efl: disallow on non-MMU systems

The libeina library uses the madvise() system call, that isn't
available on non-MMU systems. Also, several other components of EFL
use fork(). Therefore, the easiest solution is to simply disallow the
EFL as a whole on non-MMU systems.

Fixes:

  http://autobuild.buildroot.org/results/ad9/ad90baa5e07569308a7e2b2510b67c5b2a563b44//

Thanks to Ryan Barnett for helping in the investigation!

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

evas: depend on thread support

Evas has an optional mechanism to do asynchronous preloading of
images. This mechanism is optional, and in commit
b6d92bf415df33ac86487d68f5f8299f406859d2 ("libevas: async image
preload support needs threads support in toolchain"), Peter made sure
to disable the asychronous preloading when no thread support was
available.

Unfortunately, it seems like disabling the asynchronous loading is
rarely used, and it in facts fails to build: a member of structure is
not present when asynchronous preloading is disabled, but the code
continues to use it.

Since the fix is not obvious, and all this mechanism seems to have
changed completely in EFL 1.8.x, and we probably don't care much about
EFL without threads, this commit adds a dependency of libevas on
thread support. Consequently, it also reverts commit
b6d92bf415df33ac86487d68f5f8299f406859d2 which is no longer necessary.

Of course, this commit propagates this additional dependency to the
reverse dependencies of libevas.

Fixes:

  http://autobuild.buildroot.org/results/6de/6de90018a9eeb9c495d15046a8b3270eb95a5550//
  http://autobuild.buildroot.org/results/693/693df99db4ab357b48d427be3a72f6d64dd53065//

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

sdl: add support for top-level parallel make

To be sure that host-autoconf dependency is already built move the
call to autogen.sh from SDL_POST_PATCH_HOOKS to SDL_PRE_CONFIGURE_HOOKS.

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Acked-by: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

ncftp: fix cross-compilation test

The cross-compilation test is based on the ability to run
a test program on the host, which is wrong.

If it runs, then the configure script concludes
that we're doing native compilation,
if it doesn't run, we're doing cross-compilation.

The configure script needs to be regenerated to fix the
cross-compilation test.

Fixes
http://autobuild.buildroot.net/results/969/969a49ae97a50634ea846a82b9c360e4fb020ace/build-end.log

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

skeleton: /etc/fstab: make sure /tmp is world writable and sticky

ramfs (which is used instead of tmpfs if CONFIG_SHMEM isn't enabled in the
kernel configuration), defaults to mode 0755 instead of 01777 like tmpfs
uses.

/tmp should be world writable and sticky, so explictly enforce the mode so
ramfs users gets it correct instead of relying on the defaults.

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

matchbox-lib: apply upstream based patch to include a missing header

The upstream patch can't be applied as-is because the source code of
the file which is patched has changed. A version bump to the latest
release doesn't fix the problem because the upstream patch was applied
after the resease.

Upstream patch:
   http://git.yoctoproject.org/cgit/cgit.cgi/libmatchbox/commit/?id=9315bcce1f01e7d6c00161442d61d3c676794d9d

Fixes:
   http://autobuild.buildroot.net/results/e4d/e4d510ab6ba69f95401e3120ab90ccdff22deb91/

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

matchbox-lib: rename patches to follow BR guidelines

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

tvheadend: fix error compiling with ccache

The tvheadend configure script fails to correctly handle spaces in any of
the options passed to it.  If ccache is used the configure script fails to
handle the space in the --cc option and errors.  Also the --cflags option
fails to handle multiple entries and only uses the first one.  This patch
simple removes setting the --cc and --cflags options as the configure
script will correctly use the CC and CFLAGS variables already passed in
TARGET_CONFIGURE_OPTS.

Signed-off-by: Martin Bark <martin@barkynet.com>
Cc: Yann E. MORIN <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

gpsd: disallow NMEA2000 CAN driver on AVR32

AVR32 is using uClibc 0.9.31, which lacks some CAN related definitions
that prevent the gpsd nmea2000 driver from building.

The rest of gpsd, with all options and drivers enabled, builds fine on
AVR32.

Fixes:

  http://autobuild.buildroot.org/results/ee2/ee2ec848e893f08fa80caf99a67e68b73b6400e8/

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

gpm: disable on !MMU, requires fork()

Fixes:

  http://autobuild.buildroot.net/results/b2b/b2beb4876fba22f139df9c336ff0fae77354721b/

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

dhcpcd: disable on static library builds

dhcpcd uses dlopen(), so it is not compatible with pure static library
builds.

