In order to keep better track of when a feature got deprecated, and hence
when it can be removed, a new set of symbols BR2_DEPRECATED_SINCE_xxxx_xx is
introduced. These symbols are automatically selected when BR2_DEPRECATED is
selected, and thus are transparent to the user.
A deprecated feature will no longer depend on BR2_DEPRECATED directly, but
rather on the appropriate BR2_DEPRECATED_SINCE_xxxx_xx. If that symbol does
not yet exist, it has to be created in Config.in.
When removing a deprecated feature, one should also check whether this was
the last feature using the BR2_DEPRECATED_SINCE_xxxx_xx symbol, in which
case the latter can be removed from Config.in.
A followup patch will make sure the overview is added to the list of
deprecated features in the manual, so that a buildroot core developer can
easily determine which features to remove in a given development cycle.
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Acked-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
In our wrapper, we forcibly add the -march=, -mcpu= and-mtune= flags
to the actual compiler, this in an attempt to always generate correct
and optimised code for the target.
But in some cases, the caller knows better than we do, and passes its
own set, or subset of those flags. In this case, some may conflict with
the ones we pass. The most prominent offender being the Linux kernel.
For example, on the ARM Raspberry Pi, the Linux kernel will set the
-march=armv6 flag and no -mcpu= flag, but we pass -mcpu=arm1176jzf-s,
which conflicts:
drivers/scsi/scsi_trace.c:1:0: warning: switch -mcpu=arm1176jzf-s
conflicts with -march=armv6 switch
(and so for all the files the kernel compiles, pretty messy)
(note: arm1176jzf-s is not an armv6, it is an armv6zk. Yeah...)
To avoid this situation, we scan our commandline for any occurence of
the possibly conflicting flags. If none is found, then we add our owns.
If any is found, then we don't add any of our owns.
The idea behind this is that we trust the caller to know better than
we do what it is doing. Since the biggest, and sole so far, offender
is the Linux kernel, then this is a rather safe bet.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Peter Korsgaard <jacmet@uclibc.org>
Cc: Arnout Vandecappelle <arnout@mind.be>
Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The Xilinx Microblaze external toolchains that we had support for are
very old, and are causing a huge number of build issues. Thanks to
Spenser Gilliland, we now have support for Microblaze in the internal
toolchain backend, and the autobuilders have been using the internal
toolchain backend since then. Therefore, it's time to deprecate those
old and unusable external toolchains.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Add Linaro ARM 2013.10 and Linaro ARM 2013.11, and remove Linaro ARM
2013.07 and Linaro ARM 2013.08.
The main change for those versions is the switch to eglibc 2.18.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
This commit adds the support for the recently release Sourcery MIPS
2013.11 toolchain (gcc 4.8, gdb 7.6, glibc 2.18), and consequently
removes the support for the Sourcery MIPS 2012.03 toolchain.
While we're at it, also fix the incorrect help text related to the
MIPS64 multilib selection.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
This commit adds the support for the recently release Sourcery ARM
2013.11 toolchain (gcc 4.8, gdb 7.6, glibc 2.18), and consequently
removes the support for the Sourcery ARM 2011.09 toolchain.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
This commit allows to build an internal toolchain for the Microblaze
architecture, with either glibc or eglibc.
Note that we add an explicit list of architectures that are supported
by uClibc, and Microblaze is not part of them, because it currently
doesn't build for this architecture.
