diffutils : convert to autotools infrastructure

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * Installation to target is the default

Signed-off-by: cmchao <cmchao@gmail.com>

dash : convert to autotools infrastructure & bump to 0.5.5.1

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * No need to strip
 * Target installation is the default
 * No installation to staging is the default
 * Remove DASH_BINARY/DASH_SOURCE variables
 * Fix minor mistakes
 * Simplify: no need to pass CC/CC_FOR_BUILD

Signed-off-by: cmchao <cmchao@gmail.com>
merge dash

berkeleydb : convert to autotools infrastructure & bump to 4.4.20

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * Installation to target is the default
 * Use BERKELEYDB_SUBDIR to get rid of the specific BUILD_CMDS and
   INSTALL_STAGING_CMDS
 * Split the INSTALL_TARGET_CMDS to install development files on the
   target only when needed

Signed-off-by: cmchao <cmchao@gmail.com>

dialog : convert to autotools infrastructure & bump to 1.1-20100428

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * Installation to target is the default behaviour
 * No installation to staging is the default behaviour
 * Stripping is done globally

Signed-off-by: cmchao <cmchao@gmail.com>

links : convert to autotools infrastructure & bump to 1.01pre1-no-ssl

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * Installation to target is the default behaviour
 * No installation to staging is the default behaviour
 * Stripping is done globally
 * Remove incorrect cleaning rules

Signed-off-by: cmchao <cmchao@gmail.com>

fbv : convert to autotools infrastructure

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 * Target installation is the default behaviour
 * No installation to staging is the default behaviour
 * Remove FBV_BINARY/FBV_TARGET_BINARY variables

Signed-off-by: cmchao <cmchao@gmail.com>

dmalloc : convert to autotools infrastructure

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

     * Target installation is the default behaviour
     * No installation to staging is the default behaviour
     * Remove DMALLOC_TARGET_BINARY variable

Signed-off-by: cmchao <cmchao@gmail.com>

distcc : convert to autotools infrastructure & bump to 1.1-20100428

Changes by Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

     * Target installation is the default behaviour
     * No installation to staging is the default behaviour
     * Remove DISTCC_BINARY/DISTCC_TARGET_BINARY variables

Signed-off-by: cmchao <cmchao@gmail.com>

Makefile.autotools.in fix typo (s/extrators/extractors/)

Noticed by Michael S. Zick.

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

Makefile.package.in: Add support for *.tbz2 package archives

Signed-off-by: Paul Jones <paul@pauljones.id.au>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Fix u-boot 2010.03 build for AVR32 targets

Signed-off-by: Thiago A. Correa <thiago.correa@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

target/device: misc cleanup

Remove outdated and/or unused variables/files.

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

integrator926: use standard uclibc config

I don't see any reason to use a custom config here.

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

defconfigs: use the default BR2_JLEVEL setting (2)

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

module-init-tools: don't build documentation

We don't have docbook2man in buildroot, and the build breaks if you
have it on the host, but in an incompatible version (like on my machine).

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

Merge branch 'linux-cleanup' of git://git.busybox.net/~tpetazzoni/git/buildroot

Conflicts:
target/Makefile.in

Remove now unused Config.in for kernel selection

It's now linux/Config.in that allows to configure the kernel
configuration/compilation.

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

linux: add support for initramfs

In Buildroot, the kernel is built and installed *before* the root
filesystems are built. This allows the root filesystem to correctly
contain the kernel modules that have been installed.

However, in the initramfs case, the root filesystem is part of the
kernel. Therefore, the kernel should be built *after* the root
filesystem (which, in the initramfs case simply builds a text file
listing all files/directories/devices/symlinks that should be part of
the initramfs). However, this isn't possible as the initramfs text
file would lack all kernel modules.

So, the solution choosen here is to keep the normal order: kernel is
built before the root filesystem is generated, and to add a little
quirk to retrigger a kernel compilation after the root filesystem
generation.

To do so, we add a ROOTFS_$(FSTYPE)_POST_TARGETS variable to the
fs/common.mk infrastructure. This allows individual filesystems to set
a target name that we should depend on *after* generating the root
filesystem itself (contrary to normal ROOTFS_$(FSTYPE)_DEPENDENCIES,
on which we depend *before* generating the root filesystem).