Fixes:

  http://autobuild.buildroot.net/results/512/51249e8d3487e17bb6a60a99dcbd461f7b591eac/

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

coreutils: add gnulib patch to fix librt/libthread link in static case

coreutils was not building in the static link case with uClibc, due to
binaries being linked with librt (for timer functions), but not with
libpthread. However, librt uses functions from libpthread, so both
must be specified. This lack of linking against libpthread was caused
by a deficiency in one of the gnulib m4 macro, for which we add a patch.

Fixes:

  http://autobuild.buildroot.org/results/bd4/bd4cc779b46e3837e3d6c43c97eaf42fdfc3a288/

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

pv: fix minisetlocale related build failure

pv configure script is somewhat stupid: if it cannot find the host
gettext tool msgfmt, it concludes that gettext is not available, and
provides its own minimal version. Unfortunately, this minimal version
conflicts with the available target gettext. We fix this by ensuring
that host-gettext is built if gettext support is enabled, by adding it
to the dependencies.

Fixes:

  http://autobuild.buildroot.org/results/d30/d30db79190f2cf8953751d55a5d34644f60daec4/

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

gdb: cross-gdb has no support for NIOS 2

Fixes:

  http://autobuild.buildroot.org/results/df0/df09e1fe301480b599be777bf26874d66a152810//

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

libv4l: fix build on AArch64

AArch64 does not implement the SYS_open system call, but only the
SYS_openat system call, as should be done for all modern architectures
merged into the Linux kernel. This causes problems when building
libv4l 0.8.9, which are fixed by integrating an upstream patch,
available in a more recent version of libv4l. Backporting the patch is
preferred over bumping libv4l, as libv4l has a fairly large number of
reverse dependencies, and we don't want to break anything that close
to the release.

Note that this commit also renames the existing libv4l-largefile.patch
to libv4l-01-largefile.patch in order to comply with the Buildroot
patch naming scheme, and guarantee the order in which patches are
applied.

Fixes:

  http://autobuild.buildroot.org/results/ed6/ed66792e332ee0256a17f98cbe21dfcfe6e8743b/

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

libsepol: really fix static library build

In commit 52eeb04c7608b9675bd8c2d57d3a4d1373ee72ad ('libsepol: fix
static-only library build') I forgot to git add the changes made to
the libsepol.mk file itself. Those changes are needed to actually pass
STATIC=<something> when building static libraries only.

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

manual/faq: add section about why no binary packages

It comes up every now and then on the list, so better be prepared to
point at the manual, rather than rehash the same every time.

Most of the chapter is a copy-paste of the report from the Buildroot
Developpers Day in Pragues, 2011-10-28:
    http://lists.busybox.net/pipermail/buildroot/2011-November/047229.html

We consider the opinions expressed in that report to be still valid
now, two years later.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Samuel Martin <s.martin49@gmail.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
Cc: Baruch Siach <baruch@tkos.co.il>
Cc: Danomi Manchego <danomimanchego123@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

opencv: add missing indirect dependency from gstreamer

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libplayer: add missing indirect dependency from gstreamer

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

xscreensaver: add missing indirect dependencies from libgtk2

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

libsepol: fix static-only library build

Add a patch that improves the libsepol build system to support
static-only library build.

Fixes:

  http://autobuild.buildroot.org/results/30c/30cb919d2bbf51bfdf0a83c618124d927d046a61/build-end.log

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

icu: not available when BR2_BINFMT_FLAT is used

icu does not recognize 'uclinux' as a supported platform. While making
it recognize uclinux is easy, there is another problem down the road:
icu does very weird things to generate an ELF library containing
static data (libicudata.a), and the generated library being ELF, it is
not compatible with the FLAT binary format expected by uclinux
platforms such as Blackfin in FLAT format.

Therefore, we simply disallow the selection of icu on FLAT
platforms.

Note that adding a dependency on BR2_BINFMT_ELF doesn't work, because
BR2_BINFMT_FDPIC is considered to be separate (even if technically
FDPIC is a derivative of ELF). That's why the dependency we're adding
is "depends on !BR2_BINFMT_FLAT" and not "depends on BR2_BINFMT_ELF".

