The endianness handling functions in platform.h are protected behind ifdef
__GLIBC__ which musl doesn't define even though it does provide the
endianness handling interface. Work around it by ensuring __GLIBC__ is
defined.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The recent x264 version bump broke the configure step on x86/x86-64 as x264
ends up using gas instead of yasm as assembler. The reason for this is the
recent upstream commit to optionally use nasm instead of yasm if AS= is
passed:
Enables the use of nasm as an alternative to yasm.
Note that nasm cannot assemble x264 with PIC enabled since it currently doesn't
support [symbol-$$] addressing which is used extensively by x264's PIC code.
This includes all 64-bit Windows and 64-bit OS X builds, even non-shared.
For the above reason nasm is currently intentionally not auto-detected, instead
the assembler must be explicitly specified using "AS=nasm ./configure".
Also drop -O2 from ASFLAGS since it's simply ignored anyway.
But as we pass AS=$(TARGET_AS) it ends up using gas instead. Fix it by
explicitly passing AS=yasm instead.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bernd Kuhls [Sun, 30 Aug 2015 12:16:30 +0000 (14:16 +0200)]
package/x11r7/xserver_xorg-server: dri needs arm >= v6
Fixes
CC dri.lo
/tmp/ccc6IbbW.s: Assembler messages:
/tmp/ccc6IbbW.s:3114: Error: selected processor does not support ARM mode `ldrex r2,[r1]'
/tmp/ccc6IbbW.s:3117: Error: selected processor does not support ARM mode `strexeq r2,r0,[r1]'
/tmp/ccc6IbbW.s:3273: Error: selected processor does not support ARM mode `ldrex r2,[r1]'
/tmp/ccc6IbbW.s:3276: Error: selected processor does not support ARM mode `strexeq r2,r3,[r1]'
/tmp/ccc6IbbW.s:3352: Error: selected processor does not support ARM mode `ldrex r1,[r2]'
/tmp/ccc6IbbW.s:3355: Error: selected processor does not support ARM mode `strexeq r1,r0,[r2]'
/tmp/ccc6IbbW.s:3451: Error: selected processor does not support ARM mode `ldrex r3,[r2]'
/tmp/ccc6IbbW.s:3454: Error: selected processor does not support ARM mode `strexeq r3,ip,[r2]'
/tmp/ccc6IbbW.s:3522: Error: selected processor does not support ARM mode `ldrex r3,[r0]'
/tmp/ccc6IbbW.s:3525: Error: selected processor does not support ARM mode `strexeq r3,r1,[r0]'
Makefile:653: recipe for target 'dri.lo' failed
make[5]: *** [dri.lo] Error 1
make[5]: Leaving directory '/home/buildroot/buildroot/output/build/xserver_xorg-server-1.17.2/hw/xfree86/dri'
using this defconfig
BR2_arm=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_TOOLCHAIN_BUILDROOT_CXX=y
BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_EUDEV=y
BR2_PACKAGE_MESA3D=y
BR2_PACKAGE_MESA3D_DRI_DRIVER_SWRAST=y
BR2_PACKAGE_XORG7=y
BR2_PACKAGE_XSERVER_XORG_SERVER=y
/home/test/autobuild/instance-3/output/host/usr/aarch64-buildroot-linux-gnu/sysroot/usr/lib/libx264.so:
undefined reference to `x264_deblock_h_chroma_422_intra_neon'
Function "chroma_422_intra_neon" is defined in
x264/common/aarch64/deblock-a.S. The code in
x264/common/aarch64/* is only compiled if an
assembler was found:
http://git.videolan.org/?p=x264.git;a=blob;f=Makefile;h=6193c5936939fda36f9a2e00c734223318ccda14;hb=HEAD#l134
To allow this add BR2_aarch64 as an assembler-enabled arch.
This however breaks compilation due to broken NEON detection:
"no NEON support, try adding -mfpu=neon to CFLAGS
If you really want to run on such a CPU, configure with --disable-asm."
