Danomi Manchego [Wed, 23 Nov 2016 02:23:02 +0000 (21:23 -0500)]
Makefile: fix distclean removal of $(O)
The distclean target no longer removes the "output" directory for
in-tree builds, because $(O) is no longer just "output" in that
case. Change the test to be against "$(CURDIR)/output", to match
the O setting, and a similar test elsewhere in the same Makefile.
Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Max Filippov [Wed, 16 Nov 2016 23:26:01 +0000 (15:26 -0800)]
package/e2fsprogs: disable use of magic.h
RHEL 5.x does have magic.h, but it does not define all expected symbols.
In particular, the NO_CHECK symbols were only added in file 4.20 and RHEL
5.x is using 4.17. Don't use magic.h and libmagic when building host
package.
Suggested-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
By default, qextserialport will attempt to build a shared library.
qesp_static has to be set in CONFIG to build static. Note that
static+shared is not supported, in that case we just build shared.
The install target commands also have to be gated in that case,
because the *.so files don't exist. For completeness we both set
QEXTSERIALPORT_INSTALL_STAGING to NO and don't define
QEXTSERIALPORT_INSTALL_TARGET_CMDS for static builds, although one
of them would be sufficient.
qextserialport: use 'make install' to install to staging
qmake hardcodes the path to the sysroot in the install commands, so
we can't use it for target. But it's perfectly usable for staging.
By using 'make install', we get:
- the extserialport.prf file is installed to the mkspecs directory and
libqextserialport.prl in staging/usr/lib, so qmake can do its magic
and add the necessary compiler options;
- it also works for static build, when *.so files don't exist.
The QExtSerialPort and qextserialport.pc files are created by
Buildroot so they still have to be installed explicitly. Note that
upstream installs in the QtExtSerialPort directory, not QExtSerialPort,
so we follow that decision.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
[Thomas: also adjust 0003-pkgconfig.patch to fix the header path.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Fabio Estevam [Sun, 20 Nov 2016 18:03:00 +0000 (16:03 -0200)]
configs/imx28evk: Bump to kernel 4.8.8
Kernel 4.8.8 brings commit 235bde1ed3f0ff ("net: fec: Call swap_buffer()
prior to IP header alignment"), which makes Ethernet functional again
on mx28.
Bump to this version.
Signed-off-by: Fabio Estevam <festevam@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The wscript checks the following.
'func': check_libs(['atomic'],
check_statement('stdint.h',
'int64_t test = 0;'
'test = __atomic_add_fetch(&test, 1, __ATOMIC_SEQ_CST)')),
and
'func': check_statement('stdint.h',
'int64_t test = 0;'
'__typeof__(test) x = ({int a = 1; a;});'
'test = __sync_add_and_fetch(&test, 1)'),
It fails if neither of these are available.
mplayer uses a homegrown configure script that calls 'gcc -v | tail -1'
to find the gcc version. However, our toolchain wrapper adds
-Wl,-elf2flt to the gcc arguments. This triggers gcc to do an actual
compile (rather than just printing version info and exiting), and that
compile fails because of the missing definition of main(). The last
line of the output is therefore:
collect2: error: ld returned 1 exit status
and configure doesn't find a version in there. It then falls back to
using the host gcc instead, which breaks things completely a bit
further down the line.
Since mplayer probably isn't very useful on noMMU architectures anyway,
just disable it completely for FLAT.
For powerpc (and powerpc only), the openblas build system will enable
USE_OPENMP when threads are enabled. But the toolchain wasn't
necessarily built with OpenMP support. So explicitly disable OpenMP -
for all architectures, not just powerpc, just to keep things simple.
We could have autodetected whether OpenMP is available at configure
time, but that would be more of a feature patch, it's up to upstream
to do that.
libuv uses pthread_barrier_* functions, which aren't available with
linuxthreads. Notice that libuv contains a local prototype for these
functions, so libuv is able to build and the error only triggers when
applications try to link against it.
Also propagate this dependency to the reverse deps of libuv.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
xapp_xload tries to include protocols/rwhod.h, which isn't available on
musl. Fortunately, the package also has stub code that can be enabled
with -DRLOADSTUB (this is done e.g. on Cygwin).
