qt: fix mysql module build

Closes #2371

Bug 2371 was a Qt compilation issue when MySQL support was enabled,
because the Qt configuration tests were looking at host MySQL
libraries.

Even though I couldn't reproduce the bug with MySQL libraries
installed on my host, there was clearly a problem since
-I/usr/include/mysql and -L/usr/lib/mysql were part of the compilation
flags when building those tests.

This was caused by the fact that /usr/bin/mysql_config was used
instead of $(STAGING_DIR)/usr/bin/mysql_config. We fix this by passing
the path to the correct mysql_config script using the -mysql_config Qt
./configure argument. Since this mysql_config script returns the
correct flags for libraries and headers path, the hardcoded -I and -L
flag in qt.mk are no longer needed.

With this in place, Qt is no longer compiled with -I/usr/include/mysql
and -L/usr/lib/mysql, which presumably should fix the problem reported
in bug 2371.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mysql_client: bump to 5.1.53 to fix abi check problem

MySQL 5.1.47 is affected by a bug in one test called the "ABI check",
which does not work with recent gcc versions. This bug is referenced
as http://bugs.mysql.com/bug.php?id=52514 in MySQL bug tracker.

Since it has been fixed in newer versions of MySQL, we simply bump the
version to the latest available in the 5.1.x series.

The patches have not changed, they have only been refreshed.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xkeyboard-config: bump to 2.0 to fix bug 2965

Closes #2965

For some odd reason, xkeyboard-config < 1.8 was creating a symbolic
link from /usr/share/X11/xkb/xkbcomp to the xkbcomp binary. But in
cross-compilation mode, this is absurd as the xkbcomp binary to which
the link is pointing is the one in $(HOST_DIR).

This symbolic link thing has been removed completely starting from
xkeyboard-config 1.9. See
http://cgit.freedesktop.org/xkeyboard-config/commit/?id=f413dff57e77e7b01461508f74b4e92d1dc8defd.

Therefore, we simply bump xkeyboard-config to the latest available
version, 2.0.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

busybox: get rid of BR2_PACKAGE_BUSYBOX_FULLINSTALL

As discussed on IRC, this only needs to be disabled for very specific
configurations, and it can nowadays be done with a custom busybox
.config (CONFIG_INSTALL_APPLET_DONT), so get rid of the option.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

busybox: clarify BR2_PACKAGE_BUSYBOX_FULLINSTALL description

Explicitly explain what the (default) install target does, so people who
don't know busybox details understands what the option does.

Notice: Busybox can be configured to create hard links or shell wrappers
instead, but by default symlinks are used.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xlib_libX11: compile makekeys for the host correctly

As reported by Miroslav Ignjatovic in bug #2983, our hack in
xlib_libX11 to build makekeys for the host does not work, for several
reasons:

 * As we are building a tool for the host, we shouldn't pass
   -I$(STAGING_DIR)/usr/include, since the $(STAGING_DIR) contains
   headers of packages for the target.

 * Instead, we should use the headers in $(HOST_DIR)/usr/include. They
   were not used due to a typo: $(HOST_CFLAGS) must be used instead of
   $(HOSTCFLAGS).

 * Finally, in order for makekeys to find the required headers in
   $(HOST_DIR)/usr/include, we must depend on host-xproto_xproto.

This fixes bug #2983.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: fix multilib symlink for external toolchains

ARCH_SUBDIR is a shell variable, so it should be referenced with
$${ARCH_SUBDIR}. Without this, no symbolic link is created, and the
external toolchain fails to work if the non-default multilib variant
is used.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

busybox: bump 1.18.x version

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

qt: svg and multimedia modules depends on gui module as well

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

qt: phonon support depends on gui module

Otherwise configure errors out with 'Phonon enabled, but GUI disabled'.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: fix linux-% shortcut targets

linux-% shortcut targets (short for linux26-%) ignores the ouput dir
$(O) so that 'make O=output.arm linux-menuconfig' is actually run in the
default $(O) directory output/ and not in output.arm/. Fix by passing on
$(O) if set.