Due to recent changes in U-Boot, see
http://git.denx.de/?p=u-boot.git;a=commit;h=a26cd04920dc069fd6e91abb785426cf6c29f45f
re-creation of .config from defconfig by "make oldconfig" happened
incorrectly.
Default prompt for target selection was set as No that lead to missing platform
selection by automated scripts like that:
------------------>8------------------
"yes "" | make oldconfig" on defconfig
------------------>8------------------
And that lead to build failure:
------------------>8------------------
$ make
scripts/kconfig/conf --silentoldconfig Kconfig
CHK include/config.h
UPD include/config.h
GEN include/autoconf.mk
In file included from ./include/common.h:18:0:
include/config.h:5:22: fatal error: configs/.h: No such file or directory
#include <configs/.h>
^
compilation terminated.
scripts/Makefile.autoconf:72: recipe for target 'include/autoconf.mk' failed
make[1]: *** [include/autoconf.mk] Error 1
------------------>8------------------
This patch reverts mentioned change making possible to build U-Boot
2015.07 in Buildroot.
Note there's a probability that this particular patch won't make its way
in upstream U-Boot and better solution will be found there. But for now
we need to fix U-Boot building in Buildroot anyways.
Hopefully for the next release this patch won't be necessary.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Yann E. MORIN [Tue, 25 Aug 2015 21:43:33 +0000 (23:43 +0200)]
package/dropbear: add warning about volatile keys
When the rootfs is read-only, keys will be generated in a volatile
location, which is inherently bad as host keys will change on each boot,
rendering them virtually useless.
Add a warning so the user is at least aware of the issue.
Hide the rm output to avoid noisy output, now that we have a proper warning.
Move the starting message after the symlink-block, to avoid messages
collision. Move the umask as well, since /etc/dropbear/ may be world
readable; just the private host keys should be ?00 (and dropbear handles
that by itself).
[Peter: minor tweaks to commit message] Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Peter Korsgaard <jacmet@uclibc.org> Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
X.org xserver depends on libepoxy for glamor support, which depends on
EGL support. Mesa3d is not the only possible EGL provider therefore
change the dependency check.
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Sat, 29 Aug 2015 10:15:13 +0000 (12:15 +0200)]
package/setools: needs wchar
Even though setools itself does not require wide chars, the configure
script needs them when testing for C99 support.
Since:
- it is not clear whether wchar are mandatory for C99 or not,
- only uClibc may miss support for wide chars,
- using SELinux would probably involve rather big systems (i.e. glibc
would not be such a huge overhead),
just make setools depend on the toolchain having widechar.
Thomas Petazzoni [Sat, 29 Aug 2015 09:33:13 +0000 (11:33 +0200)]
ipkg: add patch to fix musl build
Like Romain Naour did for util-linux, this commit adds a patch for the
ipkg package to remove the use of the __P, which is useless for modern
compilers, and not defined by the musl C library.
Yann E. MORIN [Fri, 28 Aug 2015 22:34:41 +0000 (00:34 +0200)]
docs/website: remove broken symlink for top-level manual
Recently, we've tried to have the manual lie at the top-level of the
site, to get a shorter URL.
However, that is broken: we do not need only a link the the HTML, but
also to the CSS and to any resource linked from the html (images...).
The server configuration has now been updated to add a redirect, so we
no longer need that symlink anyway. Remove it.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Peter Korsgaard <jacmet@uclibc.org> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Fri, 28 Aug 2015 22:25:01 +0000 (00:25 +0200)]
docs/manual: makedevs only accepts numerical values
Because of a limitation in how makedevs resolves names [0], usernames
and groupnames would be resolved with UIDs and GIDs from the host, not
the target.
Fixing makedevs would involve us rewriting a /etc/passwd and /etc/group
parser (not too complex, still not trivial).
Document that the user and group fields should only be numbers, not
names. Also, the manual is currently misleading, as an example indeed
uses names for the user and group fields. Fix that and expand the
explanations.