This should probably be detected in configure, but that's a lot more
work.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
taskd: add patch for correct gnutls libraries in static build
The CMakeLists.txt was using the wrong variables names to add to
includes and libs, so the additional libraries for static build
weren't added to the link command.
Patch sent upstream to taskwarrior-dev@googlegroups.com but it doesn't
seem to be very active.
Gaël PORTAY [Mon, 21 Nov 2016 16:15:29 +0000 (11:15 -0500)]
fs/ext2: fix double quoted label
Since the commit 6dd7bbb59134799ed3d7343f238b3b02592faebf, the label does
not need anymore to be quoted. Even worse it *must* not be simple-quoted,
unless the label will contain the double-quotes from the config variable
BR2_TARGET_ROOTFS_EXT2_LABEL.
The commit mentionned above has replaced echo by printf:
- echo "$$(ROOTFS_$(2)_CMD)" >> $$(FAKEROOT_SCRIPT)
+ $$(call PRINTF,$$(ROOTFS_$(2)_CMD)) >> $$(FAKEROOT_SCRIPT)
Since this commit the rootfs label contains extra double-quotes.
$ blkid
/dev/mmcblk0: LABEL=""BR 2016.08"" UUID="xxx"
^ ^
With this fix, the extra double-quotes have disappeared:
/dev/mmcblk0: LABEL="BR 2016.11-rc2" UUID="yyy"
Signed-off-by: Gaël PORTAY <gael.portay@savoirfairelinux.com> Acked-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jerry Evans [Sun, 20 Nov 2016 13:28:36 +0000 (13:28 +0000)]
docs/manual: fix documentation about BR2_EXTERNAL_<name>_PATH variable
The BR2_EXTERNAL documentation improperly tells to use the
BR2_EXTERNAL_<name>_DIR variable to reference the location of the
BR2_EXTERNAL directory. But in fact the variable is named
BR2_EXTERNAL_<name>_PATH.
In addition, some closing double quotes were missing.
This commit fixes both of those issues.
Signed-off-by: Jerry Evans <g4@novadsp.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
kvm-unit-tests: only enable for ARM on Cortex-A{7,12,15,17}
The library uses the hypervisor instruction HVC, which is only
available on ARMv7-A with Hypervisor extensions. This is limited to
Cortex-A{7,12,15,17}.
In addition the build unconditionally passes -marm so it also needs
ARM instructions to be available, but that is already implied by the
CPU selection.
Implicitly fixes
http://autobuild.buildroot.net/results/53d109fd9055fd20387bb857aced5f89cf3086fd
though it is still not clear why ld doesn't accept the -Ttext=...
option there.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
[Thomas: A12 and A17 also have the virtualization extensions.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
kvm-unit-tests: pass the --processor configure argument
The configure script uses this to set the -mcpu option. If nothing is
passed, it defaults to -mcpu=cortex-a15 for arm, and -mcpu=cortex-a57
for aarch64, which may conflict with the actual architecture.
The --processor option is only really used for arm and aarch64, but it
doesn't hurt to also pass it for the other arches.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Francois Perrad [Sun, 6 Nov 2016 23:14:08 +0000 (00:14 +0100)]
configs/olimex_a20_olinuxino_lime_mali: bump kernel version
This commit changes the kernel version used by the
olimex_a20_olinuxino_lime_mali in order to use a version that builds
fine with gcc 5.x. This kernel repository is listed in
https://linux-sunxi.org/Linux#External_links as a possible replacement
for the aging 3.4 sunxi kernel.
Signed-off-by: Francois Perrad <francois.perrad@gadz.org> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Gustavo Zacarias [Thu, 17 Nov 2016 19:41:25 +0000 (16:41 -0300)]
mesa3d: add patch to fix wayland-scanner issues
mesa3d tries to use the pkg-config specified wayland-scanner when
wayland is enabled, which points to /usr/bin/wayland-scanner which is
incorrect.
The correct fix is trying to use the PATH wayland-scanner first if it's
available rather than the pkg-config specified one, since tweaking a
staging .pc file with a host binary is wrong. Fixes:
http://autobuild.buildroot.net/results/c0c/c0ca3778c390a33c27073553ca3bef7ac7c5784b/
Patch sent upstream:
https://lists.freedesktop.org/archives/mesa-dev/2016-November/135665.html
Signed-off-by: Gustavo Zacarias <gustavo@zacarias.com.ar> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Romain Naour [Tue, 15 Nov 2016 23:22:18 +0000 (00:22 +0100)]
package/xenomai: bfin remove inline keyword
Xenomai user space tools fail to build for bfin target since it's
provide it's own pthread_atfork(), shm_open() and shm_unlink()
definition using the inline keyword and weak attribute.
