Fixes following CVE's:
CVE-2026-70368: Fixed an out-of-bounds memory
access triggered by logging attacker-controlled protocol messages longer
than 1,024 bytes (thanks to AISLE Research and Clemens Lang).
CVE-2026-70367: Fixed a SOCKS server mode bypass of the localhost
destination filter using alternate local-address encodings and
interface-scoped IPv6 destinations (thanks to AISLE Research and Clemens
Lang).
Complete Changelog is here:
https://www.stunnel.org/NEWS.html
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Julien Olivain <ju.o@free.fr>
SDL3_gfx provides basic antialiased drawing routines such as lines,
circles or polygons, an interpolating rotozoomer for SDL surfaces,
framerate control and MMX image filters, as an extension to SDL3.
https://github.com/sabdul-khabir/SDL3_gfx
Signed-off-by: Michael Fischer <mf@go-sys.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Commit fef9cad1fe ("package/libxmlsec1:
bump version to 1.3.12") has introduced a gcc 7.x and atomic
dependency, but the Config.in comment dependencies are not correct:
+comment "libxmlsec1 needs a toolchain w/ gcc >= 7"
+ depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_7 || !BR2_TOOLCHAIN_HAS_ATOMIC
Indeed, we treat BR2_TOOLCHAIN_HAS_ATOMIC like an architecture
dependency in Buildroot, so it should be:
+comment "libxmlsec1 needs a toolchain w/ gcc >= 7"
+ depends on BR2_TOOLCHAIN_HAS_ATOMIC
+ depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_7
so that the comment does NOT show up when BR2_TOOLCHAIN_HAS_ATOMIC is
false.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Test an actual D-Bus round-trip using easydbus:
- obtain a name on the system bus
- subscribe to a signal on a chosen object/interface
- emit signal with a string payload
- run the easydbus mainloop
- verify the handler received the expected payload
This covers both the default Lua interpreter (Lua 5.4) and LuaJIT
(Lua 5.1 ABI).
Signed-off-by: Marcin Niestroj <m.niestroj@grinn-global.com>
[Julien:
- move test files into a rootfs-overlay
- move common BR2_PACKAGE_{,EASY}DBUS configs to EasyDBusBase
- make stricter check of script output
- add DEVELOPERS entry for rootfs-overlay files
]
Signed-off-by: Julien Olivain <ju.o@free.fr>
Upstream removed configure option INSTALL_EXAMPLE_CODE with commit
8434de3cbc
which was first released with version 2.11.0.
Upstream re-added a similar configure option named INSTALL_EXAMPLE_CODE
with commits
f8d989160e7c803c8538
which were first released with version 2.15.0.
Buildroot bumped the package to 2.15.0 with
8e713dcc88 and forgot to rename the
configure option.
Remove libva/x11/wayland-related configure option due to upstream commit
8434de3cbc
which was first released with version 2.11.0. Remove the selection of
optional packages for these dependencies as well.
Removed unneeded libatomic handling.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Changelog 2.5..2.6:
treewide:
Migrate documentation to readthedocs
Emit more specific error messages for USB permission-based access issues
Add unit tests
snagflash:
Support compressed input files, e.g. *.wic.bz2
Refactor fastboot-uboot logic to factor out common code and improve logging
Automatically detect when oem_run isn't supported by U-Boot, and fallback to ucmd
snagrecover:
Support Renesas RZ/N1 platforms
Support SoC model aliases, and add i.MX8MP alias
Handle relative path prefixes with the -F syntax
Changelog 2.6..2.6.1:
snagrecover:
Correct i.MX93 ROM code recovery which doesn't support control endpoint HID commands
Add multi-board support to AM335x recovery setup script
Add i.MX7Solo support
treewide:
Fix broken links in documentation
Drop Python 3.9 support
Add CI for Python 3.14
One packaging-related change is that Snagboot would now
unconditionally import the bzip2, gzip and lzma Python modules, which
are not necessarily available. Since these are only needed when
compressed images are provided as input, we submitted a patch upstream
to make those imports only when compressed files are provided as
input, relaxing those dependencies.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Upstream doesn't provide any useful changelog, but the number of
1.5.0..1.6.2 commits is reasonable:
bfc28c7 Define {MAX_,}RESERVED_BLOCKS as floats
5b93731 Recover automatic block number calculation
d6e659d Correct message for minimum block count
1b2a30e Correct also commented code to take into account reserved blocks
944a4fd Indirect blocks must be taken into account for stats
3e2fce1 Restore (MAX_)RESERVED_BLOCKS to integer division
b571221 Add missing newlines on error messages.
cd4ff51 don't call ftruncate() on stdout
6909cde Enable system locale - change from the standard (C) to system locale.
