package: use <pkg>_CONFIG_SCRIPTS in packages that used special handling
The packages changed by this commit were not only changing prefix= and
exec_prefix= during their <foo>-config fixups, they were also changing
includedir= and/or libdir=. So, they could not be directly converted
to the new <pkg>_CONFIG_SCRIPTS infrastructure.
However, a careful analysis of their default <foo>-config shows that
includedir= and libdir= is defined relatively to either ${prefix} and
${exec_prefix}. Therefore, the manual fixing of includedir= and
libdir= is useless, and fixing prefix= and exec_prefix=, as done by
the <pkg>_CONFIG_SCRIPTS mechanism is sufficient.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Samuel Martin" <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
package: use <pkg>_CONFIG_SCRIPTS wherever possible
Use the <pkg>_CONFIG_SCRIPTS mechanism in all packages for which it
does all what the package was doing. A few packages, like libxslt, are
for now left out, since they need some additional fixup (for example a
fixup of includedir).
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
package: rename <pkg>_CONFIG_FIXUP to <pkg>_CONFIG_SCRIPTS
This commit renames the newly introduced <pkg>_CONFIG_FIXUP variable
to <pkg>_CONFIG_SCRIPTS, for two reasons:
* <pkg>_CONFIG_SCRIPTS will not only "fixup" the scripts in
$(STAGING_DIR)/usr/bin, but also remove them from
$(TARGET_DIR)/usr/bin. So it is not only about doing a "fixup".
* On the principle, it is strange that the variable carries an
indication of the action that will take place on those files. It
should rather be named to say "Here are the <foo>-config scripts",
and let the package infrastructure decide if it should fix them up,
remove them, etc.
This commit also updates the documentation accordingly.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Acked-by: "Samuel Martin" <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
libffi: add patch to fix library installation location
The libffi library was installed in a wrong directory when using the
non-default variant of a multilib toolchain. We fix this by patching
the libffi Makefile.am to install its library in a normal
location. The patch has been submitted upstream.
This makes it possible to put empty directories in the overlay.
Thanks to Aras Vaichas for pointing that out.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Since the post-image config option uses the plural for "Custom
scripts", do the same for the post-build config option, for
consistency.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
docs/manual: slightly reword the solutions to customize rootfs
The order of the solutions to customize the root filesystem is
changed: we now mention the post-build script mechanism *before* the
custom root filesystem skeleton mechanism, because the former is
preferred over the latter.
In addition to this, we give a few more details about direct
customization of the root filesystem in output/target, and about the
custom target skeleton solution.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Acked-by: Luca Ceresoli <luca@lucaceresoli.net> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
[Peter: fix minor typos] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Just like we have a post-build script mechanism that gets executed
after the build of all packages but before the creation of the
filesystem images, let's introduce a post-image script mechanism, that
gets executed once all filesystem images have been generated.
This can for example be used to call a tool building a firmware image
from different images generated by Buildroot, or automatically extract
the tarball root filesystem image into some location exported by NFS,
or any other custom action.
[Peter: fix image script check] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Tested-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Acked-by: Luca Ceresoli <luca@lucaceresoli.net> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Yann E. MORIN [Thu, 7 Feb 2013 09:04:48 +0000 (09:04 +0000)]
arch/arm: fix-up the ARM Kconfig warning
Kconfig does not accepts that a symbol that is part of a choice
be affected a default value.
Fix this by introducing a dummy EABI symbol, and make the real
EABI symbol a prompt-less option that depends on !OABI.
[Peter: drop arm dependency, rename to EABI_CHOICE] Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Peter Korsgaard <jacmet@uclibc.org> Cc: Arnout Vandecappelle <arnout@mind.be> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Contrary to what the .mk file states, proxychains-ng doesn't use autotools.
Instead it uses a handwritten configure script, so convert it to
generic-package.
This handwritten configure script generates config.mak which is then
included by the Makefile. Unfortunately config.mak does:
CC ?= $(TARGET_CC)
But as make sets CC to 'cc' by default, this never does anything - So
it ends up using the host compiler (and target CFLAGS), breaking the build.