[Peter: Use EXTRAMAKEARGS]
Signed-off-by: Bjørn Forsman <bjorn.forsman@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: support 'make linux26-nconfig'

Signed-off-by: Bjørn Forsman <bjorn.forsman@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/gcc: bump 4.5.x version

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: add KERNEL_ARCH_PATH to fix x86-specific issue

Both i386 and x86_64 architectures are supported by the arch/x86
directory in the kernel. So, when we copy the kernel configuration
file to arch/$(KERNEL_ARCH)/configs/, it does not work because
arch/i386 and arch/x86_64 do not exist.

So, we introduce KERNEL_ARCH_PATH, which is the path to the
architecture specific directory in the kernel source tree.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: Add a Qemu mipsel Malta board

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: Add a Qemu PPC g3beige board

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: add support for ELF vmlinux image

The ELF vmlinux image found at the root of the kernel source tree is
the format that Qemu needs when emulating mips(el) or ppc targets, so
add support for it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: Add Qemu SH4 r2d configuration

[ Thomas:

   * renamed sh4_defconfig to qemu_sh4_r2d_defconfig, for consistency
     with other Qemu platforms supported
   * renamed board/qemu/sh4 to board/qemu/sh4-r2d
   * minor fixes in the readme.txt
   * remove useless statements in the minimal defconfig
   * switch to a fixed kernel version instead of "same as headers"
]

Signed-off-by: Philippe Reynes <tremyfr@yahoo.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xtensa: remove unused and useless skeleton patch

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: support a space-separated list of patches

The kernel being a component that often needs a fairly important set
of changes to be adapted to a particular hardware platform, having
maximum flexibility on the patching process is a nice
thing. Therefore, as per the discussions from the Buildroot Developer
Day, we add a mechanism to apply a list of patches (that could come
either from URLs, local files or local directories).

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

linux: Import minimal defconfigs as defconfigs

Now that minimal kernel defconfigs are used in Buildroot, the problem
is that copying those minimal configuration files to .config in the
kernel source tree does not work, as kconfig will ask interactively
what should be the value for all unspecified options.

On suggestion on Sam Ravnborg, the easiest way to solve this is to
import the minimal defconfig file as a defconfig inside the kernel
tree (in arch/$(ARCH)/configs) and configure the kernel with it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Add network scripting folders to fs/skeleton

Those folders are currently created using makedevs according to the
specifications in device_table.txt. However, as makedevs is no longer
executed when dynamic device creation methods are selected (devtmpfs,
udev, mdev), those folders must be created differently. We choose to
put them directly into the default filesystem skeleton.

Signed-off-by: Yegor Yefremov <yegor_sub1@visionsystems.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Free text field for serial port configuration

BR2_TARGET_GENERIC_GETTY_PORT has now a string type instead of choice.
This makes port configuration flexible and compact.

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

u-boot: add CROSS_COMPILE/ARCH when building host tools

Even when building the host tools, for some reason U-Boot tries to
execute the cross-compiler, so tell it which cross-compiler to use in
order to avoid failure such as:

/usr/bin/make -j12 -C /home/test/outputs/test-253-mini2440_defconfig/build/u-boot-custom tools
make[1]: arm-linux-gcc: Command not found
make[1]: Entering directory `/home/test/outputs/test-253-mini2440_defconfig/build/u-boot-custom'
for dir in tools examples api_examples ; do /usr/bin/make -C $dir _depend ; done
Generating include/autoconf.mk
/bin/sh: line 3: arm-linux-gcc: command not found

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: add new Qemu x86 configuration file

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: add support for Mini 2440

Paul Jones documented at
http://pauljones.id.au/blog/2010/07/using-buildroot-on-a-mini2440/ how
to use Buildroot to generate a system for the FriendlyARM Mini2440
platform. This patch integrates Paul's work into Buildroot.