9651f5d Fix alignment issues for strict architectures
4a99c22 Hardcode length of hardcoded char array instead of sizeof(long)
1f38ba4 Add ordered directory traversal
84c2d01 Add ordered directory traversal test
467ebe5 Fix formatting of man page references
810c12e Fix auto-block calculation from PR #5 merge
28988bd Skip device table entry when parent path is not a directory (from PR #33)
45191ab Fix ordered directory traversal from PR #34 merge
b615521 configure.ac: teach autoreconf to detect and install macros in m4
81028ca configure.ac: upgrade ancient snprintf macro to modern autoconf-archive edition
67661a1 Fix EXTRA_DIST: reference new ax_func_snprintf.m4 instead of deleted ac_func_snprintf.m4
907737e Remove incorrect address
4c4343d Add licensing information
cb5f584 Add license information
a60ccd9 Use strrchr to split path:dest so host paths may contain colons (fixes#23)
44ff03d Implement sparse_super feature with superblock/GDT backups (fixes#29)
8b8f2d1 Add extended attributes (xattr) support (fixes#9)
d0a9274 Optimize block/inode allocation for large filesystems (fixes#31)
b69c34e Fix -d dir:/path destination: restore colon after stats pass, auto-create intermediate dirs
2c38700 Add 9 functional tests for previously untested features
3d3bcb6 remove useless NEWS file
03dee74 Bump version to 1.6.0 (fixes#28)
c8f41b9 Make xattr copying opt-in with -X/--xattrs flag (fixes#40)
3324695 Bump version to 1.6.1
4e5d8ae Update README: add -X/--xattrs, SOURCE_DATE_EPOCH, and missing device table types
bf5b1a0 Bump version to 1.6.2
This commit also updates the COPYING license file hash, after an update
of the FSF, in upstream commit:
907737e4e8
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[Julien: fix COPYING license file hash]
Signed-off-by: Julien Olivain <ju.o@free.fr>
Changelog (https://www.bytereef.org/mpdecimal/changelog.html) 4.0.0 to
4.0.1:
features
Add Cygwin support.
Update config.guess and config.sub to the latest versions.
build fixes
Fix pkg-config files for custom paths.
Set LD/LDXX unconditionally to CC/CXX, since LDFLAGS/LDXXFLAGS from ./configure rely on it.
macOS: use libdir instead of RPATH for the install_name.
This commit also updates the COPYRIGHT.txt license file hash, after a
year update.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
[Julien: fix COPYRIGHT.txt license file hash]
Signed-off-by: Julien Olivain <ju.o@free.fr>
https://github.com/irql-notlessorequal/intel-vaapi-driver/blob/2.4.5/NEWS
This bump follows
39d2ff612f
"Intel archived intel/intel-vaapi-driver at 2.4.1 and advises forking.
irql-notlessorequal/intel-vaapi-driver is the active continuation (119
commits ahead, tagged releases, Chromium fixes), so track it at 2.4.5.
[...]
The fork renamed the wayland meson option, so switch with_wayland to
with_wayland_drm in all DISPLAYSERVER branches. Also drop the dead
01.org PKG_SITE."
Updated license file and hash due to upstream commit:
7343767d87
Also switch build system to meson.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
- Update BOTAN_VERSION to 3.11.1
- Remove 0001-Add-more-value-barriers-to-avoid-compiler-induced-side-channels.patch
as it is already integrated upstream in this version.
Fixed in 3.11.1:
- CVE-2026-34580: Resolve certificate verification bypass bug introduced
in 3.11.0 (GH #5500)
- CVE-2026-34582: Resolve TLS 1.3 client authentication bypass (GH #5599)
Fixed in 3.11.0:
- CVE-2026-32877: Fix a heap over-read during SM2 decryption (GH #5450)
- CVE-2026-32883: Fix an OCSP response forgery vulnerability (GH #5449)
- CVE-2026-32884: Fix a name constraints bypass for DNS names (GH #5448)
- Remove the --disable-altivec configuration option as it has been removed
from the Botan build system in version 3.x
- Optimize the --disable-neon logic using the yx pattern.
In Botan 3.x, the --disable-neon flag is specifically targeted at the
arm32 architecture. Furthermore, Buildroot's BR2_ARM_CPU_HAS_NEON
variable is only defined for 32-bit ARM, which previously caused a
false-positive --disable-neon flag to be passed on AArch64 builds
- Update license.txt hash in botan.hash due to the copyright year update
to 2026
https://botan.randombit.net/news.html#version-3-11-1-2026-03-31
Signed-off-by: Shubham Chakraborty <chakrabortyshubham66@gmail.com>
[Julien: fix CVE numbers in commit log]
Signed-off-by: Julien Olivain <ju.o@free.fr>
- botan dropped the usage of Boost.Filesystem in 2.10.0 [0]
- botan never actively searched for Boost.System, as it is header-only
the dependecy can be dropped too
The only part of boost botan currently uses is asio. So use --with-boost option
when boost is selected.
