debug: do not force installation of gdbserver

The BR2_ENABLE_DEBUG option selects the compilation and installation
of gdbserver on the target. This is a bit restrictive, especially for
external toolchains, which may already contain a gdbserver on the
target.

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

Added lcdproc package

[Peter: fix download URL, move to 'Hardware handling' section]
Signed-off-by: Simon Dawson <spdawson at gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Remove stray $ character from a bunch of init scripts

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

squashfs3: fix build with uClibc

The squashfs3 package uses the old get_nprocs() GNU extension which does not
exist in uClibc. This has already been fixed in newer squashfs releases
(>=4.0). The patch is similar to the one committed in
dfa2a513e1aa7f6ec47f630c5d31d98cd098f408

[Peter: add patch header]
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

new package: zeromq

[Peter: fix C++/LF deps, reformat help, install into staging]
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

util-linux: fix libmount build under uClibc

Patch taken from gentoo.

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

nbd: fix download URL

The old URL nowadays return a 404.

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

dhcp: fix download URL

Tarball is in a subdir.

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

libfuse: fix download URL

The old URL nowadays return a 404.

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

ext-toolchain: Automatically set -m64 if the architecture is x86_64

Without the -m64, choosing the x86_64 architecture with a Sourcery external
toolchain will result in a 32-bit rootfs.

Also simplified the help text to reflect this change.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

apply-patches.sh: remove any rejects before applying patches

[Peter: .rej files might be in subdirs, so just do find .. | xargs rm]
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Tested-by: Ludovic Desroches <ludovic.desroches@atmel.com>
with an armadeus_apf9328_defconfig build
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

gst-plugins-bad: bump version

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

gst-plugins-ugly: bump version

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

gst-plugins-good: bump version

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

gst-plugins-base: bump version

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

gstreamer: bump version

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

libxml2: ensure patches are applied in the correct order

Some of the patches depends on eachother, so rename them to ensure they
get applied in the correct order.

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

gcc: bump 4.4.x version

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

dnsmasq: bump to version 2.60

Bump to version 2.60 of dnsmasq with the new lua scripting option.
Switch LDFLAGS from MAKE_ENV to MAKE_OPT to make it work properly.

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

lua: fix non-shared library build

Building without BR2_PACKAGE_LUA_SHARED_LIBRARY failed because of
missing -ldl.  dl is used by lua's dynamic module loading.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

New package: parted

[Peter: needs largefile + wchar]
Signed-off-by: Nicolas Dechesne <n-dechesne@ti.com>
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libethumb: Fix automagic dependencies on libedbus and libexif

libethumb will determine during configuration phase whether to build the
optional libexif and libedbus modules. It will enable this modules if
libedbus or libexif are present on the target system. Therefore, we need
to add these packages as optional dependencies to libethumb.

Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Add target to print buildroot version

It's convenient to have a target to print the buildroot version, for use
in external scripts calling buildroot.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

freetype: security bump to version 2.4.9

Bump freetype to version 2.4.9 to fix:
CVE-2012-1126, CVE-2012-1127, CVE-2012-1128, CVE-2012-1129,
CVE-2012-1130, CVE-2012-1131, CVE-2012-1132, CVE-2012-1133,
CVE-2012-1134, CVE-2012-1135, CVE-2012-1136, CVE-2012-1137,
CVE-2012-1138, CVE-2012-1139, CVE-2012-1140, CVE-2012-1141,
CVE-2012-1142, CVE-2012-1143, and CVE 2012-1144.

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

atngw100_defconfig: drop atomic64_t patch as it is now upstream

Since 3.2.10.

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

linux: bump 3.2.x stable version

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

kernel-headers: bump 3.0.x / 3.2.x stable versions

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

avr32/defconfig: use current stable kernel instead of the old 2.6.35

Drop the patch for kernel >= 3.3

Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

pixman: bump version and compilation fix

- bump to 2.5 and change download location
- disable iwmmxt support for CPU's that don't have
  this feature

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

can-utils: bump version and change download location

SocketCAN repository moved from Berlios.de to gitorious

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

usbutils: bump version to 005

Bump usbutils version to 005 and also switch to a debian mirror since
it's still missing upstream.

