When atomic intrisics are missing, libfastjson falls back to using
pthread mutexes to manage atomicity. Of course, this is much less
efficient than atomics, but it does the job.
Propagate the new dependency to rsyslog, the sole user of libfastjson.
Note: rsyslog already depends on threads for itself, but we believe it
is better to have the exact same dependency propagated.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
libfastjson is supposed to use the __sync_*4 atomic intrinsics, but alas
it is not using them because their ./configure decides they are not
available: it uses AC_TRY_RUN to check for them, and the default is to
decide they are not available, because of cross-compilation.
Besides, one of the source files was not including the generated
config.h, so even after fixing ./configure there was still a build
error.
The first patch is a backport from upstream to fix the missing
inclusion.
The second patch is switching AC_TRY_RUN over to AC_LINK_IFELSE, as the
only thing we're interested in is to check for the presence of the
atomic intrisics, and linking is enough for that.
Fixes:
http://autobuild.buildroot.org/results/192/1923d0b570adba494f83747a9610ea6ec35f5223/http://autobuild.buildroot.org/results/23a/23ac0e742ed3a70ae4d038f8c9eadc23e708f671/
and many others...
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Thompson <peter.macleod.thompson@gmail.com>
[Thomas:
- point <pkg>_LICENSE_FILES to a file that actually exists
- add host-pkgconf to the dependencies
- remove unneeded CONF_OPTS options
- pass FREETYPE_CONFIG in the environment to point to freetype-config]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Thompson <peter.macleod.thompson@gmail.com>
[Thomas:
- remove all sub-options to select the various image formats. Between
no formats enabled and all formats enabled, the size difference of
the library is ~30 KB, so it really isn't worth having all those
sub-options:
-rwxr-xr-x 1 thomas thomas 22444 juil. 15 15:51 libSDL2_image-2.0.so.0.0.1
-rwxr-xr-x 1 thomas thomas 59216 juil. 15 15:52 libSDL2_image-2.0.so.0.0.1
consequently, we simply enable unconditionally all image formats
that don't have any dependency, and use "automatic" dependencies
for the 4 formats that require an external library
- remove the host package variant, as it isn't used anywhere.
- remove --with-sdl2-prefix and --with-sdl2-exec-prefix, and instead
add a dependency on host-pkgconf so that pkg-config is used
- remove --disable-static, this is handled by the autotools-package
infrastructure already
- point <pkg>_LICENSE_FILES to an existing file.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Thompson <peter.macleod.thompson@gmail.com>
[Thomas:
- Rewrap Config.in help text.
- Remove unused SDL2_GFX_VERSION_MAJOR variable
- Add host-pkgconf to the dependencies, since pkg-config is used to
find sdl2. This fixes the build, otherwise -I/usr/include/sdl2 was
added to the CFLAGS when sdl2-config was found on the host machine.
- Remove --with-sdl2-prefix, since this option doesn't exist, and
using pkg-config is a better solution (--with-sdl-prefix exists,
but is not useful when pkg-config is available).
- Remove --enable-static, packages are not supposed to pass such an
option, it's not automatically by the autotools package
infrastructure.
- Pass --disable-sdltest instead of --disable-sdl2test since the
latter doesn't exist, while the former does.
- Pass SDL2_GFX_AUTORECONF = YES since the configure/Makefile.in are
not up-to-date and therefore it tries to use aclocal at build time.
- Adjust _LICENSE_FILES, since there is no file named 'LICENSE', use
the smallest source file instead, since it contains the license
text.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Some old binary blobs drivers only implement an old VIDEODRV ABI. This
is the case for the AMD Catalyst driver for example.
Such a situation already exists with the nvidia-tegra23, that only
support the VIDEODRV ABI 14.
Since VIDEODRV ABIs are not backward compatible [0], lets introduce an
older Xserver version that supports such an old ABI.
0. https://www.x.org/wiki/XorgModuleABIVersions/
Signed-off-by: Romain Perier <romain.perier@free-electrons.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The cyclictest program doesn't build with musl toolchains due to
several issues with a different implementation of the "struct sigevent"
from Glibc/uClibc and musl (similar to [1]).
The pi_stress program doesn't build with musl toolchains with the
following error since musl doesn't provide _POSIX_THREAD_PRIO_INHERIT
define even if musl have Priority Inheritance mutexes [2]:
src/pi_tests/pi_stress.c:78:2: error: #error "Can't run this test without PI Mutex support"
#error "Can't run this test without PI Mutex support"
Since the fix for cyclictest is not trivial, disable rt-tests for musl
toolchains.
Fixes:
http://autobuild.buildroot.net/results/c27/c272ef7df6190cbb688e6db0b1ee210e6fbb472c
[1] https://bugs.mysql.com/bug.php?id=80322
[2] http://www.openwall.com/lists/musl/2016/03/18/6
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
* Remove fetching upstream patch to fix build without MTD support
* Update .config
* Update help text as SSL support is now needed for other features
than the webserver too
Signed-off-by: Jörg Krause <joerg.krause@embedded.rocks>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
This reverts commit b6ff742ca0.
This bump causes numerous build failures, so Vicente and Yann proposed
to revert it.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Fixes:
GNUTLS-SA-2016-2 - vulnerability that affects certificate verification
when GnuTLS is used in combination with the p11-kit trust module.
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
As described at:
4520524ba0
this commit continues a series of updates of ARC tools.
This engineering build fixes the kernel dwarf stack unwinder feature for
ARC targets.
We still keep GDB as it is of arc-2016.03 release because there're some
issues we'd like to resolve before releasing it to wider audience.
So again note this is next engineering builds of arc-2016.09 series
and it might have all kinds of breakages, please don't use it for
production builds.
Related to:
4520524ba0
Signed-off-by: Vlad Zakharov <vzakhar@synopsys.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
syslog-ng is built with the default buildroot flag localstatdir=/var
This causes three issues:
1) syslog-ng creates two files (syslog-ng.ctl and syslog-ng.persist in /var)
which shouldn't have random files generated in it.
2) SELinux expects it in it's default directory /var/run.
3) This breakes read only file systems because /var/ usually isn't mounted as
a seperate filing system.
Setting the buildflag to the default /var/run
fixes all three of thee of these issues.
Signed-off-by: Adam Duskett <aduskett@codeblue.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bump U-Boot version to 2016.07 and also remove the extra patch
as it is already part of U-Boot mainline now.
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bump U-Boot version to 2016.07 and also remove the extra patch
as it is already part of U-Boot mainline now.
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
- drop 0001-native-affinity.patch as it was replaced upstream [1]. Autoreconf
could be dropped
- drop conf env ac_cv_file__proc_stat=yes ac_cv_file__proc_meminfo=yes because
these tests no longer run when cross-compiling [2]
- upstream created a custom check function to use ncurses*-config [3] but it
does not allow to override the path to ncurses*-config, leading to a build
failure as the htop build system would search the path of host tools.
A new patch 0001-Allow-to-override-ncurses-config-path.patch allows this
override [4]. It brings back the need to autoreconf
- set path to ncurses5-config in conf env in order to avoid the htop build
system finding it in the path of host tools
[1] dfad0afb36
[2] b561956637
[3] 96c929f82b
[4] 666f12f60f
Signed-off-by: Ricardo Martincoski <ricardo.martincoski@gmail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