The initramfs code in fs/initramfs/initramfs.mk uses this to add a
dependency on 'linux26-rebuild-with-initramfs'.

In linux/linux.mk, we do various things :

 * If BR2_TARGET_ROOTFS_INITRAMFS is enabled (i.e if initramfs is
   enabled as a root filesystem type), then we create an empty
   rootfs.initramfs file (remember that at this point, the root
   filesystem hasn't been generated) and we adjust the kernel
   configuration to include an initramfs. Of course, in the initial
   kernel build, this initramfs will be empty.

 * In the linux26-rebuild-with-initramfs target, we retrigger a
   compilation of the kernel image, after removing the initramfs in
   the kernel sources to make sure it gets properly rebuilt (we've
   experienced cases were modifying the rootfs.initramfs file wouldn't
   retrigger the generation of the initramfs at the kernel level).

This is fairly quirky, but initramfs really is a special case, so in
one way or another, we need a little quirk to solve its specialness.

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

linux: add support for linux26-{menuconfig,xconfig,gconfig} targets

These targets allow the user to customize the configuration of the
Linux kernel. After changing the kernel configuration, the next time
the user runs "make", the kernel is rebuilt to take into account the
new configuration (not rebuilt from scratch).

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

linux: adjust kernel config according to the Buildroot configuration

We only adjust the configuration or ARM EABI and IPv6. The (more
complicated) initramfs case is handled in a separate commit. The user
is expected to take care of all other configuration details (like
having Netfilter enabled to make iptables work, etc.)

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

Add generic functions to enable/set/disable options in kconfig files

The KCONFIG_ENABLE_OPT, KCONFIG_SET_OPT and KCONFIG_DISABLE_OPT are
new make functions to respectively enable, set and disable options in
Kconfig-like files (as used by the kernel, uClibc or Busybox).

They can be used as follows :

 $(call KCONFIG_ENABLE_OPT,CONFIG_FOOBAR,/path/to/.config)
 $(call KCONFIG_SET_OPT,CONFIG_BARFOO,foobar,/path/to/.config)
 $(call KCONFIG_DISABLE_OPT,CONFIG_FARBOO,/path/to/.config)

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

linux: Add dependency on host-module-init-tools

In order to not depend on module init tools being installed on the
development environment of the Buildroot user, let's build module init
tools for the host.

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

module-init-tools: bump version + convert to autotools

The "remove-index" patch is no longer needed, the newer version of
module-init-tools correctly uses strchr().

The "module-init-tools" patch, whose main purpose was to disable the
generation of man pages (it requires docbook2man), is replaced by a
simpler implementation,
module-init-tools-3.11-add-manpages-config-option.patch.

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

module-init-tools: remove support for cross-depmod

The new infrastructure to build the Linux kernel doesn't use a
cross-depmod. This isn't needed with modern module-init-tools that
most distros include perfectly handle this.

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

iso9660: take into account the linux changes

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

Remove old Linux infrastructure

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

New, simpler, infrastructure for building the Linux kernel

This patch introduces a single, simple, infrastructure to build the
Linux kernel. The configuration is limited to :

 * Kernel version: a fixed recent stable version, same as kernel
   headers version (for internal toolchains only), custom stable
   version, or custom tarball URL

 * Kernel patch: either a local file, directory or an URL

 * Kernel configuration: either the name of a defconfig or the
   location of a custom configuration file

 * Kernel image: either uImage, bzImage, zImage or vmlinux.

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

target: get rid of unused BR2_COPYTO / BR2_TARGET_ATMEL_COPYTO

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

package: remove games/Config.in

Unused since the package menu restructuring.

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

docs: Document location of .config for out-of-tree builds

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

docs: fix git-over-http description

This has now finally been setup.

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

Config.in.mirrors: Use sources.buildroot.net as default backup site

The buildroot/busybox/uClibc VM is running low on disk space, and we've
been asked to move the source mirrors off-site.

A redirect has been setup between the old buildroot.net/downloads/sources/
and sources.buildroot.net, so old .configs continue to work, but we might
as well use the official one now.

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

check-host-sed.sh: use mktemp for temporary file

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

check-lxdialog.sh: use mktemp for temporary file

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

package/config: create .depend in object directory

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

kconfig: build in BUILD_DIR/buildroot-config

This way the main buildroot dir can be completely read-only for
out-of-tree builds

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

Makefile: fix make <blah>_defconfig

Make sure we only pass O=<path> if we were originally called like that,
otherwise we might end up looking for the .config in the wrong place.