Fix it by passing the correct CC setting at build time as well.
While we're at it, also cleanup the package by using 'make install-config'
to install the configuration file.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
configs/{ea3250, fdi3250, phy3250}: lock down kernel headers version
Update {ea3250,fdi3250,phy3250}_defconfig to lock down kernel headers to
an appropiate version.
Otherwise the target is building with, at the moment of this writing,
version 3.7 headers and a 2.6.34 kernel which usually isn't wise.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Ryan Mallon [Tue, 5 Feb 2013 07:00:05 +0000 (07:00 +0000)]
libsndfile: Add dependency on BR2_LARGEFILE
libsndfile requires sizeof(sf_count_t) == 8, where sf_count_t is a
typedef alias for off_t. This is not true by default for all
tool-chains, which leads to a runtime assert failure in binaries
compiled against libsndfile. See:
Add a dependency on BR2_LARGEFILE for libsndfile, and a comment if
BR2_LARGEFILE is not selected.
[Thomas: rebased patch, added more propagation of the new
BR2_LARGEFILE dependency to gst-plugins-good and mpd.]
Signed-off-by: Ryan Mallon <rmallon@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Store BR2_DEFCONFIG in .config, and use it to update the original input
defconfig file after updating the configuration. When a config is
created by using the BR2_DEFCONFIG=... option, this is saved in the
.config file; later runs of savedefconfig will update that same location.
It is also possible to configure this place in the interactive
configuration.
The BR2_DEFCONFIG value itself is not saved into the generated
defconfig, since Kconfig considers it at its default. This is
intentional, to avoid hard-coding an absolute path in the defconfig.
It will anyway be set again when the defconfig is used with the
'make BR2_DEFCONFIG=... defconfig' command.
As a side-effect of this change, the *config options have been moved out
of the BR2_HAVE_DOT_CONFIG condition. This doesn't make any functional
difference, because the .config is still not read for the *config targets.
However, the defconfig and savedefconfig targets do need to include
.config now, which makes them slightly slower.
[Peter: slightly tweak help text] Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Acked-by: Luca Ceresoli <luca@lucaceresoli.net> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The filesystem overlay is a tree that is copied over the target fs
after building everything - which is currently usually done in the
post-build script.
libffi: switch to using a Git version, brings Microblaze and Xtensa support
We already had backported patches for the Blackfin and AArch64
support, and now we would have needed to also backport the Xtensa
support and the Microblaze support. This starts to get crazy.
Let's switch to using a Git version from Github, until libffi finally
releases a new stable version.
In order to achieve this, we also need to:
* autoreconf the package, so that a libffi.pc file gets generated
* manually install the libffi.pc file, because it doesn't get
installed by libffi Makefile, for some reason
* remove the part of the target post install hook that was messing
with libffi.pc, since it was modifying the one in STAGING_DIR
(which is odd for a target post install hook), and the libffi.pc
file is anyway not installed to the target anymore.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Chris Zankel [Mon, 4 Feb 2013 12:52:08 +0000 (12:52 +0000)]
util-linux: add support for Xtensa
[Thomas Petazzoni: rename patch to not have the package version in the
patch file name, adjust the commit title.]
Signed-off-by: Chris Zankel <chris@zankel.net> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
As we are going to add an Xtensa related patch that needs to be
applied *after* the AArch64 patch, let's first add a number to each
util-linux patch file name, so that we know they will be applied in
the right order.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The external toolchain wrapper sets sysroot etc. to an absolute path.
By changing this to a relative path, it is possible to move the host
directory to a different location and still have a working build
system.
This only works for a downloaded external toolchain. For a pre-installed
external toolchain, it is possible to move the host directory to a
different location, but not the external toolchain directory (it does work
if the external toolchain directory lies within the host directory). For
an internal or crosstool-ng toolchain, there is no wrapper so updating the
sysroot path should be done in a different way.