[0] aff8cc8a7a
Signed-off-by: Michael Nosthoff <buildroot@heine.tech>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Changes made in this release:
Ben Wolsieffer (1):
nptl: fix getting main stack size on no-MMU
Max Filippov (4):
stdio: fix %a formatting of inf and nan
stdio: fix radix point output for %a
stdio: output sign with NaNs
stdio: restrict first digit of %a output to 0 or 1
Ramin Moussavi (19):
libintl.h: route stubs through format_arg inlines; fix gettext_noop
build: scope .NOTPARALLEL to locale targets only, restore parallel build
limits.h: define PAGE_SIZE/PAGESIZE like musl, fixing the gdb build
wait4: fall back to waitid(2) at runtime when wait4 returns ENOSYS
pagesize: restore PAGE_SHIFT fallback for binfmt_flat (no auxv)
buildsys: build locale headers in-tree, not via a recursive sub-make
extra/locale: skip locales the build host cannot provide
csky: fix struct semid_ds layout for TIME64
csky: make __kernel_uid_t/__kernel_gid_t 32-bit
hppa: fix inotify/signalfd/timerfd flag values
hppa: do not set IPC_64 for the SysV IPC *ctl commands
x86: fix struct semid_ds layout for TIME64
i386: add CFI to __libc_i386_syscall6 for cancellation unwinding
i386: do not set IPC_64 for the SysV IPC *ctl commands on 5.1+ kernels
buildsys: terminate .eh_frame in shared objects
m68k: make the __jmp_buf layout match what setjmp.S actually saves
inet: use the direct accept4 syscall when __NR_accept is missing
m68k: implement atomic operations with the CAS instruction
libc: fix struct shmid_ds layout for TIME64
Waldemar Brodkorb (10):
netinet/in.h: sync with glibc
bits/fcntl-linux.h: sync with glibc
add copy_file_range syscall wrapper
sys/fanotify.h: drop include to fix c-sky compilation
adjtimex: fix for architectures lacking __NR_clock_adjtime, f.e. frv
sys/user.h: restore required macros
netinet/tcp.h: sync with glibc
loongarch: add basic static only support
riscv: setcontext: fix linking error
bump version for 1.0.59 release
ramin (16):
ldso: harden _dl_tls_get_addr_soft against an unusable DTV
microblaze: implement atomic operations with lwx/swx
ldso: microblaze: add missing addend to R_MICROBLAZE_TLSDTPREL32
ldso: make the _dl_tls_get_addr_soft DTV check microblaze-specific
microblaze: fix struct semid_ds layout for TIME64
m68k: fix struct semid_ds layout for TIME64
mips: fix fallocate() 64-bit argument passing on n32
mips: fix mmap64() for large offsets on n32
mips: fix setjmp() on n32/n64 (.cpreturn, not .cprestore)
mips: always set IPC_64 for the SysV IPC *ctl commands
libc: fix struct msqid_ds layout for TIME64 on all common-header arches
linuxthreads: fix SIGSEGV in MIPS cancellable wrappers (waitpid, nanosleep)
ldso: serialize dl_iterate_phdr to fix concurrent unwind hang on SMP
nptl: cancel the current thread directly instead of via SIGCANCEL
mips: only n32/n64 always need IPC_64, not o32
m68k: use the kernel cmpxchg helper for atomics on ColdFire
Patch 0001 is applied upstream and is removed here.
Fixes:
- https://autobuild.buildroot.net/results/2bd/2bd73b0060cad4f44649ffe4ea60c4775160d86e
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Since the bump of grub2 to version 2.14, the build fails on some
configurations with:
/home/autobuild/autobuild/instance-8/output-1/host/bin/grub-mkimage: error: `/home/autobuild/autobuild/instance-8/output-1/build/grub2-2.14/build-i386-pc/grub-core//kernel.img' is miscompiled: its start address is 0x9074 instead of 0x9000: ld.gold bug?.
We are not the only ones affected: OpenEmbedded has the issue as well,
addressed in commit
66aefeca48,
and this commit points that libreboot and Gentoo have also adopted the
same temporary fix: revert two upstream commits.
So to fix the issue in Buildroot, we take the same approach. It is
temporary, and there is an upstream patch series under discussion to
really fix the problem. But as this series has 9 patches still under
discussion, we take the simpler approach of reverting two simpler
patches for the time being.
host-pkgconf is needed as an additional dependency because
configure.ac using m4 macros provided by host-pkgconf.
Fixes:
https://autobuild.buildroot.net/results/89b/89b907278cada3d7ca1fee2d08d66921ffb4d7cc/build-end.log
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
In order to verify that pyudev works fine, including with just
libudev-zero, add a basic test for this package.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Commit 3f24135c56 ("package/libinput:
only needs libudev, not udev daemon") changed libinput to depend on
libudev instead of udev.
