When sstrip is selected it tries to strip kernel modules too.
Unfortunately this fails with a "unrecognized program segment header
size" error thus interrupting the build process.
We introduce a new $(KSTRIPCMD) strip command for this, being a regular
strip when sstrip is selected and an empty stub when not stripping.
At the same time get rid of the REMOVE_SECTION_* variables, as they are
only used once.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Certain toolchain configs (E.G. on blackfin), don't use symlinks, so the
file name to match is the actual SONAME (.so.*). Support these toolchains
as well.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Fix previous commit not to strip libthread_db library. In the previous
commit the wildcard was incorrectly matching the symbolic link instead
of the real library file.
Signed-off-by: Javier Viguera <javier.viguera@digi.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
For proper threading debug support, the libthread_db.so library cannot
be stripped on the target. This is because the target gdbserver will
also load up this library at runtime and poke around its symbols.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Often times, the only reason /usr/share exists is because of documentation
installed into it. So once we're done cleaning the docs, attempt to rmdir
/usr/share. If it's empty, things will work out nicely.
[Peter: use make infrastructure to ignore errors rather than shell]
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The installed kernel modules should have useless build cruft stripped out
of them. On my system, a default build went from a very unreasonable 30MB
to a normal 3MB (on disk) and from 14MB to 3MB when compressed due to the
info stored in the kernel modules alone.
[Peter: don't hardcode binutils strip specific options]
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Ensure host packages (dependencies of target packages) are also handled
by make source / make external-deps.
This has to be done a explicitly, as we don't list these in Kconfig,
and hence also not in a variable like TARGETS, so instead we have to
look at the <PKG>_DEPENDENCIES variables for each enabled package and
extract the host packages from there.
Host packages can in turn also have dependencies, so we have to follow
those as well. Ideally this should be done recursively, but as that's
pretty hard to do in make, it is limited to 1 level for now.
Finally, host packages share source files with target packages, so
pipe output of make external-deps through sort -u to ensure duplicates
are removed.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Closes#2731
Bootloader and Linux kernel steps can add content to the rootfs, so ensure
the cleanup / postprocess steps are moved after those.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The configuration cache shared between packages, while being in
principle a nice idea to speed-up the configuration of packages by
avoiding repetitive identical checks, turned out to be unreliable due
to the subtle differences between similar but not identical checks in
different packages. After spending some time trying to fix those, we
concluded that supporting the shared configuration cache is definitely
too hard and too unreliable, and that we'd better get rid of it
altogether.
This patch therefore removes the shared configuration cache
infrastructure and usage.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
It seems that target-specific variables don't mix with target rules.
Thanks to Yann for helping debugging the issue.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Some packages use the buildroot version string (to set their own version
string). Computing the version string globally will makes it easy for
them to use it instead of re-computing it in every packages.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
This re-instates writing the version string in .config headers, and no
longer provides it as a kconfig symbol in .config (it is now a variable
in the Makefile, and in the environment).
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Move it down, users should not mess with it.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Default HOST_CFLAGS to -O2, so host tools (like the cross compiler) are
built with optimization by default.
Based on a patch by Will Newton <will.newton@gmail.com>.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Closes#2707
It's not only for kernel headers, and the kernel headers .mk file
isn't included for crosstool-ng toolchain, which broke linux/u-boot/..
builds.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
And all the infrastructure surrounding it. A broken sed implementation
is quite rare nowadays, as seen by the fact that the current host-sed
support has been broken for a while, so just get rid of it.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The top-level Makefile can be executed in parallel, as it causes problems.
We can force make to be not parallel.
It's been reported many times, and recent discussions on IRC with kos_tom,
and user nick-named knee, led to this patch.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
If building out-of-tree, add a Makefile wrapper that calls-out to the real
Makefile with proper args.
Avoids having to pass -C and O= every time we call make.
This is highly inspired from how the Linux kernel does it, and portions of
it have been used. We can't use exactly the same implementation as the
kernel does, because:
- the script writing the wrapper has been expunged of the few lines
that were too kernel-related: in buildroot we do not need the version
string in the wrapper, and we do not have a patchlevel version;
- "in-tree build" does not have the same meaning for the kernel and for
buildroot: for the kernel, $(O) point to the $(TOPDIR), while for
buildroot $(O) points to $(TOPDIR)/output.
For more complete explanations, see:
http://lists.busybox.net/pipermail/buildroot/2010-September/037815.html
[Peter: minor tweaks]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998 at anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The new DL_MODE variable dispatches between the various download
implementations of each method (Git, Subversion, Wget) to deal with the
normal download (default mode, 'DOWNLOAD'), the source-check
('SOURCE_CHECK') and to show the external dependencies for external-deps
('SHOW_EXTERNAL_DEPS').
For the latter, the legacy script wget-show-external-deps.sh is no
longer required as $(WGET) isn't called directly anymore but always
through the DOWNLOAD helper.
Signed-off-by: Maxime Petazzoni <maxime.petazzoni@bulix.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
It's not really necessary to differenciate the commands for checking out
or updating a repository. Only the path to the binary and eventual
top-level flags are useful to configure.
Signed-off-by: Maxime Petazzoni <maxime.petazzoni@bulix.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
In preparation for the re-work of the DOWNLOAD helper to support
multiple download methods and protocols, the spider feature used with
wget is removed for now until it is re-implemented on top of the new
download methods.
