Merge branch 'codesourcery-toolchain-support' of git://git.busybox.net/~tpetazzoni/git/buildroot

Add support to linux 2.6.30 kernel patches for Atmel arm in menuconfig.

Signed-off-by: Josh.Wu <josh.wu@atmel.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

busybox: add 1.16.1 sed fix

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

initramfs: fix init symlink

When initramfs was ported to the new fs structure the init symlink
macro was defined, but forgot to add it to PRE_GEN_HOOKS

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

busybox: get rid of unused initramfs.mk file

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

target: really remove the target/*/*.mk include

Commit 5faae4869027 claimed to do this, but it didn't - Causing
make to error out.

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

toolchain/binutils: Improve arm linking performance

Backported a patch from CVS that improves linking times for large
projects (eg 700s to 6s). See
http://old.nabble.com/-PATCH--Reduce-ARM-linking-time-tt27961038.html
for the original.

[Peter: add patch header]
Signed-off-by: Laine Walker-Avina <lwalkera@ieee.org>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Repair initramfs dependancies after move to fs

The restructure for building root filesystems changed the target name
for the initramfs file, to build the file the trget is now
initramfs-root but the generated file is rootfs.initramfs

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

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

mtd: Add missing 'mtdinfo'

Commit 9277cb760a4cc6e62d0cd23fa938dd953e875a13 added support
for mtdinfo to mtd.mk but missed corresponding Config.in stuff.

Signed-off-by: Michael Roth <mroth@nessie.de>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

makedevs: convert to a more normal way of building packages

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

makedevs: move to package/

Since target/ doesn't contain any .mk file in subdirectories, also
remove the include in target/Makefile.in.

python: needs WCHAR support in toolchain

And hence, portage needs it as well.

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

fs/ext2: fix blocks/inodes calculation

With the ROOTFS_TARGET conversion, EXT2_OPTS gets evaluated very early
(before TARGET_DIR is populated with files), so the calculated
blocks/inodes numbers are wrong.

Fix it by moving the calculation to a shell script wrapper around
genext2fs, so it only gets executed just before genext2fs runs.

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

Enable libutil by default for uclibc 0.9.31

Closes #1591

Signed-off-by: Kelvin Cheung <keguang.zhang@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

toolchain/gdb: target-gdb needs wchar support in toolchain

For the embedded readline/gettext stuff.

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

fs/ext2: fix typo

BR2_TARGET_ROOTFS_EXT2_RESBLKS is an int, so test against 0 rather than
the empty string - Otherwise the test is always true and a -m option
without arguments is added to the argument list, causing genext2fs to
get confused.

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

CHANGES: #1591 is now closed (portmap fails to compile)

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

CHANGES: #1537 is now closed (dev entries not created anymore)

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

Merge branch 'fs-fixes' of git://git.busybox.net/~tpetazzoni/git/buildroot

uclibc: Enable libresolv and libnsl stubs by default for 0.9.31

Closes #1555.

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>

Update defconfigs to include BR2_ROOTFS_DEVICE_TABLE definition

We only bother updating the defconfigs that need a non-default
BR2_ROOTFS_DEVICE_TABLE value.

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

Remove all TARGET_DEVICE_TABLE definitions

We have a special case for Xtensa, which was patching the generic
device_table.txt. Instead of doing this, we just keep a copy of the
device table, specific to Xtensa, with Xtensa specifities. The fact
that the patch wasn't applying anymore on the generic device table is
a sign that the existing approach wasn't working anyway.

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

documentation: update after TARGET_DEVICE_TABLE removal

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

cramfs/jffs2: use the new ROOTFS_DEVICE_TABLE variable

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

fs: change the way the device table is configured

Until now, the location of the device table was specified by a
variable in board Makefiles. Unfortunately, this variable is not
accessible from fs/common.mk, since the target/ code is included
*after* fs/common.mk.

Anyway, the general idea is to move away from these boards Makefile,
and provide configuration option for things like the device table
location.

Therefore, this patch adds a BR2_ROOTFS_DEVICE_TABLE option which
allows to specify which device table should be used.