Likewise, when changing between using O= or not (or simply running
make <blah>_defconfig in a pristine source tree), we do end up having
a .config, but no .config.cmd. It isn't really critical so just continue
even if we couldn't read it.

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

cleanup/simplify kconfig changes for O= builds

No functional change.

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

Move config files into output directory for out-of-tree build

Closes #1213

Signed-off-by: Will Wagner <will_wagner@carallon.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

CHANGES: #835 is closed

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

ncurses: bump to 5.7

Closes #2038

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

uClibc: 0.9.31.config: enable UCLIBC_SV4_DEPRECATED option

Needed by xfsprogs.

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

ntfsprogs: fix fuse dependency

The ntfsprogs configure script would add hardcoded host dirs to the
pkg-config search path, confusing configure if libraries were installed
on the host and not on the target.

At the same time, ensure that fuse is build before ntfsprogs (if enabled),
so the fuse support gets built.

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

ntfsprogs: disable crypto support

The ntfsprogs crypto support requires libgnutls, which we don't have
in BR, and the autodetection in configure gets confused when the host
has a libcrypt-config.

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

squashfs: fix build with uClibc

squashfs used the get_nprocs() GNU extension, instead of directly
calling sysconf(), breaking the build on uClibc.

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

lrzsz: convert to autotools infrastructure

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

libart: fix staging install

The libart package had a couple of typos so it never really installed to
staging. So fix it.

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain: remove gcc 4.1.2 and non-sysroot support

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

toolchain/gcc: Use 4.3.x like for 4.4, get rid of older 4.3.x versions

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

toolchain: bump gcc 4.3.x series to 4.3.5

Closes #2065

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Fix potentially conflicting build flags

Change the default target optimisation value so
it does not conflict with gcc optimization level

Signed-off-by: Paul Jones <paul@pauljones.id.au>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ffmpeg: add user-defined configure parameters

Closes #401

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ffmpeg: allow customization of codecs, (de)muxers and other components

Add the option to customize the list of decoders, encoders, muxers, demuxers,
parsers, protocols, bsfs and filters to be built into ffmpeg, and to compile or
exclude input and output devices.

[Peter: fix qstrip invocation]
Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ffmpeg: add commandline programs

Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ffmpeg: add new package

[Peter: configure doesn't support --disable-nls / --disable-largefile,
 needs largefile support in toolchain]
Signed-off-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

at91dataflashboot: fix patch target

Touch destination file so the patch target isn't rerun for every
invocation.

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

at91bootstrap: add help text

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

CHANGES: update with changes from Thomas' bootloaders-cleanup branch

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

at91bootstrap: don't depend on BR .config

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

Merge branch 'bootloaders-cleanup' of git://git.busybox.net/~tpetazzoni/git/buildroot

busybox: add 1.16.2 devmem fix for access to the highest page

From busybox git.

Devmem used to always map 2 pages in case the access would cross a page
boundary, which fails if you try to access the highest page.

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

busybox: bump 1.16.x version

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

ipsec-tools: bump version, convert to autotools format and simplify Config.in

Closes #2035

- Remove -Werror from CFLAGS
- Always enable ipv6 support if available in toolchain
- Always install into staging

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

vpnc: remove unused legacy patches

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

linux: fix dependency on mkimage

After the rework of the U-Boot configuration/compilation process, we
need to slightly rework how target/linux/Makefile.in.advanced depends
on mkimage on the host to produce an uImage.

target/linux/Makefile.in doesn't need to be fixed as it just doesn't
handle this dependency for the moment.

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

Add support for the Barebox bootloader

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

at91dataflashboot: add two patches

The first patch allows to build with an EABI toolchain. The second
patch disables the install step of at91dataflashboot Makefile since it
tries to copy files to /tftpboot.

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

at91dataflashboot: add a patch step

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

at91dataflashboot: various fixes

Fix the URL, fix the directory to which it gets unpacked, fix make
target names to be lower-case only.