Fixes:

  http://autobuild.buildroot.org/results/b41/b415fed7fae4012bad7d8b53a481bd71bdab716f/build-end.log

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

Update for 2014.02-rc2

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

webkit: disable on < ARMv5t

webkit needs the ARM blx instruction, which is only available on >=
ARMv5t, so we have to make sure the webkit package cannot be enabled
on < ARMv5t. In order to achieve this, this commit introduces the
BR2_PACKAGE_WEBKIT_ARCH_SUPPORTS hidden Config.in knob, which avoids
duplicating all over the place the complex architecture dependencies
of webkit.

Fixes:

  http://autobuild.buildroot.org/results/fdf/fdf8bc7660ac251792d0542d2729ea22753d3789/build-end.log

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

libvncserver: needs mmu

Fixes http://autobuild.buildroot.net/results/30e/30e6a879ab6d84df4688c3d29761522f68dad9fc/

fork() is only used in VNCommand.c, so it seems doable to patch the sources
(and send upstream) to make it work on !mmu if people are interested.

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

lvm2: disable static builds

When BR2_PREFER_STATIC_LIB is defined it still try to build the shared
executables and it fails.

Fixes:
http://autobuild.buildroot.net/results/b0b/b0b386969459dda9c294f1ccb4927ca225fa6bdd/

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

thrift: fixes compilation issues

Apply upstreamed patches:
https://issues.apache.org/jira/browse/THRIFT-2229
https://issues.apache.org/jira/browse/THRIFT-2367

Fixes:
http://autobuild.buildroot.net/results/efb/efbbae54b5a16b26cac7fcc69ea99bb6f8661fe8/build-end.log
http://autobuild.buildroot.net/results/f83/f837440fdf8b3bddfc6cc548902a0453640fa6b4/build-end.log

[Romain: patch 05 rebased on top of thrift-0.9.1]
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

dropbear: bump version

Fixes a number of issues. From the changelog:

- Avoid getting stuck in a loop writing huge key files, reported by Bruno
  Thomsen
- Avoid crash on exit due to cleaned up keys before last packets are sent,
  debugged by Ronald Wahl
- Fix a race condition in rekeying where Dropbear would exit if it received a
  still-in-flight packet after initiating rekeying. Reported by Oliver Metz.
  This is a longstanding bug but is triggered more easily since 2013.57
- Ensure that generated RSA keys are always exactly the length
  requested. Previously Dropbear always generated N+16 or N+15 bit keys.
  Thanks to Unit 193
- Fix DROPBEAR_CLI_IMMEDIATE_AUTH mode which saves a network round trip if the
  first public key succeeds. Still not enabled by default, needs more
  compatibility testing with other implementations.
- Fix signal handlers so that errno is saved, thanks to Erik Ahlén for a patch
  and Mark Wickham for independently spotting the same problem.

And at the same time get rid of the TIMEVAL_TO_TIMESPEC as dropbear provides a
fallback implementation if not available (and has done so since 2006).

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

linux: bump to version 3.13.3

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

linux-headers: bump 3.{2,4,10,12,13}.x stable versions

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

linux-headers: remove unused 3.3.8 patch

3.3.8 support is removed since 334dca61 (kernel-headers: remove deprecated
versions 3.1, 3.3, 3.5), so let's get rid of the patch as well.

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

mplayer: add missing reverse dependencies for nios2 and sh2a

Propagate to comment and libplayer unsupported platforms bfin and sh2a.

Fixes:
http://autobuild.buildroot.net/results/5d8/5d80ade791319b6a66ac1c342f7b1d25c023f0af/

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mplayer: remove sh2 support

Remove a leftover of the commit commit e543f5a10488b020bb4ccb03e6de0e455c64841a
(arch: remove sh2, sh3 and sh3eb support).

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

mplayer: not available on bfin

Fixes:
http://autobuild.buildroot.net/results/109/109b7c9b575c7362c4982a908a1e3695e7590ebd/

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

host python/python3: conditionally disable unicodedata

The host python always had --disable-unicodedata, regardless of the
corresponding configuration option BR2_PACKAGE_PYTHON_UNICODEDATA.
Since the host python is used to byte-compile python modules, this meant
that such modules could not contain unicode strings. For example, following
statement in a python module:
    print u"\N{SOLIDUS}"

would cause the byte-compilation to fail with message:
    SyntaxError: ("(unicode error) \\N escapes not supported (can't load
    unicodedata module)",

Instead, conditionally disable unicodedata based on
BR2_PACKAGE_PYTHON_UNICODEDATA, also for the host python.

