Dependency was made optional in 4.6.0 release here:
6058ab9dfe
python-can has a lot of optional dependencies, most of which are not
represented in buildroot. As msgpack is used for the virtual multicast
udp can interface[1], which does seem like a bit of a niche usecase,
just drop the mandatory dependency without introducing a user-visible
config option to enable it.
[1] https://python-can.readthedocs.io/en/4.0.0/interfaces/udp_multicast.html
Signed-off-by: Marcus Hoffmann <buildroot@bubu1.eu>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit a4cdb412f1)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Libiio python bindings use ctypes and specifically the find_library()
function from there to load the libiio.so shared library. This is not
working unless glibc utils (specifically ldconfig) is installed to the
target (alternatively the target would need gcc or binutils, for objdump
or ld).
The easy fix here is to just bypass the find_library() machinery
altogether as it's not needed on a buildroot system.
Signed-off-by: Marcus Hoffmann <buildroot@bubu1.eu>
Tested-by: Fiona Klute <fiona.klute@gmx.de>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 193df1cbec)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In commit a68899d49e ("package/python3:
work around GCC bug 121567"), we introduced a work around for a gcc
bug, by reducing to -O1 the optimization level on SuperH.
However, it turns out that this is not sufficient, as the build will
only succeeded at -O0.
Fixes:
https://autobuild.buildroot.net/results/31f/31f34a983036b4135c12e5797b5c2258ab33e6c2/
Which is a config with BR2_OPTIMIZE_2=y, which means
BR2_TOOLCHAIN_HAS_GCC_BUG_121567=y, and therefore -O1 is passed, but
still the build fails. At -O0 the build doesn't fail.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 0f938aed25)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The test to enable NEON on AArch64 is as following:
ifeq ($(BR2_aarch64)$(BR2_ARM_CPU_HAS_NEON),yy)
It cannot be to true as $(BR2_aarch64) and $(BR2_ARM_CPU_HAS_NEON) are
mutually exclusive. NEON is compulsory on AArch64 so remove
$(BR2_ARM_CPU_HAS_NEON) from the test.
Fixes: ba2fb599cd ("package/pixman: bump to version 0.44.2")
Signed-off-by: Sébastien Szymanski <sebastien.szymanski@armadeus.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit d297569eb4)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since the bump of pip to version 25.3 in commit
285097051d, the build of the Python
bindings of libselinux is broken for both the host and target
variants.
For the host variant, because "pip install" no longer finds the
system-provided setuptools and tries to download setuptools by itself,
causing build issues because our host-python doesn't have SSL support:
Could not fetch URL https://pypi.org/simple/setuptools/: There was a problem confirming the ssl certificate: HTTPSConnectionPool(host='pypi.org', port=443): Max retries exceeded with url: /simple/setuptools/ (Caused by SSLError("Can't connect to HTTPS URL because the SSL module is not available.")) - skipping
For the target variant, because "pip install" no longer finds the
sysconfigdata package:
ModuleNotFoundError: No module named '_sysconfigdata__linux_sparc64-linux-gnu'
[end of output]
We fix this by taking a patch from Debian, which is slightly tweaked
to also cover our host package (the original Debian patch was passing
--no-build-isolation only when DESTDIR was not empty, but in Buildroot
host packages are built with DESTDIR empty, and we do need
--no-build-isolation).
Fixes:
https://autobuild.buildroot.net/results/0e9de0c0d8b6ec57eea9f8834f02076b296ba4f1/ (host-libselinux)
https://autobuild.buildroot.org/results/1b87c659f1901b0bf33fa4a2ff0ed40b13114bba/ (libselinux)
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Co-Authored-By: Julien Olivain <ju.o@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 0e5eef911c)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The CPE 'cpe:2.3:a:antirez:linenoise:1.0:*:*:*:*:*:*:*' is valid for the
package linenoise [1].
Since the latest version is '1.0' since 2015 the CPE_ID_VERSION is set
to that version.
The CVE that applies on version 1.0 were checked with the 'cve-check'
script:
```
echo '{"components": [{"bom-ref": "linenoise", "name": "linenoise", "version": "1.0", "cpe": "cpe:2.3:a:antirez:linenoise:1.0:-:*:*:*:*:*:*"}]}' | support/scripts/cve-check | jq -r '.vulnerabilities[].id'
```
Only the CVE-2025-9810 exists and that was fixed in [2].
[1] https://nvd.nist.gov/products/cpe/detail/10423C23-6AAA-439E-B723-1FCDEB3A769F
[2] 3c7cbf97d7 package/linenoise: security bump to version e26268de5e
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 2668d121e5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
opencv3's code is not compatible with newer versions of ffmpeg, and
opencv3 is no longer maintained, so we have no choice but to disable
its ffmpeg support.
