package while detecting postgresql
support.

Fixes
http://autobuild.buildroot.net/results/1cd/1cda32983fdf9522e4a9d48cc1304cc1451771a9/

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

package/gstreamer1/gst1-libav: not available on nios2

ffmpeg is not available on nios2, this limitation was not reverse-
propagated to this package.

Fixes
http://autobuild.buildroot.net/results/267/2674395c1d804df68e7e4bc9bba1d6b7fc6d2dce/

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

package/bluez5_utils: remove include linux/if_bridge.h to fix musl build

Fixes
http://autobuild.buildroot.net/results/eba/ebaa0bcb9c325aa6ed0bbd6c7ec75d44befa7645/

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

package/bluez5_utils: remove autoreconf

0001-Link-mcaptest-with-lrt.patch was removed 2016-01-05:
https://git.busybox.net/buildroot/commit/package/bluez5_utils?id=e2274856123fcc387df95675d7a66eb2c576eaa2

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

putty: fix musl build for 64-bit targets

Add upstream patch fixing MinGW build that also fixes musl build. Both of them
do not provide __uint64_t on 64-bit targets.

Fixes:
http://autobuild.buildroot.net/results/8f7/8f7683b2e204249c37690bae915208298ea94490/
http://autobuild.buildroot.net/results/f28/f2836f3a7e27f74f5adf38d1769b9b047d59c385/

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

putty: fix musl build

Add upstream patch for fixing build with musl.

Fixes:
http://autobuild.buildroot.net/results/84b/84b8e3e73f7e70ea1f013fc7c1120cf23095a77a/
http://autobuild.buildroot.net/results/b7c/b7cd550af983165d7d930bcc68fcab5bad00159d/

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

package/kodi-screensaver-matrixtrails: add hash

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

package/kodi: Fix crash in pictureviewer when opening directories with many files

backported upstream commit:
https://github.com/xbmc/xbmc/commit/f63563615e357b7d794a38e1d37276c325d1466f

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

package/mtr: uses fork(), needs mmu

Fixes
http://autobuild.buildroot.net/results/dcb/dcbd39878b1a80d6a16032e967bd8e2b928f7a78/
http://autobuild.buildroot.net/results/e05/e05632e0b160ee995ce46aaf3ceddf46d3658adf/

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

Update for 2016.05-rc2

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

android-tools: fix build on big endian systems

This commit adds a patch to the android-tools package to make it build
properly on big-endian systems.

Fixes:

  http://autobuild.buildroot.net/results/1b8ace1a083b419c1ab913fda0d36a1d2d910f13/ (PowerPC)
  http://autobuild.buildroot.net/results/5bb304c91367f1570903c8c527b387c65e572e56/ (Xtensa)

Cc: Gary Bisson <gary.bisson@boundarydevices.com>
Cc: Antoine Tenart <antoine.tenart@free-electrons.com>
Cc: Julien Corjon <corjon.j@ecagroup.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

boost: fix build failures on NIOSII and Microblaze

The fenv support provided by glibc on NIOSII and Microblaze is not
sufficient for Boost, causing the same build failures as the ones we had
with uClibc.

To address this, we adapt (and rename) the existing
0006-uclibc-fenv.patch by forcefully disabling fenv support in Boost on
NIOSII and Microblaze, in addition to the existing uClibc exclusion.

Fixes (for NIOSII):

  http://autobuild.buildroot.net/results/9b8c5d2cd31d63ae70b41d1434868c83e65b8423/

Fixes (for Microblaze):

  http://autobuild.buildroot.net/results/0b9db4c5cfd80d75eb620dbf7a6201faed7230a1/

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

qt5multimedia: select Qt5OpenGL module in case OpenGL support is enabled

Qt5Multimedia needs Qt5OpenGL module in case OpenGL support is
enabled.

Fixes bug reported by Marco Trapanese ([1]).