The mode field should also be numbers, not a rwx combo.
Slight typographical eye-candy about major/minor fields.
[0] it uses getpwnma(3) and getgrnam(3) to resolve names to IDs, and
those only resolve names on the host (i.e. there is no way to specify
a chroot-like feature)
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Running configuration tests...
Failed to process makespec for platform 'devices/linux-buildroot-g++'
Project ERROR: Compiler <path_to_output_dir>/host/usr/bin/ccache <path_to_output_dir>/host/usr/bin/<cross_compile>-g++ not found. Check the value of CROSS_COMPILE -device-option
Could not read qmake configuration file <path_to_output_dir>/build/qt5base-5.5.0/mkspecs/devices/linux-buildroot-g++/qmake.conf.
Error processing project file: /dev/null
This was caused by Buildroot setting this in
qt5base-5.5.0/mkspecs/devices/linux-buildroot-g++/qmake.conf:
QMAKE_CXX = $${BR_CCACHE} $${CROSS_COMPILE}g++
But qt5base-5.5.0/mkspecs/features/device_config.prf expects QMAKE_CXX
to be a single valid (absolute or QMAKE_PATH_ENV-relative) path to an
existing file, which is not possible if using ccache as above.
Add a patch fixing this by testing only the first value in QMAKE_CXX.
Signed-off-by: Benoît Thébaudeau <benoit@wsystem.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Fri, 28 Aug 2015 13:03:34 +0000 (15:03 +0200)]
python-can: remove incorrect dependency
When using the python-package infrastructure, there is no need for
packages to declare a dependency on the Python interpreter: the
package infrastructure does it automatically.
Moreover, this is actually broken in the case of python-can, which can
be selected either with Python 2.x or Python 3.x. If the latter is
chosen, python-can will still trigger the build of Python 2.x, which
is incorrect.
This was discovered by Vicente Olivert Riera during the analysis of
the following build failure:
Fix the missing gstreamer1 build dependencies, which could possibly
prevent the configuration of qt5multimedia from detecting the supported
gstreamer1 features.
Fix the missing gstreamer1 install rules, which resulted in the
following runtime error:
defaultServiceProvider::requestService(): no service found for - "org.qt-project.qt.mediaplayer"
Signed-off-by: Benoît Thébaudeau <benoit@wsystem.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Wed, 26 Aug 2015 15:06:18 +0000 (17:06 +0200)]
qt5: disable for static-only builds
Even though we have some specific code to support building Qt5 for
static-only configurations, it doesn't work. The first problem is that
our custom qmake.conf always passes -ldl, which makes a number of Qt5
config.tests fail at configure time. Once this problem is fixed by
removing -ldl from QMAKE_LIBS and adding it to QMAKE_LIBS_DYNLOAD
instead, the next problem is that the plugin infrastructure of Qt5
assumes that Linux has dynamic library support: the qlibrary_unix.cpp
file includes <dlfcn.h>, and the only condition for this file to not
be included is:
Until recently, building Qt5 statically was working because our C
library was not built static-only: it provided <dlfcn.h> and
libdl.so. But now that we have a really static only toolchain, Qt5 no
longer builds.
The easiest solution is to simply make Qt5 depend on dynamic library
support.
Brendan Heading [Thu, 20 Aug 2015 22:32:13 +0000 (23:32 +0100)]
packages/linux-pam: disable when using musl
Fixes: http://autobuild.buildroot.net/results/e33/e337d69420ad00b2cc4017d639a31803926f2353/
linux-pam needs some surgery to build under musl, so for the time being
disable it until the issues are solved. Also disables dependent package
python-pam, and dependent suboptions under openvmtools, rsh-redone and
util-linux.
Signed-off-by: Brendan Heading <brendanheading@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Wed, 26 Aug 2015 06:44:30 +0000 (08:44 +0200)]
zyre: bump version to fix static linking problem
The last tagged release of zyre, v1.0.0, was made in May 2014. Since
then, they have switched to pkg-config to detect zmq and czmq, which
fixes static linking problems. However, they made a number of changes
to configure.ac, which make it difficult to backport just the relevant
changes.