Unfortunately, the kernel being 2.6.32, we can't easily use a minimal
defconfig here. The mini2440 support has been merged into more recent
kernels, but I don't have the hardware to test.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: add Qemu Versatile defconfig

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove the few remaining unused Atmel-specific bits

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: remove Valka v100sc2

This default configuration did not even build a kernel image, which is
the main point of having board default configuration. So remove it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: remove atstk1005_defconfig

Neither the kernel nor U-Boot have support for a 1005 board, so let's
get rid of this board configuration, the corresponding target
skeleton, kernel and busybox configuration and device table.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minimize atstk100x_defconfig

Use recent U-Boot and kernel version, remove target skeleton and use
the default one instead, remove useless kernel configuration, busybox
configuration and device table.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: remove support for at91sam9260pf

It was supposed to be the support for AT91SAM9260 using a parallel flash
(instead of the usual dataflash). But the provided U-Boot configuration
at91sam9260pf.h was not used anywhere, and it was unsufficient to add
correct support in U-Boot (some changes in U-Boot Makefile would also be
needed). Additionnally, this configuration has not been merged into U-Boot
upstream since 2007 (when it was added to Buildroot).

Therefore, let's get rid of this configuration. If some users are
interested, we can re-introduce it properly with their help.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minify at91sam9260dfc_defconfig

Use recent U-Boot and kernel versions, remove useless kernel
configuration file.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minify at91rm9200df_defconfig

Use recent U-Boot and kernel versions, remove useless kernel
configuration file.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: simplify kb9202 board

Minimize the board defconfig, remove custom busybox configuration,
custom kernel configuration (use the kernel defconfig instead), custom
device table and target skeleton. The only difference in the target
skeleton was the support of mdev and the usage of an automount
script. Instead of adding this in a board-specific way, we should
provide board-independent configuration options. There are already
patches contributed to add support for mdev.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: simplify atngw100 support

Minimize atngw100_defconfig, remove atngw100-base_defconfig, and
remove the target skeleton and device table. Instead of having
complete copies of new target skeletons (making them hard to
maintain), we should just have a post-build script that
adds/removes/tweaks the existing target skeleton.

Moreover, most of the tweaks in this target skeleton were for specific
packages, but the policy now is that board defconfig should just build
a basic root filesystem with Busybox, and let the user select
whichever set of packages (s)he wants.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Minify arm_nptl_toolchain_defconfig

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Minify arm_toolchain_defconfig

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: remove i386/i686_defconfig

Those are not associated with any specific hardware system (PC or
another i386 system). Moreover, the fact that those configurations
require the build of a JFFS2 filesystem, very uncommon on PC systems,
seems to indicate that those configurations are not really being used
today.

It would make more sense to have a qemu_i388_defconfig (building a
kernel with just the device drivers for Qemu) and possibly a
pc_i386_defconfig (building a kernel with many device drivers, and a
bootloader such as Grub or Grub 2).

We also remove the corresponding kernel configuration files.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove useless kernel config files for integrator926

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: remove integrator926_huge_defconfig

One config per board is enough.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: minify integrator926_defconfig

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

barebox: bump Barebox to 2010.11

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

u-boot: add U-Boot 2010.09

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

dependencies: add unzip

unzip is needed to uncompress at91bootstrap, so let's add it as a
dependency. This can be discussed, and if we think it shouldn't be a
new dependency, we can also build a host-unzip.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

at91bootstrap: no board name check when 'make source' is used

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minify at91sam9g20dfc_defconfig

Use modern U-Boot and kernel versions, get rid of the now unused
kernel configuration file since we use the kernel defconfig instead.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minify at91sam9261ek_defconfig

Use modern U-Boot and kernel versions, and get rid of the useless
kernel configuration file, since we now use the kernel defconfig.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

board: update and minify at91sam9263ek_defconfig

Use modern kernel and U-Boot versions, and get rid of the now useless
kernel configuration file since we use the kernel defconfig file
instead.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

at91bootstrap: rename config options for consistency

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

at91bootstrap: switch to the official version

The Buildroot makefile was fetching and building the special
AT91Bootstrap of Ulf, which is not the Atmel official version. While
Ulf's variant has a better configuration/build system, the Atmel
version, as officially supported, is probably a better choice for the
future.