With gcc 5+ the weak attribute is discarded by the inline keyword,
so each symbol are global and are redefined several time while
linking.
Instead of patching configure.ac we can just add the missing
libraries to LIBS. Tested with m68k/arm static builds.
This makes the patch obsolete and the autoreconf can be avoided.
Jörg Krause [Tue, 15 Nov 2016 20:46:58 +0000 (21:46 +0100)]
package/tremor: fix build with Thumb only ARM archs
Some ARM architectures like ARMv7-M only supports Thumb instructions,
but the tremor build configuration enables ARM assembly code
unconditionally for all arm triplets by defining _ARM_ASSEM_.
We are overriding this by undefining this macro for ARM architectures
not supporting ARM instructions.
Zakharov Vlad [Tue, 15 Nov 2016 15:21:31 +0000 (18:21 +0300)]
gdb: arc: remove C++ dependency and disable build with C++
gdb arc-2016.09-rc1 is based on upstream 7.12 gdb version where we can
still disable C++ build. For more information take a look at:
https://sourceware.org/gdb/wiki/cxx-conversion
We wanted to build gdb with C++ as it is used as default in 7.12
and is even a mandatory requirement since 2016/09.
But unfortunately we missed one runtime failure that C++ build causes
and so now we have to disable C++ build.
When gdb arc-2016.09-rc1 is built as a C++ application it segfaults at
runtime for ARC.
We are going to fix the issue in 2017.03 ARC toolchain release. and
before this we remove C++ dependency are adding temporary workaround
to prevent runtime segfaults.
The workaround is to disable building gdb as C++ application via
passing --disable-build-with-cxx config option when building gdb for
ARC.
Signed-off-by: Vlad Zakharov <vzakhar@synopsys.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Peter Seiderer [Mon, 14 Nov 2016 23:16:40 +0000 (00:16 +0100)]
qt5base: fix eglfs compile for odroid-mali
Avoid duplicated struct fbdev_window definition (introduced by [1]) by
renaming struct fbdev_window to shadow_fbdev_window.
Fixes [2]:
qeglfsmaliintegration.cpp:45:8: error: redefinition of 'struct fbdev_window'
struct fbdev_window {
^
In file included from /accts/mlweber1/rc-buildroot-test/scripts/instance-0/output/host/usr/aarch64-buildroot-linux-gnu/sysroot/usr/include/EGL/eglplatform.h:28:0,
from /accts/mlweber1/rc-buildroot-test/scripts/instance-0/output/host/usr/aarch64-buildroot-linux-gnu/sysroot/usr/include/EGL/egl.h:36,
from ../../../eglfs/qeglfsglobal.h:45,
from ../../../eglfs/qeglfsdeviceintegration.h:48,
from qeglfsmaliintegration.h:37,
from qeglfsmaliintegration.cpp:34:
Baruch Siach [Mon, 14 Nov 2016 21:58:30 +0000 (23:58 +0200)]
openjpeg: refresh patches
The 2.1.2 version bump did not take into account the recently added fixes.
Update the patches to apply based on upstream pull requests #866 and #867.
Cc: Peter Seiderer <ps.report@gmx.net> Cc: Samuel Martin <s.martin49@gmail.com> Signed-off-by: Baruch Siach <baruch@tkos.co.il> Acked-by: Peter Seiderer <ps.report@gmx.net> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Max Filippov [Mon, 14 Nov 2016 10:05:41 +0000 (02:05 -0800)]
package/gcc: backport xtensa trap pattern for 5.x
gcc 5.x uses trap pattern extensively to mark locations where program
execution should never get. Default pattern implementation emits a call
to 'abort' function. This function however is not always available, e.g.
it's not available for the libc dynamic linker. Backport implementation
of the trap pattern for xtensa that does not result in a call to
'abort'.