However, BR2_PACKAGE_LIBINPUT_PYTHON_TOOLS selects
BR2_PACKAGE_PYTHON_PYUDEV, which depends on udev, causing a broken
dependency chain:
WARNING: unmet direct dependencies detected for BR2_PACKAGE_PYTHON_PYUDEV
Depends on [n]: BR2_PACKAGE_PYTHON3 [=y] && BR2_PACKAGE_HAS_UDEV [=n]
Selected by [y]:
- BR2_PACKAGE_LIBINPUT_PYTHON_TOOLS [=y] && BR2_PACKAGE_LIBINPUT [=y] && BR2_PACKAGE_PYTHON3 [=y]
Turns out that after experimenting, the majority of the pyudev
functionality works fine with just libudev. Only pyudev.udev_version()
has been detected as not working as it calls into udevadm. But the
actual features to enumerate devices and get their characteristics
just fine.
So, to fix the issue, we change the dependency of python-pyudev to
just libudev instead of udev.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Commit 5394df6400 ("package/tpm2-tss:
add Mbed-TLS crypto backend support") added a dependency to
BR2_PACKAGE_TPM2_TSS_FAPI, and a comment about this
dependency. However, the comment is shown even if other
architecture-related dependencies are not met.
This commit fixes that by only displaying the comment when relevant,
i.e when BR2_TOOLCHAIN_HAS_SYNC_4.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
In commit 5394df6400 ("package/tpm2-tss:
add Mbed-TLS crypto backend support"), a depends on
!BR2_PACKAGE_MBEDTLS was added to BR2_PACKAGE_TPM2_TSS_FAPI, but this
new dependency was not properly propagated to the reverse
dependencies, causing kconfig warnings such as:
WARNING: unmet direct dependencies detected for BR2_PACKAGE_TPM2_TSS_FAPI
Depends on [n]: BR2_PACKAGE_TPM2_TSS [=y] && BR2_TOOLCHAIN_HAS_SYNC_4 [=y] && !BR2_PACKAGE_MBEDTLS [=y]
Selected by [y]:
- BR2_PACKAGE_TPM2_TOOLS_FAPI [=y] && BR2_PACKAGE_TPM2_TOOLS [=y] && BR2_TOOLCHAIN_HAS_SYNC_4 [=y]
This commit properly propagates this new dependency to
BR2_PACKAGE_TPM2_TOOLS_FAPI, fixing the kconfig warning.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
It has been good practice for a while to describe the kind of change
in the commit message of any patch that changes license file
hashes. This is important so any substantial changes don't go
unnoticed. However it wasn't explicitly stated in the manual, fix
that.
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
Localedef is built from the glibc source code, so license information
is the same.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
The COPYINGv3 license file was added with upstream commit
e067e53080386e93dcf8b07e25fb6656f2c8941e, it applies to scripts used
during build.
LGPL-3.0+ applies to sysdeps/htl/raise.c, which is used for the Hurd
platform. Buildroot builds only Linux, so the file should not be
compiled, but it is part of the downloaded (and possibly shared, if
"make legal-info" is used) archive. The same applies to the manual.
Checked against the copyright file of the Debian "libc6" package (as
of 2.42-17 and 2.43-2), note that Debian excludes the manual.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
https://gcc.gnu.org/pipermail/gcc-announce/2026/000193.html
GCC 16.2 is a bug-fix release from the GCC 16 branch
containing important fixes for regressions and serious bugs in
GCC 16.1 with more than 102 bugs fixed since the previous release.
For the list of bugs fixed in this version, see:
https://gcc.gnu.org/gcc-16/changes.html#16.2
Removed patch 0003 which was backported from the gcc-16 branch and is
included in this release.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
This reverts commit 72f7aa47e7.
The MALLOC_SIMPLE workaround is no longer needed: the root cause, a
zero _dl_pagesize on noMMU/FLAT targets, is fixed by commit
35d3927e85 ("package/uclibc: fix for noMMU targets using FLAT
binaries").
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
This reverts commit 832ff692ab.
The MALLOC_SIMPLE workaround is no longer needed: the root cause, a
zero _dl_pagesize on noMMU/FLAT targets, is fixed by commit
35d3927e85 ("package/uclibc: fix for noMMU targets using FLAT
binaries").
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
This reverts commit dd67de1bd3.
The MALLOC_SIMPLE workaround is no longer needed: the root cause, a
zero _dl_pagesize on noMMU/FLAT targets, is fixed by commit
35d3927e85 ("package/uclibc: fix for noMMU targets using FLAT
binaries").
Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Signed-off-by: Fiona Klute <fiona.klute@gmx.de>