[1] http://lists.busybox.net/pipermail/buildroot/2016-May/161288.html

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

package/jack2: fix build WRT backtrace support

This change adds a patch checking for the presence of execinfo.h header
and enabling the backtrace support depending on the check result.

Fixes:
  http://autobuild.buildroot.org/results/415/415e2100dc59d35e58646c07f7cdccabecdda966/
  http://autobuild.buildroot.org/results/43c/43ca1b103434ae582fdf93cb5912b311960f303b/
  http://autobuild.buildroot.org/results/391/391e71a988250ea66ec4dbee6f60fdce9eaf2766/
  ...

Adapted from the PR:
  https://github.com/jackaudio/jack2/pull/206

Signed-off-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

assimp: disable on Microblaze

This package triggers an infinite loop bug in gcc on the Microblaze
architecture when the optimization level is O1, O2 or O3. This bug has
been reported at https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71124.

While Buildroot by default uses an Os optimization level, assimp's build
system overrides that by O3 by default.

This problem is causing timeouts in the autobuilders that make them
consume 100% of CPU during 8 hours (the timeout used by the autobuilder
scripts).

Fixes:

  http://autobuild.buildroot.net/results/084fc537ab81aed278126f173daf99f2699ef22c/

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

glibc: disable build of glibc 2.22 with gcc 6.x

glibc-2.22 does not build with gcc 6.x. The first issues can be fixed
by:

 - Using the same 0004-gcc6.patch as we're using with glibc 2.23 since
   Buildroot commit ab8de336eb39ae1cb019a72be65bd0504c86e527.

 - Backport glibc commit 5542236837c5c41435f8282ec92799f480c36f18 to fix
   strchr() build failure.

But then, there are some more failures caused by the fact that numerous
glibc files were not using modern prototypes in the function
definitions, causing build failures such as:

../sysdeps/unix/sysv/linux/dl-openat64.c:26:1: error: 'openat64' defined as variadic function without prototype [-Werror]
 openat64 (dfd, file, oflag)
 ^~~~~~~~
In file included from ../include/fcntl.h:2:0,
                 from ../sysdeps/unix/sysv/linux/dl-openat64.c:21:
../io/fcntl.h:214:12: note: previous declaration of 'openat64' was here
 extern int openat64 (int __fd, const char *__file, int __oflag, ...)
            ^~~~~~~~

This could be fixed by backporting glibc commit
9dd346ff431fc761f1b748bd4da8bb59f7652094, but this is a large commit,
touching a significant number of files. We probably don't want to take
this path, and instead making glibc 2.22 as not available with gcc 6.x
seems like the right course of action.

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

gcc: fix build of gcc 6.x on Microblaze

Due to patch 840-microblaze-enable-dwarf-eh-support.patch, gcc 6.x
does not build:

../../gcc/config/microblaze/microblaze.c: In function 'void microblaze_expand_epilogue()':
../../gcc/config/microblaze/microblaze.c:3046: error: 'gen_rtx_raw_REG' was not declared in this scope

This patch was originally added to gcc 4.9 to make it capable of
building glibc. However, this is no longer needed with gcc 6.x, which
builds glibc just fine.

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

website/news.html: add 2016.05-rc1 announcement link

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

libinput: fix missing static_assert

uClibc-ng does not define the C11 static_assert macro. Patch configure.ac to
detect this and define a nop macro.

Fixes:
http://autobuild.buildroot.net/results/3eb/3eb32c19f90a5fd8d45a0c36676e015e8278a469/
http://autobuild.buildroot.net/results/184/1844890c65615f1676a85c6fac78937249eee9f1/
http://autobuild.buildroot.net/results/3a3/3a3f8c5624e8019a6eababbf6e7440fdd668f85f/

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

package/nginx: fix patches titles

Changed [PATCH x/y] to [PATCH] at the beginning of the subject line
since the numbering is meaningless.

Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Martin Bark <martin@barkynet.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>