[Peter: add comment about autoreconf, use snapshot.debian.org]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mpfr: bump to version 3.1.0-p6

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

alsa-lib: bump to version 1.0.25

[Peter: swapped static/shared test around]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

alsa-utils: bump to version 1.0.25

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

wipe: bump to version 2.3.1 and switch to sourceforge

The original version is gone, so switch to sourceforge wipe and bump to
version 2.3.1

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

file: bump to version 5.11

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

grep: bump to version 2.11

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

sysstat: bump to version 10.0.3

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

hdparm: bump to version 9.39

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

nbd: bump to version 3.0

[Peter: only disable fallocate for uClibc toolchains]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

libfuse: bump to version 2.8.7

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

libnl: bump to version 3.2.7

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

smartmontools: bump to version 5.42

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

ntfsprogs: remove deprecated

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

ntfs-3g: bump to version 2012.1.15

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

radvd: bump version to 1.8.5

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

util-linux: bump to version 2.20.1

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

mpd: autoreconf no longer needed

configure.ac patch got dropped with the 0.16.3 version bump.

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

mpd: bump to version 0.16.7

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

gmp: bump to version 5.0.4

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

squid: bump to version 3.1.19

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

dhcp: security bump to version 4.1-ESV-R4

Bump to version 4.1-ESV-R4 to fix CVE-2011-4539

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

libcurl: security bump to version 7.24.0

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

libxml2: add multiple security patches

Fixes for CVE-2011-1944, CVE-2011-2821, CVE-2011-2834, CVE-2011-3919,
CVE-2012-0841 and others from upstream.

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

libxslt: add multiple security patches

Add security patches for CVE-2011-1202 and CVE-2011-3970.

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

packages: remove redundant INSTALL_TARGET=YES declarations

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

packages: remove redundant INSTALL_STAGING=NO declarations

Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

packages: remove redundant AUTORECONF=NO declarations

[Peter: handle libgtk2 as well]
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

sh: Add additional ttySCs

There are a couple of Renesas SH devices with 8 serial ports used.

Signed-off-by: Phil Edworthy <phil.edworthy@renesas.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

bluez_utils: bump to 4.99

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

network-manager: remove double libnl dependency

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

xdriver_xf86-video-fbdev: bump to 0.4.2

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

feh: bump to 2.4

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

tiff: bump to 4.0.1

Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

minicom: bump version

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

New package minicom

[Peter: fix whitespace and deps (wchar, ncurses, only iconv if !locale)]
Signed-off-by: Simon Dawson <spdawson at gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

ccache: set COMPILERCHECK to 'none'

This allows ccache to re-use its cache contents even if the compiler
binary mtime has changed. It is the simplest approach to solve this
problem, and it works for the internal, external and crosstool-ng
toolchain backends.

Of course, it leaves the user responsible for invalidating the cache
when necessary, but there doesn't seem to be a real good solution that
allows both to: 1/ keep the cache contents accross builds and re-use
it and 2/ invalidate the cache automatically when the compiler chances
in an incompatible way.

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

ccache: Force ccache to use its internal zlib

The problem is that without this, ccache would link against the zlib
of the build system, but we might build and install a different
version of zlib in $(O)/host afterwards, which ccache will pick
up. This might break if there is a version mismatch. A solution would
be to add host-zlib has a dependency of ccache, but it would require
tuning the zlib .mk file to use HOSTCC_NOCCACHE as the
compiler. Instead, we take the easy path: tell ccache to use its
internal copy of zlib, so that ccache has zero dependency besides the
C library.

Fixes bug #4808.

Thanks to Raúl Sánchez Siles <rsanchezs@infoglobal.es> for reporting
the bug and testing the proposed solution.

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

microperl: bump to version 5.12.4 and convert to gentargets

* Convert microperl to gentargets infrastructure
* Bump to a more modern version 5.12.4
* Introduce the bundle options to simplify people's lives

host-microperl is a fully-fledged perl.

For the time being we can't build XS modules thus breaking
target automake support for example since it requires IO.

target-automake was broken before anyway since the automake version
bump.

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

kmod: bump version

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

barebox: fix 2012.03 typo

Thanks to Eric Bénard for noticing.

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

toolchain-external: add linaro 2012.02

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

Merge branch 'for-2012.05/ext-toolchain-updates' of git://git.free-electrons.com/users/thomas-petazzoni/buildroot

gpsd: dbus interface requires dbus-glib

gpsd uses dbus-glib as the dbus interface, so it should only be built if
libglib2 has been selected.  To simplify things, build dbus support only
if dbus-glib is selected.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

barebox: add 2012.03 release

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

microperl: fix host build

It misses -lm when compiling miniperl

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

expat: add security patch for CVE-2009-3560

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

package: fix WGET download method

In cf2486bf317e4bbf88c801fb96183ba62be78cc8, we changed from using the
-P option of wget (to set the output *directory*) to using the -O
option (to set the output *file*). Unfortunately, wget -O has a
strange behaviour: it creates an empty 0-byte file even if the
download fails (for example when there is no network connection).