Since we already had another backported fix, let's simply bump the
version of zyre to the latest available commit, which builds fine with
no change for shared and static scenarios, thanks to the use of
pkg-config.
An issue was opened upstream to ask them to tag a new release:
https://github.com/zeromq/zyre/issues/324.
Gustavo Zacarias [Tue, 25 Aug 2015 18:53:59 +0000 (15:53 -0300)]
openssh: security bump to version 7.1p1
Fixes:
CVE-2015-6563 - Fixed a privilege separation weakness related to PAM
support.
CVE-2015-6564 - Fixed a use-after-free bug related to PAM support that
was reachable by attackers who could compromise the pre-authentication
process for remote code exectuion.
CVE-2015-6565 - incorrectly set TTYs to be world-writable.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar> Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
package/ltrace: do not enable libunwind support for MIPS
Currently libunwind support in ltrace is broken for MIPS. I'm working
with upstream to fix this issue, but it's not yet ready, so let's
disable it by now.
Nathaniel Roach [Mon, 24 Aug 2015 15:44:16 +0000 (23:44 +0800)]
package/libmbim: Now needs libgudev under systemd
libmbim uses code (originally from udev) that has since been split
from the main systemd codebase into libgudev.
Fixes: http://autobuild.buildroot.org/results/638/638dbf05b785a276a33983b0237b7cad54777b85/
Tweak the package files for libmbim to require libgudev when building
with systemd.
Nathaniel Roach [Mon, 24 Aug 2015 15:44:14 +0000 (23:44 +0800)]
libgudev: New package
As libgudev recently was split from the main systemd/udev source, this
library is now required to build certain packages.
This library is only relevant to systemd, as the code it contains is
still contained in eudev.
[Thomas:
- don't show the dependency comment when systemd is not available,
since libgudev is anyway useless when you're not using systemd.
- fix the license, it's LGPLv2.1+ and not GPLv2+
- remove useless empty lines in the .mk file.]
Yann E. MORIN [Sat, 22 Aug 2015 17:53:13 +0000 (19:53 +0200)]
core/pkg-kconfig: fix configurators
Currently, the configurators are using $($(2)_MAKE_ENV), often derived
from $(TARGE_MAKE_ENV), as the environment to be set when calling the
various configurators.
This means that our host tools are used first, most notably pkg-config
(from host-pkgconf).
However, this is inherently flawed. Our pkg-config, when set for the
host, only searches .pc files in $(HOST_DIR) and never ever uses the
ones from the host. For example, since we do not build a host-qt, our
pkg-config would not find the host's QtCore.pc et al.
Consequently, on some systems (but not on others?) most of the
configurators fail to build, especially the latest kernel versions, as
they have been starting to use pkg-config two years ago.
Fix that by filtering-out sensible values out of the environment, but
only when calling the configurators.
[Thomas: rewrap comment to appropriate length.]
Reported-by: Mauro Condarelli <mc5686@mclink.it> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Cc: Mauro Condarelli <mc5686@mclink.it> Tested-by: Mauro Condarelli <mc5686@mclink.it> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Samuel Martin [Mon, 24 Aug 2015 12:47:12 +0000 (14:47 +0200)]
package/vlc: update opencv support
vlc supports both opencv-2.4 and opencv-3, so adjust the vlc package to
reflect this.
Cc: Jonathan Ben Avraham <yba@tkos.co.il> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Samuel Martin <s.martin49@gmail.com> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The dependency is changed from a 'select' to a 'depends on' to avoid a
circular dependency caused by the introduction of OpenCV-3. This means
we can also drop the threads and C++ dependencies, since they are now
inherited via the depends on OpenCV.
