elf_xtensa_gc_sweep_hook doesn't correctly unreference symbols that were
made local, that results in link failure with the following message:
BFD (GNU Binutils) 2.24 internal error, aborting at elf32-xtensa.c line
3372 in elf_xtensa_finish_dynamic_sections
elf_xtensa_gc_sweep_hook determines symbol reference type (PLT or GOT) by
relocation type. Relocation types are not changed when symbol becomes
local, but its PLT references are added to GOT references and
plt.refcount is set to 0. Such symbol cannot be unreferences in the
elf_xtensa_gc_sweep_hook and its extra references make calculated GOT
relocations section size not match number of GOT relocations.
Fix it by treating PLT reference as GOT reference when plt.refcount is
not positive.
Fixes:
http://autobuild.buildroot.net/results/3e2e24f697e26c93d4d95782b1cb7799fa620a7a/http://autobuild.buildroot.org/results/97d4c96d6f6cdc1ed4007456f4ab70be9dfa41b5/
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
postgresql currently does not build in BR2_STATIC_LIBS=y
configurations, and since there is little interest in using such a big
piece of software in statically linked configuration, this commit
makes postgresql depends on !BR2_STATIC_LIBS.
Fixes:
http://autobuild.buildroot.net/results/f9ed96d22e91cdba9ad92c4d4ea52e422bf1f1c9/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The acl and attr packages are using autoconf and libtool, but not
automake, which causes all sort of troubles. Some of them are already
worked around in acl.mk and attr.mk, but there is another one: the
libdir variable in the .la file remains empty, which causes libtool to
sometimes be confused about where the acl and attr libraries are
installed.
This was causing build issues with libarchive, but interestingly only
in situations where we had zlib *and* libxml2 also part of the build,
which maybe affects how libtool in libarchive behaves since libarchive
can use zlib and libxml2 as optional dependencies.
So what this patch does is to mungle the libattr.la and libacl.la
files to make sure they contain the right libdir values. Yocto is
doing exactly the same thing.
Upstream attr and acl have been converted fully to automake, so
hopefully the upcoming upstream release will allow us to get rid of
those hacks, but in the mean time, it seems to be the most appropriate
solution.
Here is a minimal defconfig that exhibits the problem, and which
allows to show that this patch fixes it:
BR2_bfin=y
BR2_BINFMT_FLAT=y
BR2_COMPILER_PARANOID_UNSAFE_PATH=y
BR2_INIT_NONE=y
BR2_PACKAGE_LIBARCHIVE=y
BR2_PACKAGE_LIBARCHIVE_BSDCPIO=y
BR2_PACKAGE_ZLIB=y
BR2_PACKAGE_LIBXML2=y
BR2_PACKAGE_ACL=y
Fixes:
http://autobuild.buildroot.org/results/e96/e96c8bec1039d9fe8c6c51fd48b6be0dc8be51d7/
(and many, many similar build failures)
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The newly added patch makes sure that gold linker functionality is not
used when doing host builds, since qt5 only tests the availability of
this feature with the target compiler.
Fixes:
http://autobuild.buildroot.net/results/f403a76ac0abbf8488373c0dffb4487f5d98c55d/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
There are currently two patches with the 0002 sequence number, which
isn't great. Rename one of them to use the 0003 sequence number.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
mosquitto triggers the infamous _gp issue with the NIOS 2 toolchains,
so let's not allow mosquitto in such situations.
Fixes:
http://autobuild.buildroot.net/results/b853369452115b0c6f32c6c960af2dbdf71a74af/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The build of boost-log with a uClibc toolchain on PowerPC fails with
some weird C++ issues. Since nobody ever looked into this problem,
let's "fix" the autobuilder issue by not allowing to select boost-log
in such a situation.
Fixes:
http://autobuild.buildroot.net/results/13b43105caf4d3952de70030b51f8d96cf6604ee/
(and many, many similar build failures)
[Peter: also add dependency to comment]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
-q option is missing in BOOST_INSTALL_STAGING_CMDS,
so the build doesn't stop on the first error.
This help to see what happened.
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Fixes:
http://autobuild.buildroot.net/results/d9ec39dc5829aac1b25bfea7484458fcebd68412/
add cstdlib header to ensure EXIT_FAILURE presence
Using some cross-compiling toolchains, EXIT_FAILURE is not always
properly defined without this inclusion.
Patch is upstream:
https://github.com/zeromq/zmqpp/pull/109
[Thomas: use a patch formatted with git format-patch, rather than the
output of 'git show'.]
Signed-off-by: Lionel Orry <lionel.orry@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
CMake complains about Policy CMP0037 that reserve some
target name such "test".
So remove the warning by renaming test to xmltest.
Backport an upstream patch:
0b2c22bd46be95b08f234b8b4cb262f72ab3f0e8
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The current NIOS 2 toolchains are not capable of building Boost, so
let's disable it and its reverse dependencies. Even though it's not
strictly an architecture dependency, we use the <pkg>_ARCH_SUPPORTS
paradigm for this dependency, since it simplifies a lot handling all
boost reverse dependencies, and is anyway quite similar to an
architecture dependency since we don't display a comment about this
dependency.
Fixes:
http://autobuild.buildroot.net/results/e119b1ef55c546e0d0598b85c46ceefa5c43d5a6/
[Peter: also update mpd comment]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This reverts commit 884af65fd5. Thanks
to the change of the installation path of host-autoconf-archive
macros, host-ola will no longer see them when doing its autoreconf, so
the patch that was added by the commit being reverted here is no
longer needed.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The host-autoconf-archive package was added to provide the necessary
autoconf macros for the libsigrok package, and later also used for the
thrift package.