[Thomas: add better commit log, add architecture dependencies on
uClibc, to avoid selecting uClibc on Microblaze]
Signed-off-by: Spenser Gilliland <spenser@gillilanding.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The glibc dynamic linkers for ppc64 and s390x are named ld64.so.*
so modify the check_glibc test to match them.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
uClibc only works well for O32 ABI on mips64, so default to eglibc that
works just fine with N32 & N64 ABIs as well.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since 924b8739da we no longer declare glibc and eglibc support as
experimental, so it shouldn't be mentioned in the help text anymore.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(e)glibc doesn't support a fully statically linked userspace. Even a
basic program such as Busybox fails to do authentication due to glibc
loading some libraries dynamically. Therefore, we disable the
possibility of using a (e)glibc toolchain when
BR2_PREFER_STATIC_LIB=y.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(e)glibc doesn't support a fully statically linked userspace. Even a
basic program such as Busybox fails to do authentication due to glibc
loading some libraries dynamically. Therefore, we disable the
possibility of building a glibc toolchain when
BR2_PREFER_STATIC_LIB=y.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The glibc and eglibc support has been introduced since a little bit of
time now, I believe we can remove the "experimental" statement next to
it.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This patch fixes the following whitespace problems in Config.in files:
- trailing whitespace
- spaces instead of tabs for indentation
- help text not indented with tab + 2 spaces
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Based on a suggestion from Peter, this commit factorizes the logic and
regular expressions that are used to find the sysroot and libdir for a
given compiler. It reduces a bit the duplication of code, and
centralizes the most bizarre part of this logic in one place.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
[Peter: drop extra # as pointed out by Baruch Siach]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
In a1d94aaa3a ('toolchain-external: add support for musl C
library'), we made the following change to the SYSROOT_DIR mangling
logic:
- SYSROOT_DIR=`echo $${LIBC_A_LOCATION} | sed -r -e 's:usr/lib(32|64)?/(.*/)?libc\.a::'` ; \
+ SYSROOT_DIR=`echo $${LIBC_A_LOCATION} | sed -r -e 's:(usr/)?lib(32|64)?/(.*/)?libc\.a::'` ; \
This was needed to accomodate for musl based toolchains that don't
have libc.a in usr/lib/..., but directory in lib/... Basically, the
change makes the usr/ at the beginning optional.
However, with the very permissive (.*) matching in the middle of the
path, the change above had an unexpected consequence: any path contain
'/lib' would be truncated before this lib. As an example, Peter
reported that his builds, running from /var/lib/buildbot/ were no
longer working because the SYSROOT_DIR was decided to be /var instead
of something like
/var/lib/buildbot/buildroot/output/host/opt/ext-toolchain/arm-linux-gnueabihf/libc/.
So, this commit changes (again!) this regexp by changing (.*) to
([^/]*), the idea being that it will match only *one* path
component. Note that this intermediate (.*) directory was added in
e6e60becb0 ('external-toolchain: add support for Linaro 2012.01') to
accomodate for Linaro toolchains that have a subdirectory in their
sysroot named after the target tuple:
$ ./output/host/opt/ext-toolchain/bin/arm-linux-gnueabihf-gcc -print-file-name=libc.a
/home/thomas/projets/buildroot/output/host/opt/ext-toolchain/bin/../arm-linux-gnueabihf/libc/usr/lib/arm-linux-gnueabihf/libc.a
In addition to this, this commit also makes sure that the change
making usr/ optional is properly reported on all the instances of this
regular expression.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
For configurations using a toolchain that is preinstalled on
the host, <pkg>_SITE and <pkg>_SOURCE variables must be kept
empty to avoid downloading any toolchain package.
The actual implementation has been proposed by Thomas Petazzoni.
Signed-off-by: GONZALEZ Laurent <br2@gezedo.com>
Reviewed-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
As Samuel Martin noticed, libthread_db is not only needed when
cross-gdb+gdbserver is used, but also when the native gdb is used on
the target. As a consequence, this patch modifies the glibc package
and the external toolchain logic to ensure that libthread_db is copied
to the target either when the native gdb or gdbserver is enabled, by
relying on the BR2_PACKAGE_GDB option, which is enabled when native
gdb and/or gdbserver are enabled.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Maxime Ripard reported that the Linaro toolchains were not visible
when selecting a Cortex-A, but forgetting to set the EABI to
EABIhf. While this is expected, Buildroot should normally should a
comment in this case. However, the comment is only visible when the
selected ARM architecture is not ARMv7 *and* the EABI is not
EABIhf. Instead, make the comment visible when either the selected
architecture is not ARMv7 *or* when the selected EABI is not EABIhf.
While we're at it, reword the comment text so that it actually fits
within the limits of the menuconfig screen.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reported-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
In 11ec38b695 ("toolchain-external: fix Linaro ARM toolchain
support"), we fixed the support for Linaro EABIhf toolchains by adding
a /lib/arm-linux-gnueabihf -> /lib symbolic link. This is needed
because the dynamic loader looks for libraries in
/lib/arm-linux-gnueabihf rather than the usual /lib, but Buildroot
installs all libraries in /lib.
However, we forgot that the dynamic loader also loads libraries from
/usr/lib/arm-linux-gnueabihf rather than /usr/lib, so this patch fixes
that by adding the necessary symbolic link.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reported-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: Maxime Ripard <maxime.ripard@free-electrons.com>
Previously, an external-toolchain on a mipsel/mips64el target
didn't select the appriopriate endianness but
it asked the user to set the correct CFLAGS on his/her own.