Fixes:
https://autobuild.buildroot.net/results/9ae3911583cccb6362f33cd82e5eaafb059fdc76/
It's not clear which ffmpeg version bump broken the build exactly, but
this issue is definitely present in 2025.02.x as the following
defconfig fails to build in a similar way on 2025.02.x:
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_AARCH64_GLIBC_STABLE=y
BR2_PACKAGE_FFMPEG_NONFREE=y
# BR2_PACKAGE_FFMPEG_FFMPEG is not set
# BR2_PACKAGE_FFMPEG_INDEVS is not set
# BR2_PACKAGE_FFMPEG_OUTDEVS is not set
BR2_PACKAGE_OPENCV3=y
BR2_PACKAGE_OPENCV3_LIB_VIDEOIO=y
BR2_PACKAGE_OPENCV3_BUILD_PERF_TESTS=y
BR2_PACKAGE_OPENCV3_WITH_FFMPEG=y
BR2_PACKAGE_OPENCV3_INSTALL_DATA=y
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit ef538cf4d9)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The protobuf support breaks the build, as protobuf includes
libabseil-cpp headers, which now require C++14. opencv3 doesn't have
any ENABLE_CXX14 option, so for the time being, disable protobuf
support until someone bothers enough to fix this up.
While we suspect a libabseil-cpp version bump to be responsible for
the issue, we are not 100% sure. However, the issue is definitely
present in Buildroot 2025.02.x, as it can be reproduced using the
following defconfig:
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_PACKAGE_OPENCV3=y
BR2_PACKAGE_OPENCV3_LIB_SHAPE=y
BR2_PACKAGE_OPENCV3_LIB_STITCHING=y
BR2_PACKAGE_OPENCV3_LIB_SUPERRES=y
BR2_PACKAGE_OPENCV3_LIB_TS=y
BR2_PACKAGE_OPENCV3_LIB_VIDEOSTAB=y
BR2_PACKAGE_OPENCV3_WITH_PROTOBUF=y
Fixes:
https://autobuild.buildroot.net/results/39432e7746e6bc5224592a7d2f744ca992bd529a/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 0865927da4)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
cppcms.com now points to a Github page at
https://github.com/artyom-beilis/cppcms which has a 2.0.1 version, so
let's use that. The number of differences to 2.0.0.beta2 is very
small:
$ git log --online v2.0.0.beta2..v2.0.1
b872972 (tag: v2.0.1, origin/master, origin/HEAD, master) Version to 2.0.1
a1914f7 (tag: v2.0.0) Replaced system category with one from predating C++11 (v1.2) because std::system_category does not translate WSAGetLastError results
c4febcc Merge pull request #104 from dreaming-augustin/upstream
922cd49 Python 3.12 compatibility: wrap regex in r''.
a11e9d4 Merge branch 'cpp11'
3000bc6 (origin/1.2_updates) Merge pull request #99 from dreaming-augustin/master
44e24c7 [#89] cppcms_error fix typo + consistent messages
a6d5575 (origin/cpp11) Added backtrace to system error
b3aef3b Fixed missing include for stripped down build
463a9a6 Removed IPV6 due to travis limitations
f8163c6 Merges from cpp11
31d4fe7 Added verbose log on failure
90bc996 Added Linux to build matrix - so I have fallback if normal build environment fails
f78ee39 Added Readme for github
a737d5d Merged python3 compatibility from master
0c67544 Support of python 2.7 and python 3
0d121a7 Python3 compatibility
2fc7e38 Python3 compatibility
d745869 python3 fix for tmp_cc
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 29641d1675)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Since Buildroot commit dcee99507c that
bumped package/icu to version 77-1, the build of cppcms with ICU
support enabled fails.
Indeed, ICU now requires C++17, and while cppcms.mk has some logic to
get C++ flags using icu-config, the -std=c++17 gets ultimately
overridden by the built-in -std=c++11 flag encoded in cppcms
CMakeLists.txt.
To fix this, we have submitted a patch upstream that ensures the
CMAKE_CXX_FLAGS passed on the command line take precedence over the
built-in flags defined in cppcms CMakeLists.txt.
Fixes:
https://autobuild.buildroot.net/results/9c34a08ea02499b28093ad3fa184cee10b2883ac/
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 5a8811cade)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Switched _SITE to github, old project site is down.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
[Julien: change _VERSION to use 'git describe --abbrev=40' format]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 652dbe71c3)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The VIM_REMOVE_DOCS variable is currently a post install target hook,
but it can just as well be done inside VIM_INSTALL_TARGET_CMDS
directly.