Cc: Jonathan Ben Avraham <yba@tkos.co.il> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Samuel Martin <s.martin49@gmail.com>
[yann.morin.1998@free.fr: fix dependencies] Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Jonathan Ben Avraham <yba@tkos.co.il> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Samuel Martin <s.martin49@gmail.com>
[yann.morin.1998@free.fr:
- remove legacy symbols, now
- make opencv3 depends on !opencv, not the other way around
- slitghly reword the commit log (opencv/opencv3 dependency)
] Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Our current apitrace version can't detect the host-python version
correctly, so if both host-python and host-python3 where installed, it
will take the last one and it will fail with an "invalid syntax" error.
The latest apitrace version has this problem solved and it detects the
host-python version correctly.
Yann E. MORIN [Tue, 18 Aug 2015 19:25:53 +0000 (21:25 +0200)]
package/dropbear: fix when readlink is from busybox
Busybox "readlink -f" does not canonicalise paths when the target is
missing, while coreutils do.
Fix that by:
- making an absolute symlink
- dropping "-f" when calling readlink
Fixes #8276.
Reported-by: Jason Tang <tang@jtang.org> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Paul Cercueil <paul@crapouillou.net> Cc: Alexandre Belloni <alexandre.belloni@free-electrons.com> Tested-by: Jason Tang <tang@jtang.org> Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Yann E. MORIN [Fri, 21 Aug 2015 21:58:29 +0000 (23:58 +0200)]
linux: enable all options needed by xtables-addons
Both of CONFIG_NF_CONNTRACK and CONFIG_NF_CONNTRACK_MARK are needed by
xtables-addons.
Although the current code does enable them in the linux' .config file,
the former is protected behind CONFIG_NETFILTER_ADVANCED, which may be
missing from a user-supplied (def)config file, and is missing from some
of the bundled defconfigs as well.
For example, the following defconfig fails to build:
Yann E. MORIN [Sun, 23 Aug 2015 18:58:01 +0000 (20:58 +0200)]
package: add comment about ipkg/opkg/rpm
Too often the question is raised, that ipkg, opkg and rpm do not work,
and users complain they can not install packages.
Even though we do have a clear and clearly explained section in our
manual, people do not read it (when will users read manuals? sigh...).
So, add a big fat comment about ipkg/opkg/rpm, that Buildroot does not
generate binary packages and does not provide any package database for
any of those package manager.
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> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Wed, 19 Aug 2015 23:20:14 +0000 (01:20 +0200)]
package/librtas: only show comment on supported architectures
librtas is only available for PPC and PPC64, so only show the comment
about the (e)glibc dependency for those archs.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Sam Bobroff <sam.bobroff@au1.ibm.com> Reviewed-by: Sam Bobroff <sam.bobroff@au1.ibm.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Wed, 19 Aug 2015 18:33:34 +0000 (20:33 +0200)]
libselinux: mark as not available on ARC
libselinux causes some build problems due to the toolchain on ARC,
which haven't been solved so far. As a temporary solution for Buildroot
2015.08, this commit makes libselinux (and its reverse dependencies)
unavailable on ARC. Of course, once the toolchain problem is
addressed, this commit can be reverted to re-enable libselinux on ARC.
Thomas Petazzoni [Wed, 19 Aug 2015 18:26:52 +0000 (20:26 +0200)]
boost: fix build on ARC
This commit adds a patch to Boost to make it use the eventfd()
function provided by the C library when uClibc is used, rather than
falling back to using directly the __NR_eventfd system call. This
fixes the build on ARC, which doesn't define __NR_eventfd.
The original problem is that uClibc pretends to be glibc 2.2, which
didn't had eventfd(), so Boost makes the system call
manually. uClibc-ng, in its next release, will pretend to be glibc
2.10 (see
http://cgit.uclibc-ng.org/cgi/cgit/uclibc-ng.git/commit/?id=4ff3a6c8eb91db71d6dc3d2932b66e848bd20ac3),
which will also fix the problem, but requires bumping the uClibc
version, rebuilding the external toolchains, and so on.
Ideally, Boost should be doing a compile test to detect if eventfd()
is available or not, but the Boost build system is so brain-damaged
that doing so would require way too much effort.