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

cramfs: fix installation

The installation procedure of cramfs was broken when
$(HOST_DIR)/usr/bin didn't exist (i.e, cramfsck was installed as
$(HOST_DIR)/usr/bin directly). Use install -D with a proper
destination argument to make it work.

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

busybox: make sure architecture flags are used at link time

When compiling Busybox, according to readelf -A, all object files were
properly compiled according to the select architecture (-march=armv4t
for example), but the final busybox binary could be of a different
architecture (ARMv5t even if ARMv4t was selected).

This patch changes the way we configure/compile Busybox so that our
CFLAGS aren't anymore passed through the make EXTRA_CFLAGS variable,
but through the .config CONFIG_EXTRA_CFLAGS variable. Unfortunately,
those variables don't have exactly the same semantic for the Busybox
build system.

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

external-toolchain: Support for multilib toolchains

Multilib toolchains provide different versions of the base libraries
for different architecture variants. For example, the ARM Codesourcery
toolchain provides base libraries for ARMv5 (default), ARMv4t and
Thumb2.

Depending on the -march= argument passed to gcc, the sysroot used by
the compiler is therefore different. This means that the sysroot
location in CROSS-gcc -v cannot be used. Instead, we must use
CROSS-gcc -print-sysroot when available and fall back to the old way
if unavailable.

Moreover, we cannot simply copy the full sysroot as we used to do,
because the sysroot organization of multilib toolchain is more
complicated. In Codesourcery toolchains, we have :

 /
   etc      -- for ARMv5
   lib      -- for ARMv5
   sbin     -- for ARMv5
   usr      -- for ARMv5 (includes headers)
   armv4t
     etc    -- for ARMv4t
     lib    -- for ARMv4t
     sbin   -- for ARMv4t
     usr    -- for ARMv4t (no headers!)
   thumb2
     etc    -- for Thumb2
     lib    -- for Thumb2
     sbin   -- for Thumb2
     usr    -- for Thumb2 (no headers!)

So we have the default ARMv5 architecture variant that is installed in
the main directory, and we have subdirectories for the ARMv4t and
Thumb2 architecture variants.

Copying the full sysroot to the staging directory doesn't work. All
our packages are based on the fact that they should install libraries
in staging/usr/lib. But if ARMv4t is used, the compiler would only
look in staging/armv4t/usr/lib for libraries (even when overriding the
sysroot with the --sysroot option, the multilib compiler suffixes the
sysroot directory with the architecture variant if it matches a
recognized one).

Therefore, we have to copy only the sysroot that we are interested
in. This is rendered a little bit complicated by the fact that the
armv4t and thumb2 sysroot do not contain the headers since they are
shared with the armv5 sysroot.

So, this patch :

 * Modifies how we compute SYSROOT_DIR in order to use -print-sysroot
   if it exists. SYSROOT_DIR contains the location of the main sysroot
   directory, i.e the sysroot for the default architecture variant.

 * Defines ARCH_SUBDIR as the subdirectory in the main sysroot for the
   currently selected architecture variant (in our case, it can be
   ".", "armv4t" or "thumb2"). ARCH_SYSROOT_DIR is defined as the full
   path to the sysroot of the currently selected architecture variant.

 * Modifies copy_toolchain_lib_root (which copies a library to the
   target/ directory) so that libraries are taken from
   ARCH_SYSROOT_DIR instead of SYSROOT_DIR. This ensures that
   libraries for the correct architecture variant are properly copied
   to the target.

 * Modifies copy_toolchain_sysroot (which copies the sysroot to the
   staging/ directory), so that it copies the contents of
   ARCH_SYSROOT_DIR, and if needed, adds the headers from the main
   sysroot directory and a symbolic link (armv4t -> . or thumb2 -> .)
   to make the compiler believe that its sysroot is really in armv4t/
   or thumb2/.

Tested with Codesourcery 2009q1 ARM toolchain, Crosstool-NG ARM glibc
and ARM uClibc toolchains.