We fix this by appending "-EL" to the toolchain wrapper options
if the user has selected a mipsel/mips64el target.
[Thomas: remove unneeded test on BR2_ENDIAN, since mipsel and mips64el
are always little-endian, and add the corresponding big endian case.]
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Some Config.in(.host) files have constructs like:
config FOO_VERSION
string
default "1.0" if FOO_1_0
default "2.0" if FOO_2_0
default $FOO_CUSTOM_VERSION if FOO_CUSTOM
The dollar sign here is not needed and confusing, so can be removed.
Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The Angstrom toolchains available at
http://www.angstrom-distribution.org/toolchains/ are not usable as
external toolchains in Buildroot, because they are not pure toolchains
with just the C library, but instead complete SDKs with many
cross-compiled libraries (Gtk, Qt, glib, neon, sqlite, X.org, and many
more, approximately 200 MB of libraries).
Buildroot cannot use such toolchains, and while this is documented in
our manual, some users still try to do this. Today, one such user came
on the IRC channel, reporting a build problem, which we started
investigating, only to realize after a long time that he was using an
Angstrom toolchain.
To avoid this problem in the future, we explicitly check if the
toolchain is from Angstrom by looking at the vendor part of the tuple
exposed by the toolchain: as soon as it is
<something>-angstrom-<something-else>, we reject the toolchain with an
explanation.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit fixes bug #6452 (eglibc from Linaro 2013.07 not copied to
target correctly) by:
* Copying only the relevant library loader to the target on ARMhf
(i.e ld-linux-armhf.so and not ld.so*). This is needed since Linaro
toolchains provide two library loaders, one ARMv7 hf, and one ARMv4
soft-float.
* Making sure a $(TARGET_DIR)/lib/arm-linux-gnueabihf/ symbolic link
to $(TARGET_DIR)/lib/ exists, since the dynamic loader of Linaro
toolchains expects libraries to be found in
$(TARGET_DIR)/lib/arm-linux-gnueabihf/.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The copy_toolchain_lib_root function is responsible for copying a
given library (and its symbolic link) to the target filesystem. To do
so, it looks for the library in various locations, and then iterates
over the symbolic link all the way to the library, copying them as
needed to the target filesystem.
However, the latest Linaro toolchains bring an interesting use case:
the lib/ directory in the toolchain is organized as follows:
- ld-linux.so.3 -> arm-linux-gnueabi/ld-2.17...so
- ld-linux-armhf.so.3 -> arm-linux-gnueabihf/ld-2.17...so
- arm-linux-gnueabi/
- all ARMv4T soft float libraries
- arm-linux-gnueabihf/
- all ARMv7 hard float libraries
In order to match what we do with all other toolchains, we want all
those libraries and symbolic links to be copied directly under
$(TARGET_DIR)/lib. This commit does that by adjusting the copy logic.
This is part of the fix for bug #6452 (eglibc from Linaro 2013.07 not
copied to target correctly).
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit adds support for external toolchains based on the musl C
library, as available from http://www.musl-libc.org.
Note that the pre-built musl toolchains available from
http://musl.codu.org/ are not working for the moment, since they lack
sysroot support. However, this problem has been reported to the
maintainer, who has already added sysroot support in his scripts at
https://bitbucket.org/GregorR/musl-cross, and therefore the next
version of the pre-built toolchains should work with Buildroot
out-of-the-box. In the mean time, the musl-cross script must be used
to build the toolchain.
[Peter: reword comment]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The check_glibc function contained checks to verify that the user had
properly enabled the largefile, IPv6, locale and wchar options, to
match how glibc is configured. This was useful when the support for
glibc external toolchains was introduced, but since then, we added the
BR2_TOOLCHAIN_USES_GLIBC symbol that automatically enables largefile,
IPv6, locale and wchar support when a glibc or eglibc external
toolchain is used. Therefore, many of the check_glibc checks are
useless now, so we can remove these.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit fixes various typos and mistakes in the comment at the top
of the helpers.mk file:
* usefull -> useful
* The optional stripping of libraries no longer exists, so there's no
reason to mention it.
* Indicate that the copy_toolchain_lib_root function is also used by
the glibc package, not only by the external toolchain logic.
* Separate more clearly the top comment introducing the entire file,
from the comment introducing the first function.
