The definition of TOOLCHAIN_DIR is the same regardless of whether
external or internal toolchains are used. Moreover, move its
definition together with all the other *_DIR definitions.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
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>
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 $(ARCH) at the two
locations the variable was used.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Instead of having a configuration option BR2_GNU_BUILD_SUFFIX, let's
use config.guess to guess the build system type.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Now, we just hardcode the image filenames to be rootfs.$(FSTYPE), in
the $(BINARIES_DIR).
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Lay down the path to add more than two toolchain kinds:
- check the type of toolchain as:
ifeq (toolchain_buildroot,y)
blabla buildroot-specific
else ifeq (toolchain_external,y)
blabla external-specific
endif
- prefer using positive checks, a-la:
ifeq (foo,y)
instead of:
ifneq (bar,y)
(where foo and bar are mutually exclusive)
- have the toolchain_buildroot case always appear first
- gettext is handled differently, because we want to add an option
only if not using the buildroot toolchain, hence we use ifneq.
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>
Now that we build Perl modules in $(HOST_DIR), PERLLIB must be defined
to $(HOST_DIR)/usr/lib/perl.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
During the compilation of target or host packages, host tools might be
used, and in turn, they might use host libraries installed in
$(HOST_DIR)/usr/lib. Therefore, we pass a LD_LIBRARY_PATH variable
when building packages.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
This new infrastructure allows to write simpler .mk files for packages
not using the autotools as their build system, by factorizing many
common steps (download, extract, patching), and will more easily allow
Buildroot-wide changes in how the packages are handled.
The main macro is called GENTARGETS and works similarly to the
AUTOTARGETS macro that already exists for autotools-based
packages. However, the set of variables to be defined before calling
the macro is different. Refer to the documentation for details.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Those are already in TARGET_CONFIGURE_ARGS. Also get rid of unused
BR2_AC_CV_FUNC_MALLOC_0_NONNULL variable.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
hard/softfloat is just one of the many config options, and it only
clutters the filenames.
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
To reflect the new output directory hierachy rename the Makefile variable
TOOL_BUILD_DIR to TOOLCHAIN_DIR.
Signed-off-by: Michael Roth <mroth@nessie.de>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
PKG_CONFIG_PATH adds a path to the pkg-config search path, but this is not
enough when we are building host tools - the host pkg-config default path
is still used, and this will pick up .pc files in the staging_dir tree.
PKG_CONFIG_LIBDIR overrides the pkg-config default path and ensures only
host .pc files are found. This fixes a pango build failure when building
for the host with cairo PNG support enabled.
Signed-off-by: Will Newton <will.newton@gmail.com>
In the output directory, we now have
- build/ where all the packages are built
- images/ where the final kernel and rootfs images are stored
- staging/ the staging directory (containing the development files
and libraries compiled for the target)
- target/ which contains the target root filesystem
- host/ which contains all the host programs
- stamps/ which contains the stamps files
Therefore, the build_ARCH and toolchain_build_ARCH have been
removed. People willing to use the same Buildroot sources to compile
for different architectures are invited to use the O= command line
option for out-of-tree compilation.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
In an attempt to disable the compilation and installation of
documentation of gtk packages, a DISABLE_DOCUMENTATION variable
containing --without-html-dir has been added to
package/Makefile.in. This variable is used by
package/Makefile.autotools.in so that --without-html-dir is in effect
passed to all autotools-based packaging relying on the
Makefile.autotools.in infrastructure.
Unfortunately, --without-html-dir doesn't work. It leads libglib2
./configure script to think that HTML_DIR is "no", which leads to the
installation of the documentation in $(STAGING_DIR)no (yes, with the
"no" suffix at the end). The issue is that --with-html-dir is not an
enable/disable type of option, it's an option that only allows to pass
a PATH for documentation installation.
As we don't want the documentation to be installed in this odd
$(STAGING_DIR)no directory, we simply get rid of this option.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The AUTO_CONFIGURE_OPTS is not used anywhere in the tree. autoconf
based packages should used the Makefile.autotools.in machinery
instead.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The XSERVER variable used to be defined by package/Makefile.in because
the X server package name was different depending on the type of X
server that was choosen. Nowadays, the name of the package is always
xserver_xorg-server, so there's no point in having this XSERVER
intermediate variable.