Thomas Petazzoni [Mon, 14 Nov 2016 21:20:36 +0000 (22:20 +0100)]
dtv-scan-tables: rename file to have only ASCII characters
Since the bump of dtv-scan-tables to version ceb11833b35f05813b1f0397a60e0f3b99430aab in commit b1c8794d8ac0eb3895d13ae91d8e912ec469a105, one file contains non-ASCII
characters, which causes encoding issues tvheadend. Since no other
file in the dtv-scan-tables code base contains files with non-ASCII
characters (despite having files named after cities in various
countries that definitely do have non-ASCII characters), we rename
this file so that it is named with only ASCII characters.
This fixes the build of tvheadend, which was failing when the host
Python interpreter was python3, due to a file name encoding issue.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Reviewed-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Zakharov Vlad [Mon, 14 Nov 2016 17:56:37 +0000 (20:56 +0300)]
toolchain: Bump ARC tools to arc-2016.09-rc1
As described at: 4520524ba055706236db9f00dd79f1b2e2e87fde
this commit continues a series of updates of ARC tools.
This time we're updating tools to arc-2016.09-rc1.
Other important change is that we also update gdb. Now we are
using gdb 7.12.
This version of gdb requires C++ toolchain support so we add
corresponding dependency to gdb Config.in file.
Signed-off-by: Vlad Zakharov <vzakhar@synopsys.com>
[Thomas:
- fix dependency on C++ of gdb, it must use BR2_INSTALL_LIBSTDCPP
- add comment about the C++ dependency of gdb on ARC.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Zakharov Vlad [Mon, 14 Nov 2016 18:07:00 +0000 (21:07 +0300)]
configs/snps_axs*: bump U-Boot to 2016.11
With this commit we upgrade U-Boot version to 2016.11 in
"snps_axs101_defconfig" and "snps_axs103_defconfig".
Important fixes: new version fixes U-Boot build for arc700. Since
gcc-6.x "-marc700" option is no longer supported, "-mcpu=arc700"
should be used instead.
We haven't sent it before as we were waiting for U-Boot 2016.11
release.
Signed-off-by: Vlad Zakharov <vzakhar@synopsys.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Zakharov Vlad [Thu, 10 Nov 2016 15:53:27 +0000 (18:53 +0300)]
config: bump linux kernel to 4.8.6 in synopsys defconfigs
With this commit we update ARC defconfigs with the following:
- "snps_axs101_defconfig", "snps_axs103_defconfig" and
"snps_hs38_smp_vdk_defconfig":
- bump linux kernel version to 4.8.6
- set up host linux headers to 4.8
Signed-off-by: Vlad Zakharov <vzakhar@synopsys.com>
[Thomas: as reported by Alexey, this commit is in fact a bug fix: it
makes the defconfig work properly. Indeed, since the ARC compiler has
moved to gcc 6.x, the ABI has changed, and a Linux 4.8+ kernel is
needed for userspace to work on ARC.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Thomas Petazzoni [Sun, 13 Nov 2016 17:03:57 +0000 (18:03 +0100)]
quota: fix build on musl
This commit adds 3 patches to the quota package, that allows to build
it with musl. One patch was borrowed from OpenEmbedded, the other two
patches are really trivial.
Peter Seiderer [Fri, 11 Nov 2016 22:54:11 +0000 (23:54 +0100)]
openjpeg: fix static linking of liblcms2
Fixes [1]:
.../host/usr/bfin-buildroot-uclinux-uclibc/sysroot/usr/lib/liblcms2.a(cmsplugin.o): In function `_cmsDeleteContext':
cmsplugin.c:(.text+0x1c2): undefined reference to `_pthread_mutex_lock'
cmsplugin.c:(.text+0x1f6): undefined reference to `_pthread_mutex_unlock'
Signed-off-by: Samuel Martin <s.martin49@gmail.com> Cc: Olivier Schonken <olivier.schonken@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Romain Naour [Sat, 12 Nov 2016 15:30:12 +0000 (16:30 +0100)]
package/strace: disable libmpers with musl toolchains
strace bundle some kernel headers to build libmpers, this mixes userspace
headers and kernel headers which break the build with musl.
The stddef.h from gcc is used instead of the one from musl.