However, it installs its autoconf macro in
$(HOST_DIR)/usr/share/aclocal, which is part of the include directory
of autoconf macros used for all autoreconf invocations. Unfortunately,
using those macros for all packages is causing a number of
regressions: one on host-ola fixed in commit
884af65fd5, and one in snmpp. Those
packages are autoreconf'ed, but if they are autoreconf'ed with the
autoconf-archive macros installed, they no longer work.
To solve this problem, this commit implements a simple solution that
makes using host-autoconf-archive macros an opt-in mechanism. To
achieve this, the autoconf-archive package is modified to install its
macros in a special directory, and only the packages that really want
to use those macros will have to pass a -I flag pointing to this
directory in their <pkg>_AUTORECONF_OPTS variable.
Fixes:
http://autobuild.buildroot.org/results/712/712c8083b9eeafa41b60790961028c40b28f59ad/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Romain Naour <romain.naour@openwide.fr>
Tested-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This is the most recent version of the package.
Also it fixes static builds (see
http://autobuild.buildroot.net/results/1cc/1ccf0829c13896c000aed1956155a34844207b2f)
with commit http://sourceforge.net/p/arma/code/3569/ that was made right
before 4.400.0 release.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Samuel Martin <s.martin49@gmail.com>
Acked-by: Samuel Martin <s.martin49@gmail.com>
Tested-by: Samuel Martin <s.martin49@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The patches touching configure.ac were dropped in 9d9bbf9b79 (rsyslog: bump
to version 8.8.0, 2015-02-25). Remove autoreconf as well.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
--disable-install-libiberty configure option is broken
in gcc 4.8.x, so libiberty.a is always installed in HOST_DIR.
This library broke the host-gdb build due to a fpic/fPIC issue.
Since binutils use a copy of gcc's libiberty sources, it's
affected by the same bug.
binutils-arc-2014.11 which is based on 2.23 branch doesn't have
the PR56780 fixies, so backport them.
Fixes:
http://autobuild.buildroot.net/results/ca4/ca45297aa0ffbc9062ed92dc7ac070b0b33001de/
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Acked-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Contrary to what commit f5866484f (libnftnl: exclude for CS PPC toolchains)
stated, a missing sa_family_t definition IS specific to the kernel headers
version, and not something particular to the CS toolchains, so instead
require 3.1+ headers.
The specific kernel change fixing the headers is:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6602a4baf4d1a73cc
The only reverse dependency of libnftnl (nftables) already needs 3.4+, so
this dependency doesn't need to be propagated.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
commit dfbfa6ef (dhcpcd: blacklist Sourcery PowerPC toolchains) blacklisted
specific toolchains, but the issue is actually that the kernel headers
before 3.1 didn't define sa_family_t, so instead require 3.1+ headers.
The specific kernel change fixing the headers is:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6602a4baf4d1a73cc
The only reverse dependency of dhcpcd (network-manager) already needs 3.7+,
so this dependency doesn't need to be propagated.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Even though lxc uses the autotools, it does not use libtool to build
its libraries, and does it manually. It forces building a shared
library. So let's comply with this decision, and mark lxc as not
available in BR2_STATIC_LIBS configurations.
Fixes:
http://autobuild.buildroot.org/results/588/588dc28315a9114374f148bbc64c4ec3d7b4e295/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Instead of hardcoding -lssl -lcrypto as the linker flags to link with
OpenSSL, use pkg-config to fetch those flags. This allows to behave
properly when doing static linking, since pkg-config returns the
entire set of libraries we should link against.
Fixes:
http://autobuild.buildroot.org/results/d71/d710ecfa9ed0a8e3f97d2010d3dabb372485c7f2/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Gergely Imreh <imrehg@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add a patch to U-Boot to use pkg-config to get the OpenSSL link flags
when pkg-config is available. This allows to make sure that static
linking works properly. The patch requires adding $(TARGET_MAKE_ENV)
so that the Buildroot provided pkg-config is used instead of the host
one.
The U-Boot patch has been submitted upstream.
Fixes:
http://autobuild.buildroot.org/results/a9f/a9f316cd076a74b8730ce4cdcdb8176da4ed9eb3/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Gergely Imreh <imrehg@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This reverts commit aa5254f1a2.
Now that zeromq's pkg-config file is fixed, we no longer need this
workaround for static linking.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
These compiler options shouldn't be used when linking statically, and in
particular FLAT binaries cannot use -fPIE and -pie code, so let's teach
libmemcached to not use such flags in those setups.
Fixes:
http://autobuild.buildroot.org/results/8cb/8cbf57f9136cb42be31c88e0f1f32d3d2353e5ff/
and many previous similar build failures
[Peter: rework/reword to do this based on BR2_STATIC_LIBS]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Reviewed-by: Romain Naour <romain.naour@openwide.fr>
Tested-by: Romain Naour <romain.naour@openwide.fr>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
--disable-install-libiberty configure option is broken
in gcc 4.8.x, so libiberty.a is always installed in HOST_DIR.
This library broke the host-gdb build due to a fpic/fPIC issue.
Note: host-binutils-arc-2014.12 install libiberty.a in HOST_DIR
but it was overwritten by the gcc one. The host-binutils's
libiberty.a also broke the host-gdb build. This should be
fixed in a followup patch.
Signed-off-by: Romain Naour <romain.naour@openwide.fr>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Qemu doesn't understand sh4a and sh4aeb as separate architectures, it
only knows about sh4 and sh4eb. However, it does support sh4a, since
it emulates the SH7785 CPU which is a SH4A.
Therefore, with this commit, sh4a/sh4aeb are translated as sh4/sh4eb
for Qemu.
Fixes:
http://autobuild.buildroot.org/results/8f9/8f99a776f35d4ab0841d9fd2db509850ef603bd4/
and many previous similar build failures
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