[Peter: minor tweaks to description] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Thomas Petazzoni [Wed, 19 Aug 2015 20:54:45 +0000 (22:54 +0200)]
mono: depend on shared library support
The mono source code includes <dlfcn.h>, which is only available when
shared library support is available. While it might be possible to do
a fully static installation of Mono, it probably isn't very useful.
While we're at it, this commit also makes sure that the Config.in
comment is not visible when the architecture doesn't support Mono.
Matthew Shyu [Wed, 19 Aug 2015 09:48:15 +0000 (17:48 +0800)]
qt5base: install libQt5XcbQpa.so for xcb backend
When the XCB backend is selected, the libqxcb.so plugin is installed,
and is linked against libQt5XcbQpa.so. However, until, Buildroot was
not installing this library. This commit fixes this.
[Thomas: tweak commit log.]
Signed-off-by: Matthew Shyu <matthew.shyu@amlogic.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Wed, 19 Aug 2015 11:08:45 +0000 (13:08 +0200)]
libpthsem: fix setjmp/longjmp detection
libpthsem has been affected since quite a while by build issues, but
occuring only on Nathaniel Roach's autobuilder. The failure message
is:
error: #error "Unsupported Linux (g)libc version and/or platform"
This message comes from pth_mctx.c file, which implements five methods
for machine context initialization:
VARIANT 1: THE STANDARDIZED SVR4/SUSv2 APPROACH
VARIANT 2: THE SIGNAL STACK TRICK
VARIANT 3: LINUX SPECIFIC JMP_BUF FIDDLING
VARIANT 4: INTERACTIVE SPECIFIC JMP_BUF FIDDLING
VARIANT 5: WIN32 SPECIFIC JMP_BUF FIDDLING
The "Unsupported (g)libc version and/or platform" only appears when
"VARIANT 4" is used, since VARIANT 4 only supports a very limited
number of platforms. So when building with Nathaniel's autobuilder,
VARIANT 4 is chosen.
However, when you build libpthsem on some other machine than
Nathaniel's autobuilder, VARIANT 2 is chosen, and works regardless of
the glibc version or architecture.
On both Nathaniel's autobuilder, and on a different machine, the
PTH_MCTX_MTH macro gives sjlj:
#define PTH_MCTX_MTH_use PTH_MCTX_MTH_sjlj
However, on a "normal" machine, the PTH_MCTX_DSP macro gives ssjlj:
#define PTH_MCTX_DSP_use PTH_MCTX_DSP_ssjlj
While on Nathaniel's autobuilder, it gives:
#define PTH_MCTX_DSP_use PTH_MCTX_DSP_sjljlx
This explains why VARIANT 4 is being used on Nathaniel's autobuilder,
while VARIANT 2 is used when building on other platforms.
The decision of the value for PTH_MCTX_DSP is derived as follows in
configure.ac:
AC_CHECK_SJLJ(sjlj=yes, sjlj=no, sjlj_type)
[...]
elif test ".$sjlj" = .yes; then
mctx_mth=sjlj
mctx_dsp=$sjlj_type
[...]
AC_DEFINE_UNQUOTED(PTH_MCTX_DSP_use, [PTH_MCTX_DSP_$mctx_dsp], [define for machine context dispatching])
So basically, the value of PTH_MCTX_DSP is $sjlj_type, as returned by
the AC_CHECK_SJLJ autoconf macro, implemented in
acinclude.m4. However, reading this macro is quite informative: it
does a number of tests that are not cross-compile
friendly. Especially, it looks at the kernel version with 'uname -r'
to decide whether the Linux system is braindead or not. If the system
runs a 2.2.x kernel or newer 2.x, or a 3.x kernel, everything is fine,
the system is not braindead, and sjlj_type is set to ssjlj. However,
if the build system runs a 4.x kernel, then it is considered as
braindead, and sjlj_type is set to sjljlx.