Thomas Petazzoni [Fri, 11 Nov 2016 21:09:04 +0000 (22:09 +0100)]
tinymembench: add patch to fix build on Cortex-M
This commit adds a patch to the tinymembench package that fixes the
build on Thumb-only architectures such as ARM Cortex-M. The patch has
been submitted upstream in
https://github.com/ssvb/tinymembench/pull/13.
Baruch Siach [Sat, 12 Nov 2016 19:29:56 +0000 (21:29 +0200)]
jasper: disable -pedantic-errors
The -pedantic-errors gcc option turns -pedantic warnings into errors. This
mostly affects older gcc versions that default to the ISO90 C standard. Use
the --disable-strict configure option to remove -pedantic-errors.
Romain Naour [Fri, 11 Nov 2016 21:37:42 +0000 (22:37 +0100)]
package/sane-backends: fix musl build issue
We should include config.h from sanei_backend.h in order to use the
correct if/else HAVE_FOO.
For some reason with Glibc or uClibc there is no problem but with musl
we have the following weird issue:
In file included from epsonds.h:41:0,
from epsonds-jpeg.c:18:
../include/sane/sanei_backend.h:99:33: error: expected ';', identifier or '(' before 'int'
../include/sane/sanei_backend.h:99:33: warning: useless type name in empty declaration
That's because HAVE_SIGPROCMASK is not defined although it's correctly
detected by the configure script.
Rahul Bedarkar [Tue, 8 Nov 2016 18:36:41 +0000 (00:06 +0530)]
wget: fix ssl detection in static libs configuration
When building wget with openssl in static libs configuration, wget
build system fails detect openssl because it doesn't specify LD flags
for private libs used by openssl. This specifically happens when we
pass --with-libssl-prefix to configure which tries to find ssl using
custom flags. If we don't specify --with-libssl-prefix, it relies on
pkg-config files to detect ssl and it's LD flags which helps with static
linking.
This commit removes --with-libssl-prefix conf opts. Since this case is
similar to gnutls, we remove same conf opts for gnutls as well.
wget can be built with either gnutls or openssl crypto libraries, so
separate optional support for both is not required. This commit also
does minor optimization by checking for either gnutls or openssl while
at it.
Thomas Petazzoni [Fri, 11 Nov 2016 20:35:29 +0000 (21:35 +0100)]
go: remove powerpc64 big-endian from supported architectures
cgo currently doesn't properly support powerpc64 big-endian, as noted
in https://github.com/golang/go/issues/13192, and indeed, we have a
large number of build failures of Go packages on this
architecture. This commit therefore disables Go on PowerPC64
big-endian (PowerPC64 little-endian is fine).
Romain Naour [Fri, 11 Nov 2016 17:01:59 +0000 (18:01 +0100)]
package/kvmtool: fix a musl build issue on ARM
We simply bump the version rather than backport the patch, as there has
been only very few, minor commits in-between.
The kvmtool build with musl still produce some warning but they are
fixed with the current master 0093df80 (kvmtool: Makefile: disable PIE
build for bios and pre_init). So the next version bump should fixes
them.
Baruch Siach [Thu, 10 Nov 2016 17:54:39 +0000 (19:54 +0200)]
jasper: security bump to version 1.900.22
Fixes:
CVE-2016-8693: Double free vulnerability in mem_close
CVE-2016-8692: Divide by zero in jpc_dec_process_siz
CVE-2016-8691: Divide by zero in jpc_dec_process_siz
CVE-2016-8690: Null pointer dereference in bmp_getdata triggered by crafted
BMP image
CVE-2016-2089: matrix rows_ NULL pointer dereference in jas_matrix_clip()
CVE-2016-8886: memory allocation failure in jas_malloc
CVE-2016-8887: Null pointer dereference in jp2_colr_destroy
CVE-2016-8884, CVE-2016-8885: Null pointer dereference in bmp_getdata
(incomplete fix for CVE-2016-8690)
CVE-2016-8880: Heap buffer overflow in jpc_dec_cp_setfromcox()
CVE-2016-8881: Heap buffer overflow in jpc_getuint16()
CVE-2016-8882: Null pointer access in jpc_pi_destroy
CVE-2016-8883: Assert in jpc_dec_tiledecode()
Drop upstream patches.
Change SITE to the official download location, since the current one does not
have the updated version. Unfortunately, the official site only offers tar.gz.