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

at91dataflashboot: fix configuration options

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

at91bootstrap: fix AT91BOOTSTRAP_BINARY

Use AT91BOOTSTRAP_BOARD instead of BOARD_NAME. Remove
AT91BOOTSTRAP_VERSION from the final binary image name.

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

at91bootstrap: fix install procedure

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

at91bootstrap: don't use the ATMEL_MIRROR thing

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

at91bootstrap: bump version to 2.13

At the same time, remove the unused AT91BOOTSTRAP_PATCH_LEVEL and
AT91BOOTSTRAP_PATCHED_VERSION variables.

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

at91bootstrap: add a check to verify that BR2_TARGET_AT91BOOTSTRAP_BOARD is set

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

at91bootstrap: remove unused/useless code

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

at91bootstrap: fix incorrect config option comment

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

at91bootstrap: fix the boot source selection

The boot source configuration options were depending on U-Boot
configuration options. Let's make it independent and just allow the
user to select whichever boot source is appropriate.

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

at91bootstrap: simplify configuration

There no need to depends on BR2_TARGET_AT91BOOTSTRAP when the
configuration options are already inside a if
BR2_TARGET_AT91BOOTSTRAP.

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

at91bootstrap: do not default to $(BOARD_NAME)

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

syslinux: depends on host-nasm

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

nasm: new package, needed for syslinux

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

syslinux: remove the unnecessary SYSLINUX_SUPPORTED_ARCH check

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

syslinux: bump version and fix build

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

grub2: only available on x86 and PowerPC and mark as BROKEN

The build process of grub2 breaks the compilation. It breaks with:

./configure: line 4766: syntax error near unexpected token `external'
./configure: line 4766: `AM_GNU_GETTEXT(external)'

In addition to this, it later requires Ruby. Do we really want to make
Buildroot depend on Ruby being installed on the host ? Do we really
want to build our own Ruby ? Do we even care about Grub2 ?

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

grub2: remove non-existing options

Much of the grub2.mk seems to have been copy/pasted from
grub.mk. However, all the network/splashimage related ./configure
options do not exist in grub2.

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

grub2: fix build

grub2 now builds fine, but some work remains to make it usable. What
should be installed exactly in the TARGET_DIR ? What is the
installation procedure and what should Buildroot do ?

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

grub2: bump version

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

grub/grub2: update splash image location

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

grub/grub2: drop unnecessary dependency on uclibc

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

grub2: remove GRUB2_ARCH_SUPPORTED

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

grub: remove GRUB_SUPPORTED_ARCH

grub can already only be selected for x86 and x86_64. No need to check
again for this in grub.mk.

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

defconfigs: remove references to old U-Boot versions

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

u-boot: use BR2_TARGET_UBOOT_VERSION instead of BR2_UBOOT_VERSION

In order to be coherent with all other BR2_TARGET_UBOOT_* options.

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

u-boot: beautify option prompts

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

u-boot: re-add infrastructure to specify custom tarball

The bootloader being very specific to the hardware, being able to
build U-Boot from an arbitrary tarball available on the web might be
needed.

Therefore, for U-Boot, we provide two methods :

 * Get a given stable version from U-Boot official FTP server

 * Get an arbitrary tarball

This should hopefully satisfy most needs, without complicating too
much the U-Boot build procedure on Buildroot side.

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

u-boot: add a few comments

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

u-boot: remove unused U_BOOT_TARGET_TOOLS variable

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

u-boot: further simplify the configuration

Remove all the bootsource selection mechanism and the horribly
complicated BR2_TARGET_UBOOT_DEFAULT_ENV thing, which wanted to be
generic, but was in fact very AT91-specific.

Just keep things simple: we build U-Boot with the board configuration
file specified in BR2_TARGET_UBOOT_BOARDNAME.

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

u-boot: support building for mkimage

To build mkimage for the host (which is needed to build an uImage of
the kernel), it is not necessary to configure U-Boot, and therefore to
have a particular board selected.

Therefore, this commit:

 * Adds a verification at U-Boot configure step that a U-Boot board
   name has been defined

 * Sets a default U-Boot version if none has been specified, so that
   even when U-Boot isn't selected but we want to build mkimage for
   the host, a particular U-Boot version is picked.

 * Make the host mkimage target depend only on U-Boot being
   downloaded/extracted/patched, and the target mkimage/fw_printenv
   targets depend on U-Boot being fully configured.

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