And indeed, Nathaniel's autobuilder is running a 4.x kernel, while all
other autobuilders run 2.x or 3.x kernels.
Since for all sane Linux systems, this AC_CHECK_SJLJ macro concludes
that the setjmp/longtmp type is ssjlj, this commit takes the simplest
route of forcing this value, skipping the broken detection.
Note that we're overriding ac_cv_check_sjlj instead of using the
--with-mctx-* options, since the latter do not work properly in the
context of Nathaniel's autobuilder, as the broken cross-compilation
tests continue to cause problems.
Vivien Didelot [Tue, 11 Aug 2015 21:23:50 +0000 (17:23 -0400)]
fs: iso9660: change boot menu entry text
Replace Grub-specific "menu.lst" with "menu config" in the
BR2_TARGET_ROOTFS_ISO9660_BOOT_MENU Kconfig entry text, and mention
missing grub.cfg for Grub 2.
Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Tue, 18 Aug 2015 19:23:18 +0000 (21:23 +0200)]
package/protobuf: needs dynamic libraries
Eventhough it should be theoretically possible to build protobuf in
static-only, configure.ac includes an m4 macro, ACX_PTHREAD defined in
m4/acx_pthread.m4, which forcibly checks for threads *with* shared libs,
and is completely broken for static-only (as it forces -shared whatever
the user selection), ending up with these configure results:
checking for the pthreads library -lpthreads... no
checking whether pthreads work without any flags... no
checking whether pthreads work with -Kthread... no
checking whether pthreads work with -kthread... no
checking for the pthreads library -llthread... no
checking whether pthreads work with -pthread... yes
checking for joinable pthread attribute... PTHREAD_CREATE_JOINABLE
checking if more special flags are required for pthreads... no
checking whether to check for GCC pthread/shared inconsistencies... yes
checking whether -pthread is sufficient with -shared... no
checking whether -lpthread fixes that... no
checking whether -lc_r fixes that... no
configure: WARNING: Impossible to determine how to use pthreads with shared libraries
checking whether what we have so far is sufficient with -nostdlib... no
checking whether -lpthread saves the day... no
configure: WARNING: Impossible to determine how to use pthreads with shared libraries and -nostdlib
Fixing this macro is far from trivial; protobuf in a static-only
scenario is probably not too common. So, just disable protobuf for
static-only builds.
Yann E. MORIN [Tue, 18 Aug 2015 19:23:17 +0000 (21:23 +0200)]
package/protobuf-c: threads dependency is a direct dependency
Currently, protobuf-c's dependency on threads is labelled as being
inherited from protobuf.
This is wrong, as protobuf-c does not depend on protobuf, and such
dependency was removed in e16865a (protobuf-c: Don't require protobuf on
target), but forgot to remove the corresponding comment.
Remove it now.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The problem is caused by the -fvisibility-inlines-hidden switch.
Removing the switch is probably the least intrusive way we can make the
problem go away. The first solution that was considered was to move the
definition of the offending method to the .cpp file. However, with
other combinations of compilers and platforms, I suppose it could happen
with other methods as well. Removing the switch ensures we catch them
all.
Built-tested with the config from the build bot, as well as with all OLA
options/plugins enabled.
Signed-off-by: Simon Marchi <simon.marchi@polymtl.ca> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Viallard Anthony [Thu, 13 Aug 2015 12:36:01 +0000 (14:36 +0200)]
msmtp: don't try to use IDN feature
autoconf can't detect if AI_IDN is really supported
(http://sourceforge.net/p/msmtp/code/ci/907916e8089ea9303a7c091b763087530c88a063/),
so a --disable-gai-idn flag was added to explicitly disable the usage
of IDN.
Before this patch, msmtp failed to send mail with the following
output:
msmtp: cannot locate host smtp.mandrillapp.com: Bad value for ai_flags
[Thomas: tweak commit log.]