Fix license. It is "based on the MIT license", but not exactly the same
(http://www.ece.uvic.ca/~frodo/jasper/; under "Legal Issues").
Drop autoreconf; the autotools version has been updated since commit 324ccec90d (jasper: autoreconf to fix rpath issue) that introduced it.
Cc: Maxime Hadjinlian <maxime.hadjinlian@gmail.com> Signed-off-by: Baruch Siach <baruch@tkos.co.il> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Romain Naour [Fri, 11 Nov 2016 11:45:11 +0000 (12:45 +0100)]
package/domoticz: disable for Sourcery CodeBench SH 2012.09
Like for prboom, domoticz trigger an internal compiler error:
[...]/hardware/MySensorsBase.cpp: In member function 'void MySensorsBase::Do_Send_Work()':
[...]/hardware/MySensorsBase.cpp:2190:6: internal compiler error: Segmentation fault
Thomas Petazzoni [Thu, 10 Nov 2016 12:54:48 +0000 (13:54 +0100)]
configs/atmel_sama5d4_xplained: re-add defconfig
This defconfig was mistakenly removed by commit 78ac39cb89a6933c36508f6b10d30150ddb1ba05 due to an improper conflict
detection in "git am". This commit restores it.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
board/atmel/readme.txt: remove mention of sama5d4ek
The sama5d4ek defconfig has been removed in commit 9cb0b37cbd267ec1c2beea3deeaa86846bc2d037, but the board/atmel/readme.txt
file was not updated accordingly. This commit fixes that.
The Atmel sama5d4ek board is no longer sold. Since there is the Xplained
board and we won't maintain anymore the EK board, the defconfig can be
removed. In addition, the defconfig was failing to build with gcc 5.x.
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Yann E. MORIN [Tue, 8 Nov 2016 22:18:35 +0000 (23:18 +0100)]
package/pseudo: provide better legacy handling for fakeroot
We now have a wrapper that makes pseudo behaves like the fakeroot of the
good ol' days. So the symlink will just magically keep old scripts
working as they did before the switch to pseudo.
However, using fakeroot is deprecated, and we want people to stop using
it altogether and switch to pseudo.
So, make the wrapper recognise how it's called, and if called as
fakeroot, print a warning message.
Yann E. MORIN [Tue, 8 Nov 2016 22:18:33 +0000 (23:18 +0100)]
package/pseudo: wrap the real pseudo with a wrapper
Running pseudo is more involved than running fakeroot. In the transition
from using fakeroot, we just did not account for the extra requirements.
First, we explicitly tell pseudo where it is, otherwise it tries to
guess. Its guess is correct, but it prints a warning, which is not nice.
Second, we tell it where to find the passwd and group files in case it
has to emulate access to them. We currently do not use that feature, but
better safe than sorry.
Third, pseudo spawns a background daemon, and talks to it (when fakeroot
would emulate the state all in the current process' state, pseudo uses
the daemon to coordinate the state across multiple processes). We are
not much interested in the daemon lingering around, so we just tell it
to terminate as soon as the last clients quits (this can take up to one
second).
Fourth and last, pseudo always stores its internal database when
exiting, and reloads it when spawned. The database is by default stored
in a sub-directory of the prefix it was installed in, but this is
impractical for us. We want the database to be specific to the one
config dir we are building, so we store the database in a (hidden)
sub-dir of the build dir, thus ensuring it is never shared with another
build. That directory is hidden (starts with a dot) because we consider
that to be our internal state that we do not want to expose to the user.
The wrapper has to be relocatable, so we avoid using hard-coded paths
in there: we derive those paths fom the runtime path of pseudo. However,
the build directory $(BUILD_DIR) is not available in the environment
(we do not export it because it conflicts with some buildsystems).
Instead, we use $(BASE_DIR) which is exported.
Finally, when relocated, the wrapper would not be used in the Buildroot
environment, so may not have access to TARGET_DIR or BASE_DIR, unless
the user sets them. If he does not, we still want the wrapper to be
working (to avoid the warning about the prefix, and to exit the daemon
asap); thus we leave the passwd and localstatedir variable alone if we
don't have what it needs to set them, rather than set them to incorrect
values.