See http://lists.busybox.net/pipermail/buildroot/2012-February/050371.html
for information about others things to do to make the host directory
relocatable.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Acked-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
While the fix makes gnutls build without wchar, it doesn't actually work as
there's no rpl_wctomb implementation so the .so ends up with an undefined
reference to wctomb:
./host/usr/bin/arm-linux-nm -D staging/usr/lib/libgnutls.so.28|grep wctomb
U wctomb
Causing linker errors for packages trying to use it:
CCLD msmtp
host/usr/arm-buildroot-linux-uclibcgnueabi/sysroot/usr/lib/libgnutls.so:
undefined reference to `wctomb'
collect2: ld returned 1 exit status
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
gdb 7.4.1 fails with the following error
gdb-7.4.1/gdb/linux-nat.h:79:18: error: field 'siginfo' has incomplete type
because gdb use 'struct siginfo' instead of 'struct siginfo_t'.
Crosstool-NG allready use the patches to prevent this error so use this
patches for buildroot too.
see also http://permalink.gmane.org/gmane.comp.gcc.cross-compiling/15305
Stefan Fröberg [Wed, 30 Jan 2013 02:46:40 +0000 (02:46 +0000)]
pkg-infra: add <pkg>_CONFIG_FIXUP to fix *-config files
This patch will add <pkg>_CONFIG_FIXUP variable to buildroot infra.
It's purpose is to inform buildroot that the package in question
contains some $(STAGING_DIR)/usr/bin/*-config files and that we
want to automatically fix prefixes of such files.
It is often the case that many packages call these
files during their configuration step to determine 3rd party
library package locations and any flags needed to link against them.
For example:
Some package might try to check the existense and linking flags
of NSPR package by calling $(STAGING_DIR)/usr/bin/nspr-config --prefix.
Without this fix. NSPR would return /usr/ as it's prefix which is
wrong when cross-compiling.
Correct would be $(STAGING_DIR)/usr.
All packages that have <pkg>_INSTALL_STAGING = YES defined and
also install some config file(s) into $(STAGING_DIR)/usr/bin must
hereafter also define <pkg>_CONFIG_FIXUP with the correspondig
filename(s).
This commit fixes the exact same problem than 21a0c11a90f9ff19c78a5f32e0a2e9064fd28054, but for the gdbserver
build. The problem is that when you use the Crosstool-NG toolchain
backend, gawk gets built as a dependency of Crosstool-NG. So the gdb
configure scripts detects it, and assumes it is in the PATH (because
the gdb configure step gets run with TARGET_MAKE_ENV).
But then, the build fails, because it tries to run gawk, but gawk
isn't in the PATH, because we forget to use this TARGET_MAKE_ENV
variable when building gdbserver.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
tvheadend requires an atomic_add operation. Either you have a gcc >=
4.3 toolchain and it uses the gcc intrinsics, or it has special code
for x86, x86-64, PPC and ARM. So in the context of Buildroot, the only
really problematic architecture is avr32, which uses gcc 4.2.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
We have a patch that allows neard to build even on toolchain using old
kernel headers that lack the nfc.h header. However, after the bump to
neard 0.9, this patch was not updated to take into account the two new
plugins that neard has added.
Marek Belisko [Mon, 7 Jan 2013 10:43:18 +0000 (10:43 +0000)]
lvm2: add option for application library
[Peter: make mutual exclusive with dmsetup-only, drop special install / make] Signed-off-by: Marek Belisko <marek.belisko@open-nandra.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Marek Belisko [Fri, 18 Jan 2013 12:08:46 +0000 (12:08 +0000)]
lvm2: Install devmapper.pc to staging/target pkgconfig directory.
Without this change if some package which have dependency on device mapper
and use pkgconfig to check version always fail to find library because
devmapper.pc file wasn't installed.
Signed-off-by: Marek Belisko <marek.belisko@open-nandra.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Simon Dawson [Sun, 20 Jan 2013 11:55:41 +0000 (11:55 +0000)]
avr32: support uClibc 0.9.33.2
Add a uClibc patch from OpenWRT, and tweak an existing patch to cope with
the lack of a dup3 Linux syscall on avr32. This allow uClibc 0.9.33.2 to be
built for avr32.
[Peter: add upstream url for openwrt patch] Signed-off-by: Simon Dawson <spdawson@gmail.com> Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
projects / coffee / buildroot.git / log