The Atmel version only needed a small tweak to work with EABI
toolchains.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

documentation: Update to explain how board support works

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Re-add target/generic/Config.in options to menuconfig

The options to customize the hostname, the banner and the serial port
configuration are now inside a menu named 'System configuration'.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove crazy set of empty directories

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove Busybox skeleton from target/device/Atmel

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove all Config.in/Makefile.in files in target/device

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Stop referencing the crazy target/device stuff

We don't need Config.in and Makefile in target/device: defconfig files
are sufficient to describe the specificities of a board (architecture,
compilation flags, bootloader and kernel details, etc.).

However, a placeholder such as target/device will be kept in order to
host things such as kernel configuration files or various
board-specific patches.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove mini device table, it isn't used

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Use the new BR2_ROOTFS_SKELETON_CUSTOM options

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Move Config.in.mirrors out of target/device

Having Config.in.mirrors (which also to select various download sites)
inside target/device sounds strange. This commit moves the contents of
Config.in.mirrors directly into the main Config.in file.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove useless target/device/mips

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove completely useless set of Makefiles specific to Atmel

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Update defconfigs after Atmel mirror related options removal

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove unused Atmel-specific mirror definitions

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

defconfigs: keep only one defconfig per board

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

CHANGES: #561/#1651/#1735 are resolved

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

l2tp: convert to gentargets and change to xl2tp

l2tp is no longer being developed, and xl2tpd is forked from l2tpd and is
maintained by Xelerance Corporation.

[Peter: remove invalid pppd dependency]
Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libtheora: bump version to 1.1.1

This allows to disable the build of the format specification
documentation, which was causing issues in cross-compiling mode.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mplayer: convert to gentargets, bump to a recent SVN version

Based on the conversion to gentargets done by Martin Banky, several
issues were fixed, and the mplayer package was improved:

 * Updated to a recent SVN version

 * Removed mandatory dependency on libmad

 * The AVR32 specific patch has been removed. It was a pain to remain
   blocked at the old 1.0-rc2 just for this patch. All this
   optimization work should have been submitted upstream, Buildroot is
   not the place to carry such heavy modifications.

 * Options were added to select whether mplayer and/or mencoder should
   be built/installed.

 * Support for additional options if packages have been selected:
   tremor, libmad, libtheora, libpng, jpeg, xlib_libX11,
   xlib_libXv. More could be added in the future.

 * Support for ARM-related optimization options. Similar improvements
   could be done for x86 and PowerPC architectures.

[Peter: fix build with !IPV6]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libaio: add patch to support AVR32

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

netplug: convert to gentargets and bump to 1.2.9.2

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

alsa-utils: convert to autotargets and bump to 1.0.23

[ Thomas: remove all patches, since they have been merged. Implement
  the target uninstallation step, fix many issues. ]

[Peter: fix target install / uninstall steps]
Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

hotplug: remove package

The package has been marked deprecated, is no longer developed
upstream, so let's get rid of it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libfloat: remove package

This package has been marked deprecated in Buildroot last year, no-one
complained, and it isn't developed upstream anymore. So get rid of it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ng-spice-rework: remove package

This package has been marked deprecated in October, and no-one
complained. Moreover, it doesn't sound like a package that is widely
used on embedded systems. So we get rid of it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

microcom: remove package

This package has been marked deprecated some time ago, because it has
been integrated into Busybox. Let's get rid of it now.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

memtester: convert to gentargets and bump to 4.2.1

Also, memtester requires largefile support

[ Thomas: add patch to fix memtester's Makefile at installation
  stage. Bump to 4.2.1. ]

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

lvm2: convert to autotargets and bump to 2.02.78

[ Thomas: simplifications and bump to 2.02.78 ]

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

lsof: convert to gentargets and bump to 4.84

lsof is packaged as a tarball inside a .tar.bz2 package, so we have a
post-extract hook to extract the source code tarball, so that for the
next steps, lsof looks like a normal package.