This fixes bug #6542 (https://bugs.busybox.net/show_bug.cgi?id=6542)

Reported-by: Gernot Vormayr <gvormayr@gmail.com>
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

kexec: fix zImage probe

Add upstream patch to fix probe of zImage kernels.

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

transmission: fix uTP enabled build with ccache

Transmission would incorrectly determine the C++ compiler when ccache is
enabled, causing a build with uTP to fail at the configure step.
This patch adds a patch against transmission, fixing the problem.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

qtuio: correct install-target/staging when Qt library type is 'static'

If BR2_PACKAGE_QT_STATIC is set, qtuio will not build a .so file, but .a.
However, the custom INSTALL_TARGET_CMDS and INSTALL_STAGING_CMDS
unconditionally attempted to copy the .so file.

This commit checks the requested Qt library type and copies the right
library for each case, taking into account that the static .a file does not
need to be copied to the target directory.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

vlc: explicitly disable opencv support

vlc can use opencv in video filters, but it requires a couple of
features enabled in opencv; so, simply force disabling opencv support.

Note that vlc does includes the legacy opencv/cv.h header, which pulls a
number of opencv's features (core, video, imgproc, ...).

Fixes:
  http://autobuild.buildroot.org/results/39c/39c77ffb5a5599d0b09422433c747b2bac185c4f/
  http://autobuild.buildroot.org/results/a79/a79b055da09e4d8ede263251df8461bc8a64569a/

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

beaglebone: add DT for BeagleBone Black

The devicetree data for BeagleBone Black is the am335x–boneblack.dts file
(includes "am33xx.dtsi" and "am335x-bone-common.dtsi")

BeagleBone White uses the am335x-bone.dts file.

Signed-off-by: Marcelo Gutiérrez <kuyurix@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

vlc: needs dependency on host-lua if Lua is selected

If Lua support is requested in VLC, its configure script tries to find the
luac byte compiler, which fails if host-lua is not yet built.

This can be easily reproduced by setting a minimal config with vlc and Lua
enabled, and running: 'make clean toolchain vlc'. Final output is:

checking for LUA... no
configure: WARNING: Package lua5.2 was not found in the pkg-config search path.
Perhaps you should add the directory containing `lua5.2.pc'
to the PKG_CONFIG_PATH environment variable
No package 'lua5.2' found, trying lua 5.1 instead
checking for LUA... no
configure: WARNING: Package lua5.1 was not found in the pkg-config search path.
Perhaps you should add the directory containing `lua5.1.pc'
to the PKG_CONFIG_PATH environment variable
No package 'lua5.1' found, trying lua >= 5.1 instead
checking for LUA... yes
checking for luac... no
configure: error: Could not find the LUA byte compiler.
make: *** [<buildroot>/output/build/vlc-2.1.2/.stamp_configured] Error 1

Fix this problem by setting host-lua as a dependency to vlc.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

opencv: always enable opencv_core module when opencv is enabled

It does not make much sense enabling opencv without its core module.

This configuration leads to build nothing (since all modules depend on
the core one), but install the configuration files (*.pc and *.cmake)
anyway.

This absurd situation may break the build-system of other packages
that would correctly find the *.pc (but does not check for the modules
they actually use), but would not build because of missing headers and
libraries.

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

jimtcl: don't attempt to handle ccache internally

jimtcl tries to use 'ccache' (a non-buildroot host version) which may not
exist on the host system. If ccache is enabled in buildroot, the compiler
used by jimtcl is:
ccache <buildroot>/.../ccache <buildroot>/.../<tuple>-gcc

If ccache is not present on the host, this results in the build error:

ccache <buildroot>/.../ccache <buildroot>/.../<tuple>-gcc
-D_GNU_SOURCE -Wall  -I. -fpic -pipe -Os    -c -o jim-subcmd.o jim-subcmd.c
make[1]: ccache: Command not found

This patch passes 'CCACHE=none' to the 'configure' script, disabling the
internal handling of ccache, so that ccache can be transparently passed
through CC.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

imagemagick: needs mmu

Fixes:
http://autobuild.buildroot.net/results/d2f/d2f673b71f771ab5a6435805e39338fd023f04c8/build-end.log

The complete log file says:
libMagickCore-6.Q16.so: undefined reference to `_fork'

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

feh: needs mmu

Fixes http://autobuild.buildroot.net/results/8f4/8f447412460e404aa723242dd8281b40317a017b/

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

collectd: fix static linking

collectd-tg indirectly uses pthread, but forgets to link with -pthread.