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

external-toolchain: support libraries outside of /lib

The copy_toolchain_lib_root function was making the assumption that
all libraries were stored inside the /lib directory of the sysroot
directory. However, this isn't true for certain toolchains,
particularly for the libstdc++ library.

The function is therefore reworked to find the library and its related
symlink either in /lib or /usr/lib in the sysroot, and copies it at
the same location in the target directory.

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

external toolchains: take into account architecture variant

Until now, many TARGET_CFLAGS where missing when using an external
toolchain, due to how package/Makefile.in was written. Now, a lot more
definitions are common between the Buildroot toolchain case and the
external toolchain case.

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

openssl: fix build without ipv6 support

Closes #1567

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

pciutils: convert to the generic-tools infrastructure

Signed-off-by: Thomas Brandstetter <thomas.brandstetter@me.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

mtd: host compilation needs host-zlib

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

mtd: mkfs.jffs2 requires zlib to build

[Peter: lzo needed as well]
Signed-off-by: Will Wagner <will_wagner@carallon.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

iso9660: menu.lst fixup after move to fs

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

initramfs: trivial fix following move to fs

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

busybox: 1.16.1 dnsd / hwclock fixes

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

libdrm: enable/disable features according to driver selection

Also cleanup autotools symbols and add libatomic_ops dependency
for Intel part.

Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Add libatomic_ops 1.2 needed by libdrm Intel support

Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>

Merge branch 'autotools-conversion' of git://git.busybox.net/~tpetazzoni/git/buildroot

libid3tag: convert to the autotools infrastructure

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

libmad: convert to the autotools infrastructure

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

aumix: convert to the autotools infrastructure

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

madplay: convert to the autotools infrastructure

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

findutils: convert to the autotools infrastructure

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

ipkg: remove useless configure options

These options are part of the default set of options passed to
./configure by the AUTOTARGETS infrastructure.

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

argus: convert to the autotools infrastructure

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

tar: convert to the autotools infrastructure

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

bison: convert to the autotools infrastructure

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

Remove incorrect 'dummy' example

Since the introduction of the AUTOTARGETS infrastructure, the dummy
example is no longer correct. Moreover, the documentation now contains
extensive details about how new packages should be added, either using
the GENTARGETS or the AUTOTARGETS infrastructures.

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

libcgi: convert to the autotools infrastructure

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

libsysfs: convert to the autotools infrastructure

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

libraw1394: convert to the autotools infrastructure

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

libcgicc: convert to the autotools infrastructure

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

libosip2: convert to the autotools infrastructure

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

patch: convert to the autotools infrastructure and bump version

Bumping the version from 2.5.9 to 2.6 allows to avoid adding a patch
(merged upstream) to fix patch's Makefile install target.

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

smartmontools: convert to the autotools infrastructure

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

setserial: convert to the autotools infrastructure

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

udpcast: convert to the autotools infrastructure

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

x11vnc: cleanup useless config options

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

vtun: convert to the autotools infrastructure

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

Update defconfigs after BR2_RECENT removal

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

Remove the BR2_RECENT option

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

Get rid of all usages of BR2_RECENT

Since BR2_RECENT was enabled by default, we do not want entries marked
BR2_RECENT (and thus appearing by default in Buildroot) to disappear.

Therefore, all the entries marked BR2_RECENT are converted as
non-deprecated. We can later decide, on a per-entry basis, to add
BR2_DEPRECATED to some of them. But at least, this commit doesn't
change the default current behaviour of Buildroot.

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

Get rid of the OPTIMIZE_FOR_CPU variable

This variable, together with the FIXME comment, has been added has
part of Eric Andersen's « Major buildroot facelift, step one » commit
that occured in October 2004.

Since then, no real usage has been made of OPTIMIZE_FOR_CPU, and the
initial intention has probably been lost in the memories of the
implementors.

Therefore, get rid of the variable, and just use $(ARCH) at the two
locations the variable was used.