[Peter: reword top comment, add missing 'by']
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Until now, the copy_toolchain_lib_root function took as argument the
base name of a library (e.g: libm.so), and was assuming that the usual
scheme libm.so.<x> being a symbolic link to the real library was used.
However, with musl based toolchains, the C library is named libc.so
directly, with no symbolic link at all. Therefore, this commit changes
the copy_toolchain_lib_root to move the responsibility of using a
wildcard or not after the library name the caller's responsibility.
So, all the existing LIB_EXTERNAL_LIBS values are modified to have a
.* at the end, so that the behavior is effectively unchanged.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The external toolchain code makes the assumption that all C libraries
have a ld*.so, libc.so, libcrypt.so, libdl.so, libgcc_s.so, libm.so,
libnsl.so, libresolv.so, libutil.so, and when thread support is
enabled, libpthread.so, etc.
However, this is not the case with the musl C library, which
integrates all the functionalities in a single libc.so file. In
preparation of the support of the musl library, we make the current
value of LIB_EXTERNAL_LIBS conditional to glibc or uClibc.
The addition of additional libraries through
BR2_TOOLCHAIN_EXTRA_EXTERNAL_LIBS is kept outside the condition, at
the end.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit converts the 'toolchain-external' logic to the package
infrastructure.
The TOOLCHAIN_EXTERNAL_DIR variable (which points to where the
toolchain is located) is renamed to TOOLCHAIN_EXTERNAL_INSTALL_DIR,
because the former conflicts with the package infrastructure (which
defines the <pkg>_DIR variable for each package as pointing to its
build directory).
The new _EXTRA_DOWNLOADS mechanism is used for Blackfin toolchains.
The extract, configuration and installation steps are converted inside
the <pkg>_EXTRACT_CMDS, <pkg>_CONFIGURE_CMDS and
<pkg>_INSTALL_STAGING_CMDS.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit converts the toolchain-buildroot logic to the package
infrastructure. The package is fairly simple as it only defines
BUILDROOT_LIBC, and depends on host-gcc-final to get the overall
internal toolchain build logic started.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This commit introduces a virtual package called 'toolchain', located
in 'toolchain/toolchain', which simply depends on
'toolchain-buildroot' or 'toolchain-external' depending on the
selected toolchain backend.
For now, toolchain-buildroot and toolchain-external are still manual
make targets, but the following patches convert those backends to use
the package infrastructure as well.
In addition to this:
* The main Makefile is modified to always make BASE_TARGETS point to
this new toolchain virtual package.
* The main Makefile is changed to include all the toolchain/*/*.mk
files: the toolchain virtual package, and the toolchain-buildroot
and toolchain-external directories.
* The dependency of the toolchain on prepare dirs and dependencies is
moved to the toolchain virtual package. It is moved as a
prerequisite of the "toolchain-source" rule to ensure that all
directories are prepared before we even start extracting the
toolchain.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Acked-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The ext-tool.mk logic uses the TOOLCHAIN_EXTERNAL_PREFIX variable
unconditionally, even if the external toolchain is not used. Until now
this wasn't a problem since ext-tool.mk was only included when the
external toolchain backend was selected, but the next patches are
going to include this file unconditionally.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
In order to avoid the work of converting the toolchain-crosstool-ng
logic to the package infrastructure, we remove it from Buildroot,
since it has been deprecated since quite some time.
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 <peter@korsgaard.com>
While the idea of skipping the intermediate gcc step seems to work
fine in most situations, it causes problems with the SSP
support. Until we can figure out a proper solution for this problem,
we need to revert back to the previous solution of a three stages
build.
This reverts commit 2babed4a50.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
In case there are many arguments passed to the tools, the command line
can get very long, and difficult to parse visually.
For example, the Linux kernel passes a lot of arguments to gcc (at least
45, which gives 53 with our hard-coded args). Looking at such a command
line is daunting.
So, add the possibility to print each argument on its own line.
Also, enclose all args between single quotes, so the command line
can be safely copy-pasted without special chars (spaces, $) being
inrerpreted by the shell.
Add blurb about toolchain-wrapper to documentation at the same
time.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Thomas De Schampheleire <patrickdepinguin@gmail.com>
Acked-by: Luca Ceresoli <luca@lucaceresoli.net>
Tested-by: Luca Ceresoli <luca@lucaceresoli.net>
Acked-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Update to the latest update for the 2013.05 toolchain
released on 11 September 2013.
https://sourcery.mentor.com/GNUToolchain/release2554
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Acked-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