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

netsnmp: fix static linking with openssl

Fixes http://autobuild.buildroot.net/results/084/0845bde5d234bd00d23c7a1b7e18746882dcdde4/

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

uclibc: copy config file from configure iso patch step

The three typical packages that use .config files in buildroot copy the
config file at different times in the build process:

    busybox copies its .config from the post-extract hook.
    linux copies its .config in the configure_cmds.
    uclibc copies its .config from the post-patch hook.

Copying the .config file from the configure step is the only way to properly
support an OVERRIDE_SRCDIR that does not yet have the .config file, because
the extract and patch steps are skipped in that case.

In a previous patch, the situation was already fixed for busybox. This patch
applies the same fix to uclibc: copy the config file from the configure
step, as linux is doing.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

busybox: copy config file from configure iso extract step

The three typical packages that use .config files in buildroot copy the
config file at different times in the build process:

    busybox copies its .config from the post-extract hook.
    linux copies its .config in the configure_cmds.
    uclibc copies its .config from the post-patch hook.

Copying the .config file from the configure step is the only way to properly
support an OVERRIDE_SRCDIR that does not yet have the .config file, because
the extract and patch steps are skipped in that case.

For example, when setting a BUSYBOX_OVERRIDE_SRCDIR to a cleanly extracted
busybox tarball:

$ make busybox-dirclean busybox
rm -Rf [..]/output/build/busybox-custom
>>> busybox custom Syncing from source dir
>>> /home/tdescham/repo/contrib/busybox-1.21.1
rsync -au --exclude .svn --exclude .git --exclude .hg --exclude .bzr
            --exclude CVS /home/tdescham/repo/contrib/busybox-1.21.1/
            [..]/output/build/busybox-custom
>>> busybox custom Configuring
/bin/sed -i -e "/\\<CONFIG_NOMMU\\>/d"
                            [..]/output/build/busybox-custom/.config
/bin/sed: can't read [..]/output/build/busybox-custom/.config:
                                            No such file or directory
make: *** [[..]/output/build/busybox-custom/.stamp_configured] Error 2

This patch modifies busybox.mk to copy the config file from the configure
step instead, as linux is doing, and fixing the described scenario.

This fixes bug #5030: https://bugs.busybox.net/show_bug.cgi?id=5030

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

docs/web: give a little bit more exposure to our Google+ page

Add a small icon that links the Buildroot home page to our Google+
page.

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

nut: disable parallel builds

The build system has a race condition causing build failure,
so ensure make -j1 is used.

Fixes:
http://autobuild.buildroot.net/results/0dd/0ddd856bcbec2db0500791fd428ba053d6e4fa1b/build-end.log

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

nut: disable hal

[Peter: add to existing _CONF_OPT line as suggested by Yann]
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

systemd: fix build with uClibc toolchains

Currently, the build of systemd is broken with the internal toolchain
backend, because we have uClibc patches that add support for
execvpe(), but we also have a patch for systemd that adds execvpe(),
which was added when the internal uClibc didn't support execvpe().

However, simply dropping the patch is not a solution, as it would
break the build of systemd with any other uClibc than the ones built
with Buildroot. For example, a Crosstool-NG uClibc toolchain would
fail.

Beyond this execvpe() problem, there are also other problems that
prevented systemd from being built with unpatched uClibc.

This patch does the following:

 * Add sequence numbers of systemd patches.
 * Regenerate them to avoid fuzz.
 * Improve the uClibc compatibility patch so that it:
   - Detects whether execvpe() is available or not, and if not
     available provide an implementation.
   - Detects if dup3() is available or not, and if not available
     provide an implementation.
   - Detects if MSG_CMSG_CLOEXEC is defined or not, and if not, define
     it to the appropriate value.

This has been tested with uClibc internal toolchain, uClibc external
toolchain built with Buildroot, uClibc external toolchain built with
Crosstool-NG, and glibc external toolchain.

Fixes bug #6776.