Signed-off-by: Anthony Viallard <viallard@syscom-instruments.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Christian Stewart <christian@paral.in> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Thu, 13 Aug 2015 12:13:23 +0000 (14:13 +0200)]
toolchain/external: ensure gcc version is known
Currently, when a preconfigured prebuilt toolchain forgets to specify
its gcc version, the error message is a bit misleading, like:
Incorrect selection of gcc version: expected .x, got 4.9.2
Add a an explicit check for the gcc version being set, that reports a
better error message.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Thu, 13 Aug 2015 13:48:21 +0000 (15:48 +0200)]
package/systat: fix linking with -lintl
When linking, libraries should be listed in the reverse-dependency order,
that is the library/ies with no dependency last, preceded by libraries
that depend on them.
And most importantly, libraries should be listed *after* the local
objects, .o files.
Currently, sysstat uses CFLAGS to shoehorn the -lintl linker flags,
which is incorrect, as the CFLAGS are passed before the object files.
However, sysstat has a LFLAGS variable that, presumably, is to be used
to pass linker flags. This is where we want to shoehorn -lintl.
[Thomas: keep only the locally calculated sha256 hash, and do not
mention the commented SF.net hash, since SF.net is no longer the
official upstream location.]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Baruch Siach <baruch@tkos.co.il> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jörg Krause [Thu, 13 Aug 2015 20:16:38 +0000 (22:16 +0200)]
package/libpfm4: disable Werror
Set `DBG` to an empty value to disable -Werror when building libpfm4. Build
aborts with a musl toolchain because of warnings about redirecting incorrect
header includes.
So -Werror shouldn't be used in released code since it can cause random build
failures on moderate warnings. It also depends on the used toolchain since
different toolchains may or may not print the same warnings.
Yann E. MORIN [Thu, 13 Aug 2015 21:57:45 +0000 (23:57 +0200)]
package/audit: not available for musl
audit uses strndupa() which is missing from musl.
Even though the implementation of strndupa is not too complex, we won't
go as far as duplicating it in audit, and we just disable audit for the
musl C library.
Signed-off-by: Alex Suykov <alex.suykov@gmail.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Sun, 16 Aug 2015 21:46:38 +0000 (23:46 +0200)]
package/oprofile: fix static linking
When getext is enabled, popt links with -lintl, specifies it in its
popt.pc and has done so for the past 6+ years. But oprofile does not
use pkconfig to find popt, so misses -lintl, which is important for
a static build. We have to do the call to pkgconfig manually...
Noé Rubinstein [Mon, 17 Aug 2015 07:10:44 +0000 (09:10 +0200)]
pkg-kernel-module: die if kernel module support is disabled
Test the configuration of the kernel to see if loadable module support
is enabled, and error out otherwise. This makes build failures of
external kernel modules less confusing.
[Thomas: tweak the commit log.]
Signed-off-by: Noé Rubinstein <nrubinstein@aldebaran.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Fri, 31 Jul 2015 08:23:20 +0000 (10:23 +0200)]
fs/common: add <fs>_POST_GEN_HOOKS support
The common filesystem infrastructure already supports a
<fs>_PRE_GEN_HOOKS variable, which allows filesystem makefiles to
register some actions to be done before the root filesystem image is
generated.
This commit adds a similiar <fs>_POST_GEN_HOOKS variable, which will
allow filesystem makefiles to do some actions after the filesystem
image has been generated. It will initially be used by the iso9660
filesystem to delete the temporary directory it creates.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Noé Rubinstein <nrubinstein@aldebaran.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Alexey Brodkin [Mon, 10 Aug 2015 13:36:36 +0000 (16:36 +0300)]
tinc: disable PIE on ARC
Even though ARC gcc understands "-pie" option and attempts to generate
PIE binaries as of today PIE is not really supported for user-space
applications.
So we disable PIE detection if building for ARC.
That fixes http://autobuild.buildroot.net/results/2b1/2b1aa474fc432f3f79b03c4577d68642bb7e0f9c
and also prevents execution of non-supported PIE binary in runtime.
[Thomas: fix indentation.]
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
projects / coffee / buildroot.git / log