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

target: support generic getty with sysvinit inittabs

In sysvinit inittab the "id" field (first field) must be no longer
than 4 bytes, and is not used by init to determine the output
terminal. Therefore, we adjust the strategy used to modify the inittab
file according to the getty configuration.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

sysvint: convert to gentargets and bump to 2.88

Also, changed the site to Debian, to get the latest patches

[ Thomas: add inittab compatible with sysvinit. Minor fixes. ]

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

sysklogd: convert to gentargets and bump to 1.5

Also, changed the site to Debian, to get the latest patches

[ Thomas: Minor fixes. Removed S25sysklogd since our default
  /etc/inittab already starts klogd and sysklogd. ]

[Peter: only install syslog.conf if not available in skeleton]
Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mrouted: convet to gentargets and bump to version 3.9.4

mrouted is once again being actively developed.
See http://github.com/troglobit/mrouted

[ Thomas: bump to 3.9.4, patches updated, prefix fix. ]

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mii-diag: convert to gentargets and bump to 2.11.3

[ Thomas: minor fixes. ]

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

input-tools: fix jscal build, ensure target dir exists

And at the same time only build enabled applets.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

input-tools: convert to gentargets

Signed-off-by: Martin Banky <Martin.Banky@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

package: pass TARGET_MAKE_ENV at install/clean/uninstall steps

TARGET_MAKE_ENV contains the PATH to the host tools built in
$(HOST_DIR), and not having this directory in the PATH while
installing packages may cause troubles.

The specific example that trigerred the problem is the following
one. libglib2 ./configure tries to find a awk version, either gawk or
mawk. On my limited testing system, only mawk is available, so
normally, libglib2 detects mawk and uses it, and it works perfectly.

However, when host-gawk has been compiled before libglib2 (which
happens for example if you have used the Crosstool-NG backend), gawk
is well detected in $(HOST_DIR)/usr/bin at ./configure time
($(HOST_DIR)/usr/bin is in the PATH at ./configure time, thanks to
TARGET_MAKE_ENV). But at install time, when libglib2 uses gawk, it
doesn't find it anymore because it is not in the PATH.

This also makes things coherent with host installation, in which
$(HOST_MAKE_ENV) was already used.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: more verbose in crosstool-ng backend

By default, Crosstool-NG hides all the details of the toolchain build
process. However, this is inconsistent with the typical Buildroot
build, and the rotating progress thing makes log file resulting from
redirecting the Buildroot output a bit unusable. Therefore, switch the
Crosstool-NG configuration to a more verbose default.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: expose thread options in the Crosstool-NG backend

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: BR2_NEEDS_GETTEXT must be y when uClibc is used

BR2_NEEDS_GETTEXT was already set to y for Buildroot internal
toolchain and external toolchains based on uClibc, but wasn't set for
toolchains based on uClibC built by the Crosstool-NG backend.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

gcc: disable libgomp when thread support is disabled

libgomp needs thread support on the target to build properly.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: check that the thread option selection is correct

When an external toolchain is used, the user is responsible for
telling Buildroot what the configuration of the toolchain is. In order
to make sure that the selection is correct, we already perform checks
against the configuration of the uClibc in the selected external
toolchain. This commit adds the check for thread support.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: rework thread options

The selection of linuxthreads, linuxthreads old or NPTL doesn't make a
lot of sense for external toolchains. So, instead, we :

 * Introduce an hidden BR2_TOOLCHAIN_HAS_THREADS option, which must be
   selected by toolchain specific options when thread support is
   available. Package needing to test thread support should use this
   option.