The problem is that then Buildroot thinks the download was successful
and therefore goes on with extracting the tarball.

The following succession of events makes Buildroot think that the
download has been sucessful:

 * Buildroot calls the DOWNLOAD_WGET macro with the URL of the
   official site

   * It tests if the file exists in the download directory, it doesn't
     exist.

   * It calls wget. wget fails to download the file and returns an
     error code, but leaves an empty file with the correct name in the
     downloaded directory.

 * Since the previously download failed, Buildroot tries another
   download from the Buildroot mirror (sources.buildroot.net)

   * It tests if the file exists in the download directory... and it
     exists! So this second download returns with success, and
     Buildroot assumes the file has been downloaded properly.

This scenario brings us with the following result, where the download
fails, but Buildroot continues its execution and tries to extract the
tarball:

$ rm /opt/dl/glib-2.30.2.tar.bz2
rm: cannot remove `/opt/dl/glib-2.30.2.tar.bz2': No such file or directory
$ rm -rf build/host-libglib2-2.30.2/
$ make
make -C /home/thomas/projets/buildroot O=/opt/outputs/udisks/.
>>> host-libglib2 2.30.2 Downloading
--2012-03-03 12:06:25--  http://ftp.gnome.org/pub/gnome/sources/glib/2.30/glib-2.30.2.tar.bz2
Resolving ftp.gnome.org... failed: Name or service not known.
wget: unable to resolve host address `ftp.gnome.org'
>>> host-libglib2 2.30.2 Extracting
bzcat /opt/dl//glib-2.30.2.tar.bz2 | tar --strip-components=1 -C /opt/outputs/udisks/build/host-libglib2-2.30.2  -xf -
bzcat: Compressed file ends unexpectedly;
perhaps it is corrupted?  *Possible* reason follows.
[...]
tar: This does not look like a tar archive
tar: Exiting with failure status due to previous errors
make[1]: *** [/opt/outputs/udisks/build/host-libglib2-2.30.2/.stamp_extracted] Error 2
make: *** [all] Error 2
$ ls -l /opt/dl/glib-2.30.2.tar.bz2
-rw-r--r-- 1 thomas thomas 0 Mar  3 12:12 /opt/dl/glib-2.30.2.tar.bz2

Therefore, this commit modifies DOWNLOAD_WGET so that it removes the
downloaded file if wget returns with an error.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

at91sam9260dfc_defconfig: fix kernel defconfig name

at91sam9260_defconfig contains support for the EVM (since v3.2).

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

lite: Remove excess slashes from pkg-config file.

Signed-off-by: Will Newton <will.newton@imgtec.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

glib-networking: fix build without gnutls

configure errors out when it doesn't find gnutls unless explicitly
disabled.

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

atk: fix build after glib2 bump

Remove G_CONST_RETURN references, similar to how we did for gamin.

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

imagemagick: re-add xml2-config patch

The patch somehow got removed from eb5498 (bump version, fix build) by
accident, so add it again.

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

mtd: build after busybox if enabled

Busybox also provides flash applets nowadays, so ensure the mtd version
takes precedence if both are enabled.

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

gcc: bump 4.6.x version

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

imagemagick: bump version, fix build with zlib

The recent zlib bump broke imagemagick. This has been fixed upstream
in 6.7.5, but the xml2-config fix is still not upstream and 6.7.5
needs autoconf 2.67 to autoreconf (and we have 2.65), so we cannot
easily use that.

Instead move to the most recent version using autoconf 2.64 and
backport the fix from imagemagick svn. At the same time also
ensure zlib+bzip2 support is picked up if enabled.

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

external-toolchain: add support for Linaro 2012.01

Linaro has just released a new pre-built toolchain, available as a
tarball, which is a pure toolchain (only the C library is
included). This makes this new Linaro 2012.01 toolchain usable in
Buildroot, so let's integrate the support for it.

In addition to simply adding the new external toolchain at the usual
locations, this patch allows need to adapt a few things to support
Linaro toolchains. Most toolchains store their libraries in the "lib/"
or "usr/lib" directories relative to the toolchain. Buildroot
toolchains on the other hand, store the libraries in the
"usr/<target-name>/lib" directory. And the Linaro toolchain has
choosen to use the "lib/<target-name>" directory. Therefore, this
patch adjust:

 * The logic to search a particular library when that library needs to
   be copied to the target directory

 * The logic to deduce the sysroot directory from the libc.a file
   location in the toolchain: removing "(usr/?)lib(64?)" is no longer
   sufficient, we need to take into account the "lib/<target-name>/"
   case.