The hook was registered conditionally based on BR2_PACKAGE_VIM_RUNTIME
because prior to commit f7a07f42f7, the
hook's logic was:
find $(TARGET_DIR)/usr/share/vim -type f -name "*.txt" -delete
which was failing if BR2_PACKAGE_VIM_RUNTIME was not enabled, as
$(TARGET_DIR)/usr/share/vim would not exist.
But since this commit, the hook logic is:
$(RM) -rf $(TARGET_DIR)/usr/share/vim/vim*/doc/
which obviously won't fail if $(TARGET_DIR)/usr/share/vim doesn't
exist.
So let's simplify the whole logic.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 7bcc99b57b)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Instead of calling $(MAKE) multiple times, let's call it once, with
all installation targets needed. We introduce a VIM_INSTALL_TARGETS
variable to collect the list of make install targets that need to be
invoked.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 525a234303)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
In Buildroot, we more commonly do:
$(MAKE) -C $(@D)/src
than:
cd $(@D)/src; $(MAKE)
so let's adopt this more conventional style.
This coding style in vim.mk dates from when the package was introduced
by Peter Korsgaard back in 2010.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 21e613753e)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
While not very common, it is nice when package re-installation
works. Unfortunately the "installlinks" target of vim installs links
with "ln -s", causing a package reinstallation to fail with:
cd /home/thomas/buildroot/br/output-all/target/usr/bin; ln -s vim ex
cd /home/thomas/buildroot/br/output-all/target/usr/bin; ln -s vim view
cd /home/thomas/buildroot/br/output-all/target/usr/bin; ln -s vim rvim
cd /home/thomas/buildroot/br/output-all/target/usr/bin; ln -s vim rview
cd /home/thomas/buildroot/br/output-all/target/usr/bin; ln -s vim vimdiff
ln: failed to create symbolic link 'ex': File exists
ln: failed to create symbolic link 'view': File exists
make[2]: *** [Makefile:2749: /home/thomas/buildroot/br/output-all/target/usr/bin/ex] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: *** [Makefile:2752: /home/thomas/buildroot/br/output-all/target/usr/bin/view] Error 1
ln: failed to create symbolic link 'rvim': File exists
ln: failed to create symbolic link 'rview': File exists
To fix this, we remove the target links before proceeding with the
installation.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 67e84345c1)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Fixes the following vulnerability:
- CVE-2025-68146
A Time-of-Check-Time-of-Use (TOCTOU) race condition allows local
attackers to corrupt or truncate arbitrary user files through symlink
attacks.
For more informations, see:
- https://nvd.nist.gov/vuln/detail/CVE-2025-68146
- 18a9988008
(cherry picked from commit d9c1379d1f)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The tests check if all supported hash algorithms are usable in
mkimage, for both host and target packages. Additionally, as a
necessary tool, it verifies the previous fix for FIT output from
dumpimage.
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
[Julien: use builtin kernel for faster testing]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 2dbe71dba6)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Some host commands need to call other host commands: For example,
"mkimage" from host-uboot-tools needs to run "dtc". This would fail or
call system commands without adding the host bin dir to PATH.
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
[Julien: use python functions/constants to build path]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit befb6ae81d)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Otherwise "dumpimage -l" produces only a newline when processing a FIT
image.
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 39b925a0a6)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The build of the following basic configuration enabling the
imagination Vulkan driver
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN=y
BR2_TOOLCHAIN_EXTERNAL_BOOTLIN_AARCH64_GLIBC_STABLE=y
BR2_PACKAGE_MESA3D=y
BR2_PACKAGE_MESA3D_VULKAN_DRIVER_IMAGINATION=y
fails with:
meson.build:847:3: ERROR: Feature llvm cannot be disabled: CLC requires LLVM
Adding just LLVM as a dependency is not enough, as then libclc is
needed, then LLVMSPIRVLib, then clangBasic, then the pco_clc tool.
In fact, like the Panfrost driver, building the Imagination driver
requires building host tools using host-mesa3d. To fix this we:
- Make the BR2_PACKAGE_MESA3D_OPENCL option selectable
- Make sure that BR2_PACKAGE_MESA3D_VULKAN_DRIVER_IMAGINATION depends
on BR2_PACKAGE_MESA3D_LLVM and select
BR2_PACKAGE_MESA3D_NEEDS_PRECOMP_COMPILER (the latter being needed to
build host-mesa3d)
- Make sure the host-mesa3d builds imagination
tools (-Dtools=imagination) and install
pco_clc (HOST_MESA3D_INSTALL_PCO_CLC). This requires introducing
HOST_MESA3D_TOOLS as a list of tools to build, which then gets used
to construct the -Dtools argument, as we can now have both
"panfrost" and "imagination" in this list.
With all this, the defconfig above builds successfully.
This has been broken since Buildroot commit
5e818c16a3, which introduced the vulkan
driver support.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit e1d159c5d5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