Fixes:

  http://autobuild.buildroot.org/results/591/591c6055430da334bd1e46e7d01497add45da837/build-end.log
  http://autobuild.buildroot.org/results/e58/e5851939d2837d4e35a4c1d9ca6df3c93ed5a34b/build-end.log
  http://autobuild.buildroot.org/results/92e/92e4c40c69a3feb4046b2fe5a0d8c69d5a44a157/build-end.log

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

trivial: update buildroot e-mail address to buildroot@buildroot.org

As the e-mail address buildroot@buildroot.org is now enabled, update the
e-mail addresses in the source tree from @uclibc.org and @busybox.net to our
own proper domain.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

lxc: needs setns system call, disable on nios2

Fixes http://autobuild.buildroot.net/results/f59/f592a011321429426be81bd1ecf664ad9a2c0161/

Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

package/elfutils: disable for avr32

Disable elfutils for avr32, because:
  - elfutils uses canonicalize_file_name, which came after uClibc-0.9.31.
  - avr32 uses a toolchain with uClibc-0.9.31

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

boost: test module uses fork()

Fixes http://autobuild.buildroot.net/results/17dd7946631354d59336259d5f31aa899e3599b8/

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

coreutils: correct next version in patch description

The patch that commit a728e2fe35 (coreutils: fix build against uclibc
snapshot) adds mentions the wrong coreutils version in its description.

Also fix the sign-of tag while at it.

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

gcc: xtensa: fix __builtin_frame_address use with shared libgcc

The xtensa port uses __xtensa_libgcc_window_spill in libgcc to implement
__builtin_frame_address. This symbol is local/hidden in libgcc. This is not a
problem when linking against static libgcc. But g++ defaults to
-shared-libgcc, thus breaking link against C++ shared libraries that are using
__builtin_frame_address as follows:

ld: test: hidden symbol `__xtensa_libgcc_window_spill' in .../libgcc.a(lib2funcs.o) is referenced by DSO

Add upstream patches that make libgcc_s.so a linker script that links in
unresolved symbols from the static libgcc, similar to the ARM and PowerPC
ports.

Fixes:
http://autobuild.buildroot.net/results/e2d/e2d1a763fa86b8575e2e48e6d73c018175f43e7c/

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

coreutils: fix build against uclibc snapshot

uClibc development version adds support for POSIX spawn routines. However,
unlike glibc these routines are in librt. This breaks gnulib autoconf
detection. Teach gnulib autoconf to look for POSIX spawn in librt.

Fixes:
http://autobuild.buildroot.net/results/246/246b3778a1a646afd1c8b9c17b4579fb5a27120e/

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

gst1-libav: fix build for xtensa

gst1-libav needs the same fix for xtensa like ffmpeg got in commit
4ec35e76b8a3 (ffmpeg: fix build for xtensa).

Fixes:
http://autobuild.buildroot.net/results/33f/33fcbb79809719f68f3ee2cdb0fcacad82b9e0eb/

[Peter: pass the correct flags, not the ones from ffmpeg]
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

iozone: needs thread support in toolchain

Fixes:

  http://autobuild.buildroot.org/results/2a3/2a33d2c7535a9d867d76dd5cf05e1bcc3f5cdc38//

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

docs/news.html: add 2014.02-rc1 announcement link

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

busybox: ensure SHA passwords are supported if used

As reported on IRC, our default busybox configuration doesn't support SHA
encoded passwords, breaking login if enabled under system configuration.

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

Update for 2014.02-rc1

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

CHANGES: update with recent changes

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

sdl: fix autoreconf timing issue

Context:
The autobuilders were failing on the symbol _XData32 being in conflicts.
A patch had been added to SDL to add a check to the configure.in

Problem:
Sometimes, the build would fail, because of an _XData32 symbol being in
conflicts eventhrough the patch was here.

What was happening:
Following the classic buildroot workflow:
   - Extract
   - [...]
   - Apply 001 patch, which touches configure.in AND configure
   - Apply 002 patch, which touches configure.in
   - Invoke autogen.sh
   - [...]

Right before running autogen.sh, we have configure.in which is more
recent than configure, which is fine.
We then, execute autogen.sh which, basically, runs autoconf.

If your machine was lighty loaded, the time difference between
configure.in and configure was really tiny (ms order), which seems to be
neglected by autoconf.
The results was that the configure was *NOT* generated. And our second
patch was not taken into account.