This patch makes all packages use xserver_xorg-server directly as a
dependency, and removes the XSERVER variable from package/Makefile.in.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The option BR2_PACKAGE_TINY, BR2_PACKAGE_XORG and BR2_PACKAGE_XGGI
don't exist, so do not use them to find the value of the $(XSERVER)
variable.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Since the move of many host binaries to $(HOST_DIR), the TARGET_PATH
for internal toolchain had been changed to include $(HOST_DIR)/bin and
$(HOST_DIR)/usr/bin, but this change was not reflected to the
TARGET_PATH used in external toolchain configuration. This patches
fixes this.
We also remove $(STAGING_DIR)/bin and $(STAGING_DIR)/usr/bin from the
TARGET_PATH in external toolchain configuration, since the STAGING_DIR
is not supposed to contain host binaries.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Some configure scripts seems to ignore CXX settings if it is set to
the empty string, and goes back to the default (<arch>-linux-g++),
so use false instead, as that will loudly break the build if the
C++ compiler is ever used.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Change the definition of TARGET_LDFLAGS to use --sysroot
$(STAGING_DIR) instead of -L$(STAGING_DIR)/lib
-L$(STAGING_DIR)/usr/lib. It fixes the following failure while trying
to build mtd-utils :
/usr/local/xtools/arm-unknown-linux-uclibcgnueabi/bin/arm-unknown-linux-uclibcgnueabi-gcc -L/home/thomas/local/buildroot-output/build_arm/staging_dir/lib -L/home/thomas/local/buildroot-output/build_arm/staging_dir/usr/lib -o /home/thomas/local/buildroot-output/build_arm/mtd_orig/flash_eraseall /home/thomas/local/buildroot-output/build_arm/mtd_orig/crc32.o /home/thomas/local/buildroot-output/build_arm/mtd_orig/flash_eraseall.o
/usr/local/xtools/arm-unknown-linux-uclibcgnueabi/lib/gcc/arm-unknown-linux-uclibcgnueabi/4.3.2/../../../../arm-unknown-linux-uclibcgnueabi/bin/ld: cannot find /lib/libc.so.0
collect2: ld returned 1 exit status
make[1]: *** [/home/thomas/local/buildroot-output/build_arm/mtd_orig/flash_eraseall] Error 1
At the same time, simplify the definition of TARGET_CFLAGS, because
the -I$(STAGING_DIR)/include -I$(STAGING_DIR)/usr/include
-I$(TOOLCHAIN_EXTERNAL_PATH)/$(TOOLCHAIN_EXTERNAL_PREFIX)/include are
no longer necessary since we sysroot the toolchain in $(SYSROOT_DIR).
This patch has no effect on non-external toolchain builds.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Move stamp (dependency) files outside the (version specific) source
directories, so other packages can hardcode dependencies on them instead
of having to use <PACKAGE>_VERSION variables.
This is important as the variables in the make rules are evaluated when
the rules is seen, which might be before the dependent makefile is parsed
(and hence <PACKAGE>_VERSION variable is known, screwing up stuff.
The downside of this is that the package isn't automatically rebuilt
when the version changes (E.G. by a svn update) and you now also have to
remove the stamp files next to $(BUILD_DIR)/<PACKAGE>-* to force a rebuild.
This matches upstream tarball, doesn't screw up existing .config's with
BR2_PACKAGE_PKGCONFIG and makes sure the patch gets applied for target
compilation.
gcc < 4.2.0 doesn't support -Wno-overlength-stings, but gcc-4.3.x configure
fails to detect that, breaking the build.
Work around it by detecting the host gcc version (and store in HOSTCC_VERSION)
and set the proper configure variables for gcc < 4.2.0.
The VFP is only available for a few ARM CPUs at the moment,
so this breaks the liboil build.
A patch is available upstream which only enables "-mfpu=vfp"
if "--enable-vfp" is given to "configure".
Autotools needs to be run for liboil for this to take effect.
A new configuration BR2_VFP_FLOAT is added to allow enabling vfp.
If this is "yes", then "-mfpu=vfp" is added to CFLAGS.
TARGET_PATH didn't contain $(STAGING_DIR)/usr/bin, which means that
programs installed in $(STAGING_DIR)/usr/bin were not considered for
execution during Buildroot build process. This was a problem with host
automake/autoconf/libtool, which could not be found.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
* In toolchain/external-toolchain/ext-tool.mk, copy the contents of
the sysroot directory to the staging dir.
* In package/Makefile.in, add a --sysroot CFLAGS pointing to the
staging dir
* Remove the CFLAGS and LDFLAGS definition from
TARGET_CONFIGURE_OPTS. I haven't investigated exactly why, but with
these options, DirectFB fails to build because it cannot find
PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP, even if DirectFB's Makefile
properly sets -D_GNU_SOURCE.