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

Get rid of useless OpenMP related code

It seems that there was an intention to add BR2_ENABLE_OPENMP someday,
but it was in June 2007 (commit
c81807a9d71fba9f35eeb7e3f3b56bda4b2e0edd) and since then, nothing
occured. Therefore, get rid of this code, and just pass
--disable-openmp to gettext to keep the current behaviour.

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

Remove variables useless after BR2_PREFER_IMA removal

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

Get rid of unused variables

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

Update defconfigs after BR2_ENABLE_LOCALE_PREGENERATED removal

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

Remove BR2_ENABLE_LOCALE_PREGENERATED

The option is marked broken since october 2009, and even the uClibc
configuration help text says that using pregenerated locales is highly
discouraged.

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

Get rid of now unused HOST_ARCH computation code

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

Update defconfigs after BR2_GNU_BUILD_SUFFIX removal

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

Guess build system and remove BR2_GNU_BUILD_SUFFIX

Instead of having a configuration option BR2_GNU_BUILD_SUFFIX, let's
use config.guess to guess the build system type.

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

Move the global config cache option to an Advanced submenu

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

Bring the three path-related options together

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

Move the command-related options to a submenu

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

Update defconfigs after BR2_HAVE_{INFOPAGES,MANPAGES} removal

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

Get rid of BR2_HAVE_{MANPAGES,INFOPAGES} config options

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

Use BR2_HAVE_DOCUMENTATION instead of BR2_HAVE_{MANPAGES,INFOPAGES}

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

Do not remove documentation on a per-package basis

Since the documentation cleanup is done by the main Makefile in a
global way, there's no need to do that on a per-package basis in the
generic package infrastructure.

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

Remove all documentation when BR2_HAVE_DOCUMENTATION is not set

Until now, we had BR2_HAVE_MANPAGES, BR2_HAVE_INFOPAGES and
BR2_HAVE_DOCUMENTATION, but there's no need to have such a
fine-grained control over what documentation gets kept or not.

Therefore, just remove all the documentation (man pages, info pages
and documentation in $(TARGET_DIR)/usr/doc and
$(TARGET_DIR)/usr/share/doc) when BR2_HAVE_DOCUMENTATION is not set.

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

Do not let packages remove man pages, info pages and documentation

The cleanup of $(TARGET_DIR)/usr/share/man, $(TARGET_DIR)/usr/man,
$(TARGET_DIR)/usr/share/info, $(TARGET_DIR)/usr/info,
$(TARGET_DIR)/usr/share/doc and $(TARGET_DIR)/usr/doc is already done
globally in the main Makefile. Therefore, there's no need to handle
that on a per-package basis.

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

Update defconfigs after BR2_UPDATE_CONFIG removal

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

Update config.{guess,sub} unconditionally

Remove the BR2_UPDATE_CONFIG option which allowed to globally
enable/disable the replacement of config.{guess,sub} for
autotools-based packages. Doing it unconditionnally just works.

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

Update defconfig after BR2_PREFER_IMA removal

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

Remove BR2_PREFER_IMA

This option is barely used, no-one is maintaining it or extending
it. So let's remove it.

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

cramfs: fix help text typo

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

Merge branch 'squashfs3' of git://git.busybox.net/~tpetazzoni/git/buildroot

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

Add version selection for the SquashFS root filesystem

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

squashfs3: new package

Add support for the old squashfs 3.x tools, so that squashfs 3.x root
filesystems can be generated.

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

fs/*/Config.in: remove useless configuration comments

The dependency on gzip, bzip2 and lzma are properly handled
automatically. No need to tell the user about this.

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

Update defconfigs after fs cleanup

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

Coherent naming for .mk files in fs/

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

Move all filesystem generation code to fs/

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

Remove IMAGE and related configuration options

Now, we just hardcode the image filenames to be rootfs.$(FSTYPE), in
the $(BINARIES_DIR).

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

iso9660: rework in a similar way to ROOTFS_TARGET

Unfortunately, it cannot use the ROOTFS_TARGET infrastructure, due to
the specifities of the iso9660 build process.

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

target/common: do not use IMAGE

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

busybox initramfs: do not use the IMAGE variable

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