If your machine was under heavy load, the time difference between the
two files would have been greater and then *maybe* picked up by
autoconf. And then the configure file was re-generated.

When the 0001 patch was introduced, SDL package did *NOT* run it's
autogen.sh, which is why it touches also the configure.
This came later, causing this behavior.

Fixes:
  http://autobuild.buildroot.net/results/d1c/d1c36f634dbf6b6e5d18444c2a23dfd129202b80/

Signed-off-by: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libsigsegv: disable for ARC

Fixes http://autobuild.buildroot.net/results/5cd/5cd051937a9471f7fb47363e868cb4b61d788d75/build-end.log

The ARC uClibc port doesn't provide ucontext_i.sym file required to enable
context manipulation routines that libsigsegv needs.

Signed-off-by: Anton Kolesov <anton.kolesov@synopsys.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

rpcbind: uses fork(), needs MMU

Fixes:

  http://autobuild.buildroot.net/results/d7c/d7c788749b54275154cc3934a7a32385cd72be61/

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

Makefile.legacy: fix recursive invocation with BUILDROOT_DL_DIR and _CONFIG

The legacy support for the old BUILDROOT_DL_DIR and BUILDROOT_CONFIG
breaks down when make is invoked recursively - which is done in a few
cases, e.g. silentoldconfig, external-defs, ... These targets always
give a legacy error.

For BUILDROOT_DL_DIR, this is fixed by making sure that the original
value of BR2_DL_DIR taken from the environment is also exported again.

For BUILDROOT_CONFIG, this is fixed with an additional comparison of
the environment variable's value with the fake value that we introduce
ourselves.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

jpeg-turbo: fix "jpegtran" spelling in target-install hook

Use correct bin name so that JPEG_TURBO_REMOVE_USELESS_TOOLS fixup
actually removes it.

Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libjpeg: fix "jpegtran" spelling in target-install hook

Use correct bin name so that LIBJPEG_REMOVE_USELESS_TOOLS fixup
actually removes it.

Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

CHANGES: update with recent changes

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

configs/raspberrypi: remove packages not strictly required

rpi-userland and libcoffi are not strictly required to boot up the
Raspberry Pi, so remove them from the defconfig.

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

configs/raspberrypi: bump kernel version

Bump from rpi-3.10.27 to rpi-3.10.29.

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

cpanminus: remove it

cpanminus was marked as broken, now it becomes useless with the Perl infrastructure.

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

perl: remove useless patch

The Perl infrastructure uses a host version of Module-Build
(see previous commit "host-perl-module-build: new package").
The target version is never used, so no need to patch it.

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

host-perl: export and rename PERL5LIB

The official Perl documentation refers only to PERL5LIB.
PERLLIB is obsolete, we don't need to be compatible with Perl4.

PERL5LIB is used in all environment, so it is simpler to just export it.

Signed-off-by: Francois Perrad <francois.perrad@gadz.org>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

beaglebone: change to zImage

The default U-Boot configuration is now to use a zImage instead of uImage.

Signed-off-by: Anders Darander <anders@chargestorm.se>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

beaglebone: update to latest U-Boot

Update to the default version of U-Boot, at this time 2013.10.

Tested with an old uEnv.txt (i.e. using uImage) and with the
currently prefered zImage.

Signed-off-by: Anders Darander <anders@chargestorm.se>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

am33x-cm3: update to latest version

This goes hand-in-hand with the kernel update.
The version is taken from meta-ti.

The new kernel version, 3.12, requires a newer PM FW.

This should correspond to v05.00.00.02, though that tag
has not been pushed...

Signed-off-by: Anders Darander <anders@chargestorm.se>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

beaglebone: update to TI 3.12 kernel

Git repo version is taken from meta-ti.
It's a 3.12.10 kernel with TI's own patches on top.

The defconfig has been kept rather close to our old defconfig.
The kernel has been boot-tested using an NFS-root.

Apart from the kernel change, the linux headers has also been
updated from 3.8.12 to 3.12.10.