Signed-off-by: Maxime Petazzoni <maxime.petazzoni@bulix.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The previous commit has removed calls to conf_write_autoconf(), which
is the function that generates the KCONFIG_AUTOCONF,
KCONFIG_AUTOHEADER, KCONFIG_TRISTATE files and the split config (with
one file per config item). Therefore, those things were not generated
anymore before the build.
In order to get them generated before the build, we use the same
mechanism as the kernel: run a silentoldconfig when the .config file
is newer than the KCONFIG_AUTOCONF file.
In Buildroot, all those elements are not really used today, except the
split config which is used a little bit in the toolchain build, in a
try to make sure the toolchain gets rebuilt properly when the
configuration changes. It does not seem that this work has been
completed.
However, as we want to keep the same behaviour as previously, we have
to generate all those elements before starting the build.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
savedefconfig allows to create a minimal defconfig file from an
existing configuration. For example :
make O=/path/to/some/buildroot/build/directory savedefconfig
will generate a minimal 'defconfig' file in the main Buildroot source
directory.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Instead of duplicating the definition of KCONFIG_AUTOCONFIG,
KCONFIG_AUTOHEADER and BUILDROOT_CONFIG, let's define them in a
COMMON_CONFIG_ENV variable, which is used by all the xconfig, gconfig,
menuconfig, nconfig, config, oldconfig, randconfig, allyesconfig,
allnoconfig, randpackageconfig, allyespackageconfig,
allnopackageconfig, defconfig, %_defconfig targets.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Our kconfig-to-buildroot2.patch hasn't been kept up to date with all
the changes made into package/config, and a single patch wasn't very
practical to maintain all our changes. Therefore, this commit adds a
package/config/patches directory, which contains a Quilt series of
patches that correspond to our modifications to the kconfig mechanism.
The huge kconfig-to-buildroot2.patch has been split into 16 smaller
patches. The purpose of some of the modifications has been clearly
identified, while some others were not identified.
The 16 patches together do match exactly the old
kconfig-to-buildroot2.patch. We have been very careful in making sure
that we wouldn't loose any of our modifications.
The only modifications made are :
* Instead of renaming the kernel Makefile to Makefile.kconfig and
naming the Buildroot Makefile just 'Makefile', we instead keep the
original package/config/Makefile from the kernel
scripts/kconfig/Makefile and name the Buildroot Makefile
package/config/Makefile.br. The main Buildroot Makefile is modified
accordingly.
* The documentation README.buildroot2 is updated to explain how to
upgrade to a newer version of scripts/kconfig.
* The kconfig-language.txt documentation is removed, as anybody can
easily find it in the kernel sources so there's no need to
duplicate it here.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
There's no need to have a .defconfig, when you run make menuconfig,
make xconfig or anything else for the first time, it will just start
with the default configuration anyway. This is what the kernel does.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Recent versions of ldconfig require the directory /var/lib/ldconfig,
otherwise they bail out with an error. Moreover, having a
/etc/ld.so.conf file also avoids a warning.
In addition to this, we remove the redirection of stderr, so that
errors remain visible to the user.
Thanks to Baruch Siach, Andy Gibbs and Darcy Watkins for investigating
the issue and proposing solutions.
This fixes the build on recent distributions such as Fedora 13 or
Gentoo.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Based on patch by Will Wagner <will_wagner@carallon.com>.
Now that the .config is located in the output directory when O=<dir>
is used, we have to pass the O= option to make for
source/source-check/external-deps to get them to look in the right place
for the .config.
Fix it by introducing an EXTRAMAKEARGS variable and use it whenever we
call back into the toplevel Makefile.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
* Remove $(TARGET_DIR)/usr/share/aclocal from target-finalize when not
installing devfiles and
* Remove some (now) redundant cleanup from individual packages
Signed-off-by: Malte Starostik <m-starostik@versanet.de>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Including a bunch of Makefiles with wildcard makes it impossible to add
new toolchain backends. Avoid that by namely including needed files.
The external toolchain still needs to include all the toolchain/*/*.mk
sub-makefiles, as they are needed to build a toolchain that runs on the
target. It is to be noted that the cross-toolchain is not built in this
case, as the make-targets to build the cross-toolchain are not present
in the $(BASE_TARGETS) variable, which is later used to create the
dependency rules.
Also, the comment 'Explicit ordering' has been removed, as it is mis-
leading. It is make's responsibility to create the proper ordering based
on the dependency rules it finds in the Makefiles
Signed-off-by: Yann E. MORIN <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
The helper functions used for external toolchains may also be useful
to alternate toolchain backends (currently, the external toolchain is
the sole user).
Signed-off-by: Yann E. MORIN <yann.morin.1998@anciens.enib.fr>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Commit ed0d45fdd (Choose host/target ldconfig based on availability)
added a runtime check for a cross-ldconfig being available.
Unfortunately this checks runs too early (at package/Makefile.in parsing
time), so it always fails when using an internal toolchain as ldconfig
isn't built yet.
Fix it by moving the check to the only place it is used (target-finalize).
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
We now have for quite some time a configuration cache used to speed-up
the execution of ./configure scripts when compiling programs for the
target. This commit introduces a similar concept when Buildroot
compiles programs for the host.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