 * Move the none/linuxthreads/linuxthreads old/NPTL selection to
   Buildroot internal toolchain configuration.

 * Add an option in external toolchain to tell if thread support is
   available or not in the external toolchain. We assume that glibc
   without threads is not possible, as Ulrich Drepper said in
   http://sourceware.org/ml/libc-alpha/2005-08/msg00091.html

ffmpeg, dmalloc and openvpn are fixed to use the new
BR2_TOOLCHAIN_HAS_THREADS option. For openvpn, --enable-threads=posix
is no longer used, as the configure script doesn't even understand
this option.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ltp-testsuite: bump version and use autotargets

Reworking ltp-testsuite is needed in order to get rid of some
thread-specific options that will be cleaned-up in the next commit.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: move Stack Protection Support option

The BR2_USE_SSP option is only used inside the uClibc build, so only
meaningful for Buildroot internal toolchains. Therefore, the option is
moved to the right location so that it isn't visible when working with
external toolchains.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: rework C++ options

Instead of having BR2_GCC_CROSS_CXX and BR2_INSTALL_LIBSTDCPP, with
BR2_GCC_CROSS_CXX not being visible (and therefore being useless),
let's just keep BR2_INSTALL_LIBSTDCPP to enable C++ in the toolchain
and install C++ libraries on the target.

We also take that opportunity to make BR2_INSTALL_LIBSTDCPP an hidden
option, which is selected by an option in Buildroot toolchain support
or an option in External toolchain support, just as we did for other
toolchain features.

Some work definitely remains to be done :

 - The name BR2_INSTALL_LIBSTDCPP is ugly, but we keep it for the
   moment in order to avoid changing all packages.

 - We should clarify the other language-related options (Fortran,
   Java, Objective-C, etc.).

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: remove toolchain-specific stripping

We already handle the stripping of libraries in $(TARGET_DIR) at the
global level, so there's no need to have toolchain-specific option and
code for this.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: add support for external toolchain profiles and download

Instead of letting the user define all the details of his external
toolchain, we define a set of profiles for well-known external
toolchains (CodeSourcery ones only at the moment, can easily be
extended with other toolchains).

Once a profile has been choosen, the user is offered the choice of
either letting Buildroot download and install the external toolchain,
or (as before) to tell Buildroot where the toolchain is installed on
the system.

We of course provide a "custom profile", through which the user can
configure Buildroot to use a custom external toolchain for which no
profile is available.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: remove ending semi-colon in helpers

Some helpers had their final line ending with a semi-colon, some did
not. For consistency, remove the final semi-colon from all helpers,
it's the responsability of the caller to add the final semi-colon as
needed.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: Improve C library option selection

Turn BR2_LARGEFILE, BR2_INET_IPV6, BR2_INET_RPC, BR2_USE_WCHAR,
BR2_ENABLE_LOCALE and BR2_PROGRAM_INVOCATION into hidden options.

Then, for Buildroot toolchains, external toolchains and Crosstool-NG
toolchains, provide visible options that selects the hidden options.

This allows :

 * To show a different label and help text in the case of Buildroot
   toolchain (do you want to enable feature X ?) and in the case of
   external toolchain (is feature X available in your toolchain ?)

 * To not show any option when a glibc external toolchain is selected
   (since glibc is assumed to support all of largefile, IPv6, RPC,
   WCHAR, locale and program invocation) and have them all selected in
   that case.

There is some amount of duplication between Buildroot toolchain config
options and Crosstool-NG toolchain config options, because kconfig
doesn't allow to source the same Config.in file twice (even if under
mutually exclusive conditions). This duplication is more readable that
the hack that consists in splitting files in multiple pieces.

However, this commit changes the name of the options visible in the
configuration interface, so existing .config files will have to be
updated accordingly.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>