[Peter: drop UEVENT_HELPER_PATH setting]
Signed-off-by: Anders Darander <anders@chargestorm.se>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

madplay: fixes build when toolchain has locale support

Otherwise build fails in intl/relocatable.c with:
./relocatable.c: In function 'libintl_relocate':
./relocatable.c:402:40: error: 'INSTALLPREFIX' undeclared (first use in this function)
./relocatable.c:402:40: note: each undeclared identifier is reported only once for each function it appears in

[Peter: drop redundant 'else' part]
Signed-off-by: Julien Boibessot <julien.boibessot@armadeus.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

libtorrent: disable on ARC arch

Libtorrent has dependencies on atomic operations. However ARC GCC doesn't
implement those at the moment so package should be disabled. Rtorrent
depends on libtorrent.

This fixes:
http://autobuild.buildroot.net/results/51414e22c8ea47c38df20d864526a370c5a895d7/

Signed-off-by: Anton Kolesov <anton.kolesov@synopsys.com>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

barebox: bump to version 2014.02.0

Signed-off-by: Fabio Porcedda <fabio.porcedda@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

docs/news.html: add report from FOSDEM meeting

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

linux: don't automatically set uevent_helper with mdev /dev management

When mdev /dev management is chosen in the buildroot configuration, the
Linux configuration is updated automatically to set option
CONFIG_UEVENT_HELPER_PATH to "/sbin/mdev". However, the help text of this
option explicitly recommends not setting this option due to large
performance impact during boot (experienced first hand by the reporter ánd
author).

The mdev startup script S10mdev already sets the helper during userspace
boot, which will make sure mdev is working correctly.

Fixes bug #6596: https://bugs.busybox.net/show_bug.cgi?id=6596

Reported-by: Andreas Koop <andreas.koop@zf.com>
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

CHANGES: update with recent changes

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

infra: replace BUILDROOT_CONFIG with BR2_CONFIG

To make the naming consistent (all user-visible options should be
prefixed with BR2_).

An entry is added to Makefile.legacy to warn users who have set
BUILDROOT_CONFIG but not BR2_CONFIG.

Still export BUILDROOT_CONFIG but pointing to some phony value, to
make sure that scripts that still use it fail in a predictable way.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Tested-by: Jérémy Rosen <jeremy.rosen@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

vlc: fix typo when setting VLC_CONF_ENV

Fixes:
  http://autobuild.buildroot.org/results/917/917c26656557d50f869639162ec7e9d0f3399824/build-end.log

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

vlc: fix glx support

glx support is provided by the mesa3d package; so, enable it only when
mesa3d is part of the package selection.

Fixes:
  http://autobuild.buildroot.org/results/9b4/9b46021ea0064850ce8f1adcb5455d6da5bff459/build-end.log

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

kexec: fix build failure

Add patch retrived from upstream and fixing undefined bzImage_support_efi_boot
build failure.

Fixes:
  http://autobuild.buildroot.org/results/c35/c3525dd25a37db7a976c7019901160e449363b5d/build-end.log

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

kexec: rename patch to add sequence number

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

trace-cmd: needs mmu

Fixes:
http://autobuild.buildroot.net/results/0b7/0b73c684b3d7edcd818cb113fda534e552028d58/build-end.log

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

trace-cmd: disable on avr32

AVR32 does not have TLS support.

Fixes:
http://autobuild.buildroot.net/results/e24/e24af3ceb125da50f0f5cbffb0e3ca091c68af33/build-end.log

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

infra: replace BUILDROOT_DL_DIR with BR2_DL_DIR.

To make the naming consistent (all user-visible options should be
prefixed BR2_).

An entry is added to Makefile.legacy to warn users who have set
BUILDROOT_DL_DIR but not BR2_DL_DIR.

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

kmod: really pass --disable-manpages

The line below overwrites KMOD_CONF_OPT, so move the --disable-manpages
options below.

Fixes http://autobuild.buildroot.net/results/842/842131307c171e1bc5a1e918996d53b00ddb1be3/

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

pkg-autotools: fixup pkg_CONFIGURE_CMDS after DISABLE_DOCUMENTATION removal

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

infra: remove separate definition of DISABLE_DOCUMENTATION

Since we now always disable documentation on target, there is no need to
have a separate variable DISABLE_DOCUMENTATION anymore.

Suggested-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>

kernel-headers: remove deprecated versions 3.1, 3.3, 3.5

The Linux kernel headers versions 3.1, 3.3, and 3.5 have been deprecated
since 2013.02 and thus can be removed in 2014.02.
For legacy handling, we automatically select versions 3.2, 3.4, and 3.6,
respectively.

Additionally, this patch removes the now unused symbol
BR2_DEPRECATED_SINCE_2013_02.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>