I have already sent this patch on December, 2nd to the mailing-list,
but got no feedback. So let's commit and see what happens :-)
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The following changes allow for use of a central configure cache
file. This speeds up configuration of packages.
Its use is configurable at the top level (BR2_CONFIG_CACHE - default n).
Old style makefiles can use it if they use the following MACRO in makefiles:
$(AUTO_CONFIGURE_TARGET) see my change to directfb.mk.
New style Autotools.in will use it if you set the global option.
However you can enable the global option and on a per package overrule it by doing
the following: $(PKGNAME)_USE_CONFIG_CACHE = NO see fontconfig.mk for an example
of this.
Finally I have removed a few config variable settings which indicated no CXX compiler
as this is wrong and breaks the build when using this central cache.
Config.in | 8 ++++++++
package/Makefile.autotools.in | 5 ++++-
package/Makefile.in | 28 +++++++++++++++++++++++++++-
package/atk/atk.mk | 2 +-
package/directfb/directfb.mk | 7 +------
package/fontconfig/fontconfig.mk | 3 +++
package/libglib2/libglib2.mk | 2 +-
package/libgtk2/libgtk2.mk | 1 -
8 files changed, 45 insertions(+), 11 deletions(-)
I would appreciate feedback on this change (I have been testing for 2-3 weeks)
But I can never test all cases! If you enable the BR2_CONFIG_CACHE option some
Makefile.autotools.in based packages may now break - I cannot build them all.
In this case you may need to remove config options that are being hardcoded all
over the place (like gtk saying we have 2 CXX compiler) or disable the use
of CONFIG CACHE file like I have done in fontconfig.
I can build all packages required to get WebKit on DirectFB up and running
and it runs fine.
I will try to resolve any issues this creates as fast as I can.
Signed-off-by: Daniel Laird <daniel.j.laird@nxp.com>
This patch is a new version of a patch already sent several times on
the mailing-list, committed and reverted a few times by Daniel Laird,
due to several imperfections. This version is a new try at finding a
solution that works for everybody. Hopefully it'll work :-)
The original problem is that external toolchain builds failed because
packages couldn't find their dependent libraries at configure time and
could not be linked with them. To fix these two problems, two things
are added:
* The TARGET_LDFLAGS variable was exposed as LDFLAGS at ./configure
time thanks to TARGET_CONFIGURE_OPTS. The TARGET_LDFLAGS variable
contains -L options with the path in the STAGING_DIR for the
libraries. It allows ./configure scripts to properly compile the
small test programs testing whether a dependency is properly
installed.
* The TARGET_CFLAGS contains a new -Wl,--rpath-link option for both
$(STAGING_DIR)/lib and $(STAGING_DIR)/usr/lib. It allows library
depending on other libraries to link properly. The TARGET_CFLAGS is
exposed as CFLAGS in TARGET_CONFIGURE_OPTS.
This new version fixes a problem encountered by hartleys
<hartleys@visionengravers.com> when building the kernel. The problem
was that the -Wl,--rpath-link options were added to LDFLAGS, while
there are options for the C compiler, not the ld linker. Moving them
to CFLAGS seems to fix the issue.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Revert the rpath patch, it looked good up until someone tried
to build a kernel as well. This seems to break as a result.
Will post a new patch soon and see how that goes..
Signed-off-by: Daniel Laird <daniel.j.laird@nxp.com>
Apply the patch I posted some time ago that fixes
rpath issues with external toolchains.
Has been tested by users of buildroot and feedback looks good.
Signed-off-by: Thomas Petazzoni
Signed-off-by: Daniel Laird <daniel.j.laird@nxp.com>
As per various email discussions add -rpath-link
to the LDFLAGS.
This definately fixes a few issues for Thomas and myself
Any objections and it can be pulled again.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@...>
Signed-off-by: Daniel Laird <daniel.j.laird@nxp.com>
Upgrade to pkgconfig 0.23 which has native sysroot support (buggy,
but easily fixable), which allows us to get rid of pkgconfig-filter.sh.
At the same time cleanup the makefile.
Packages that need to pass additional CFLAGS in their .mk have to do something
like this: ...configure $(foreach i,$(foo_CFLAGS),CFLAGS+=$$i) --prefix=...
Will need to try to copy eventual pre-existing project-specific deps back
to package/config in order not to mess up the corresponding timestamps (to avoid superfluous rebuilds)..
quite work yet for me, but this clearly is a huge project and not having it
quite work on the first pass is hardly unexpected. We definately want this
stuff in buildroot.