Since the Linaro toolchain generates code for Cortex-A processors
only, the selection of this toolchain is limited to Cortex-A8 and
Cortex-A9.

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

external-toolchain: fix description of PowerPC 2010.09 from Mentor Graphics

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

external-toolchain: adjust check for uClibc static toolchains

The current check for uClibc toolchain was verifying that a
ld-uClibc.so dynamic loader was present. However, with static-only
uClibc toolchains, this does not work. Instead, we check for an
uClibc-specific header file in the sysroot.

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

external-toolchain: run checks even on extracted toolchains

When the mechanism that allows Buildroot to download external
toolchains automatically was added, all the sanity checks on the
external toolchains were not performed. This commit re-enables those
checks that we already do on external toolchains that are not
downloaded/extracted by Buildroot. This makes the toolchain checks
more consistent accross various configurations.

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

external-toolchain: details on selecting the multilib variants

Each multilib variant need to be selected using a special combination
of flags, requiring specific choices of the Buildroot options. This
commit documents those configuration choices to make it easier to use
the various multilib variants.

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

external-toolchain: Blackfin.uclinux.org 2011R1-RC4

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

external-toolchain: add Sourcery CodeBench x86/x86_64 2011.09

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

external-toolchain: add Sourcery CodeBench PowerPC 2011.03

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

external-toolchain: bump MIPS 2011.03 from -93 to -110

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

external-toolchain: add Sourcery CodeBench ARM 2011.09

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

external-toolchain: remove 2009q1 ARM Sourcery toolchain

We are going to add one more ARM Sourcery toolchain version, so it's
time to remove the oldest version.

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

Expand external toolchain wrapper to custom flags

The BR2_TARGET_OPTIMIZATION flags were not used by the external
toolchain wrapper, which broke the multilib selection logic of
multilib external toolchains. It also simplifies the compilation of
external programs since all flags are properly passed automatically by
the toolchain wrapper.

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

Support multilib variants in sub-subdirectories

When an external toolchain has multiple variants organized in
sub-directories, Buildroot only copies the selected sysroot and not
all sysroots. In order to make this work, Buildroot creates a symbolic
link of the name of the original selected sysroot to the main sysroot
to trick the compiler so that it finds its libraries at the expected
location.

I.e, if the toolchain as the following organization (example take on
the ARM CodeSourcery toolchain) :

     .      for ARMv5T
     armv4   for ARMv4T
     thumb2  for ARMv7-A/Thumb

and ARMv4T is selected, then Buildroot will copy the contents of
armv4t/ from the toolchain into its $(STAGING_DIR) and then create a
$(STAGING_DIR)/armv4t symbolic link to $(STAGING_DIR).

However, our logic to do so only works when there was one directory
level for multilib sysroots. But in the MIPS CodeSourcery toolchain
there are multiple levels. For example, the MIPS16 soft-float
little-endian sysroot variant is in mips16/soft-float/el/ compared to
the main sysroot.

This patch improves our logic to support this case. The logic is a bit
more complicated as we don't want to create a symbolic link to an
absolute path, but a symbolic link to a relative path, because we want
the host/ directory to be relocatable.

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

Improve external toolchain logic to support IA32 Sourcery CodeBench toolchain

The IA32 Sourcery CodeBench toolchain has a relatively special
structure, with the following multilib variants:

 * Intel Pentium 4, 32 bits, the multilib variant is in ./ relative to
   the main sysroot, with the libraries in the lib/ directory.

 * Intel Xeon Nocona, 64 bits, the multilib variant is in ./ relative
   to the main sysroot, with the libraries in the lib64/ directory.

 * Intel Atom 32 bits, the multilib variant is in atom/ relative to
   the main sysroot, with the libraries in the lib/ directory.

 * Intel Core 2 64 bits, the multilib variant is in core2/ relative to
   the main sysroot, with the libraries in lib64/ directory.

So the first two variants are in the same sysroot, only the name of
the directory for the libraries is different.

Therefore, we introduce a new ARCH_LIB_DIR variable, which contains
either 'lib' or 'lib64'. This variable is defined according to the
location of the libc.a file for the selected multilib variant, and is
then used when copying the libraries to the target and to the staging
directory.

In addition to this, we no longer use the -print-multi-directory to
get the ARCH_SUBDIR, since in the case of the 64 bits variants of this
toolchain, it returns just '64' and not a real path. Instead, we
simply compute the difference between the arch-specific sysroot and
the main sysroot.

We also take that opportunity to expand the documentation on the
meaning of the different variables.

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