Jérôme Pouiller [Wed, 9 Nov 2016 10:57:27 +0000 (11:57 +0100)]
Makefile: fix stripping of ld.so and libpthread for merged /usr
If 'lib' is a symlink (as is the case when BR2_ROOTFS_MERGED_USR=y),
'find lib' does not return the correct result. So, until now,
libpthread*.so* and ld-*.so* were not stripped when 'lib' was a symlink.
We fix this by using 'find lib/' instead of 'find lib'. For consistency
reason, we also do the same change for the 'find' that removes .a and
.la files.
Signed-off-by: Jérôme Pouiller <jezz@sysmic.org>
[Thomas: slightly improved the commit log.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The Calao Systems company has gone bankrupt in April 2016, and the
domain name no longer exists, so there is no chance to ever contact
Gregory Hermant at a @calao-systems.com address, so let's remove his
entry from the DEVELOPERS file.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
The calao defconfigs no longer build with gcc 5.x, due to Linux kernel
versions used being too old. However, it is unlikely that anyone will
ever update them, since Calao Systems has gone bankrupt in April 2016.
Therefore, let's remove them. If anyone is interested again at some
point, it will be easy to revive them from the Git history.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jérôme Pouiller [Wed, 9 Nov 2016 09:51:24 +0000 (10:51 +0100)]
valgrind: do not use host mpicc
Valgrind may use 'mpicc' (from openmi project) to compile libmpiwrap-*.so.
Without any option, it will detect and use 'mpicc' from host and generate
libmpiwrap-*.so for host:
$ file target/usr/lib/valgrind/libmpiwrap-arm-linux.so
target/usr/lib/valgrind/libmpiwrap-arm-linux.so: ELF 64-bit LSB shared object,
x86-64, version 1 (SYSV), dynamically linked, not stripped
We simply disable openmi support for valgrind.
Signed-off-by: Jérôme Pouiller <jezz@sysmic.org>
[Thomas: refactor with other unconditional CONF_OPTS.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jérôme Pouiller [Wed, 9 Nov 2016 09:25:08 +0000 (10:25 +0100)]
python3: fix double format detection
Python is not able to detect if compiler double representation is
compliant with IEE754:
checking whether C doubles are little-endian IEEE 754 binary64... no
checking whether C doubles are big-endian IEEE 754 binary64... no
checking whether C doubles are ARM mixed-endian IEEE 754 binary64... no
Accordingly 'legacy' mode isused. It is possible to check this at
runtime by check if 'sys.float_repr_style' contains 'short' or
'legacy'. Calculus correctness is not garanteed with 'legacy'.
However, all gcc architecture use a representation compliant with
IEE754. So, we can enable it unconditionnaly.
Signed-off-by: Jérôme Pouiller <jezz@sysmic.org>
[Thomas: rework condition to not use strip, as suggested by Baruch.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jérôme Pouiller [Wed, 9 Nov 2016 09:25:07 +0000 (10:25 +0100)]
python: fix double format detection
Python is not able to detect if compiler double representation is
compliant with IEE754:
checking whether C doubles are little-endian IEEE 754 binary64... no
checking whether C doubles are big-endian IEEE 754 binary64... no
checking whether C doubles are ARM mixed-endian IEEE 754 binary64... no
Accordingly 'legacy' mode isused. It is possible to check this at
runtime by check if 'sys.float_repr_style' contains 'short' or
'legacy'. Calculus correctness is not garanteed with 'legacy'.
However, all gcc architecture use a representation compliant with
IEE754. So, we can enable it unconditionnaly.
Signed-off-by: Jérôme Pouiller <jezz@sysmic.org>
[Thomas: adjust condition to avoid usage of qstrip, suggested by Baruch.] Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Jan Viktorin [Wed, 9 Nov 2016 13:27:39 +0000 (14:27 +0100)]
olimex_a20_olinuxino_lime2: bump U-Boot to 2016.05
The patch fixes build by GCC 5 (unsupported by the older U-Boot).
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com> Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
It has a bug in the FPU that can generate incorrect results in certain
cases. The problem shows up when you have several fused madd
instructions in sequence with dependant operands.
Using the -mno-fused-madd option prevents gcc from emitting these
instructions. This patch adds changes to the toolchain wrapper to use
that option.
Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
projects / coffee / buildroot.git / log