Since Buildroot commit [1] "package/gcc: switch to GCC 15.x as the
default", compiling freescale_imx91frdm_defconfig fails in TF-A with
error:
include/drivers/nxp/trdc/imx_trdc.h:7: error: header guard 'IMX_TRDC_H' followed by '#define' of a different macro [-Werror=header-guard]
7 | #ifndef IMX_TRDC_H
include/drivers/nxp/trdc/imx_trdc.h:8: note: 'IMX_XRDC_H' is defined here; did you mean 'IMX_TRDC_H'?
8 | #define IMX_XRDC_H
The issue happen because this defconfig is using a Buildroot
(internal) toolchain which uses gcc version 15 by default.
This issue was fixed upstream in [2], first included in TF-A v2.13.
This commit updates U-Boot, TF-A and Linux to use the NXP
tag lf-6.18.20-2.0.0, which include this fixes.
Note that, as described in the commit log of [3] which was used as a
base for this defconfig, U-Boot and Linux were forks. Now the
necessary patches are included in the vendor BSP, this commit
switches those repos to use the vendor ones.
This commit also removes the linux-headers.hash custom hash file
which was never needed, because the defconfig used an external
toolchain.
Also, due to the upstream uboot commit [4], this commit adds a uboot
config fragment to disable CONFIG_EFI_CAPSULE_AUTHENTICATE, to
remove the dependency on the command "cert-to-efi-sig-list", provided
the host efitools package, which is not yet in Buildroot.
This commit was run tested on board 700-94610 Rev A1.
[1] a75b8623f5
[2] 2e9198d0e0%5E%21/
[3] 2e05f22ff8
[4] fd58c275f6
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 25c65e0428)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Since Buildroot commit [1] "package/gcc: switch to GCC 15.x as the
default", compiling freescale_imx93frdm_defconfig fails in TF-A with
error:
include/drivers/nxp/trdc/imx_trdc.h:7: error: header guard 'IMX_TRDC_H' followed by '#define' of a different macro [-Werror=header-guard]
7 | #ifndef IMX_TRDC_H
include/drivers/nxp/trdc/imx_trdc.h:8: note: 'IMX_XRDC_H' is defined here; did you mean 'IMX_TRDC_H'?
8 | #define IMX_XRDC_H
The issue happen because this defconfig is using a Buildroot
(internal) toolchain which uses gcc version 15 by default.
This issue was fixed upstream in [2], first included in TF-A v2.13.
This commit updates U-Boot, TF-A and Linux to use the NXP
tag lf-6.18.20-2.0.0, which include this fixes.
Note that, as described in the commit log of [3] which introduced
this defconfig, U-Boot and Linux were forks. Now the necessary
patches are included in the vendor BSP, this commit switches those
repos to use the vendor ones.
While at it, this commit also switches the defconfig to use an
external glibc stable Bootlin toolchain, to follow recommendations
from [4]. Since an external toolchain is used, the linux-headers.hash
custom hash file is no longer needed and is also removed. At the
time of this commit, the Bootlin stable toolchain is a gcc 14.
Also, due to the upstream uboot commit [5], this commit adds a uboot
config fragment to disable CONFIG_EFI_CAPSULE_AUTHENTICATE, to
remove the dependency on the command "cert-to-efi-sig-list", provided
the host efitools package, which is not yet in Buildroot.
This commit was run tested on board 700-94611 Rev B1.
[1] a75b8623f5
[2] 2e9198d0e0%5E%21/
[3] 2e05f22ff8
[4] https://elinux.org/Buildroot:DeveloperDaysELCE2024#Rules_for_defconfigs
[5] fd58c275f6
Reported-by: Robert P. J. Day <rpjday@crashcourse.ca>
Tested-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit fcc6df2e10)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Commit [1] "package/uboot-tools: fix host FIT signature support"
slightly changed the logic in which the file
$(@D)/include/generated/autoconf.h is created.
When there is no Kconfig sub-option selected, the file is no longer
created and the build fails with error:
include/linux/kconfig.h:4:10: fatal error: generated/autoconf.h: No such file or directory
The issue can be reproduced with the with the simple commands:
cat >.config <<EOF
BR2_aarch64=y
BR2_TOOLCHAIN_EXTERNAL=y
BR2_PACKAGE_HOST_UBOOT_TOOLS=y
EOF
make olddefconfig
make host-uboot-tools
This commit fixes the issue by simply adding a touch of the file
at the beginning to make sure it will always exist.
[1] 928cc5dc5c
Cc: James Hilliard <james.hilliard1@gmail.com>
Reported-by: Robert P. J. Day <rpjday@crashcourse.ca>
Tested-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 9f17ac02b8)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
U-Boot host tools use the tools configuration namespace when code calls
CONFIG_IS_ENABLED(). With USE_HOSTCC, CONFIG_IS_ENABLED(FIT_SIGNATURE)
resolves to CONFIG_TOOLS_FIT_SIGNATURE, while
CONFIG_VAL(FIT_SIGNATURE_MAX_SIZE) resolves to
CONFIG_TOOLS_FIT_SIGNATURE_MAX_SIZE.
The host-uboot-tools package only generates the old CONFIG_FIT_SIGNATURE
define. This makes tools/Makefile include fit_check_sign, but the host code
sees FIT signature support as disabled and the OpenSSL-backed signing and
verification objects are omitted. As a result, mkimage accepts a signature
node without writing its value or injecting a required public key.
fit_check_sign then has no required key and reports success without
checking the configuration signature.
A FIT-support-only configuration never exercises this path, which is why
the existing hash-only runtime test still passes.
Generate the tools FIT signature options needed by the host code and pass
CONFIG_TOOLS_LIBCRYPTO=y so the OpenSSL-backed signing, verification and
cipher objects are selected. Keep CONFIG_FIT_SIGNATURE=y in the make
options because U-Boot tools/Makefile still uses it to build
fit_check_sign.
Extend TestHostUbootTools to create an RSA-signed FIT, require a 256-byte
configuration signature and a required public key, verify the FIT, corrupt
the signature, and require verification to fail.
Signed-off-by: James Hilliard <james.hilliard1@gmail.com>
[Julien: add comments in runtime test]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 928cc5dc5c)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
https://www.samba.org/samba/history/samba-4.24.5.html
This is a security release in order to address the following defects:
o CVE-2026-6949:
TSIG packet with name compression can crash DNS
Incorrect size calculations when a TSIG record contains
compressed names can lead to a large out-of-bounds write
causing the server to crash.
https://www.samba.org/samba/security/CVE-2026-6949.html
o CVE-2026-58216:
An authenticated user could possibly crash a KDC process
A kpasswd packet that contains malformed ASN.1 might cause
the server to access 6 bytes of unallocated memory. This
memory is not exposed to the user, but in some
circumstances the server could crash.
https://www.samba.org/samba/security/CVE-2026-58216.html
o CVE-2026-58218:
DNS signing DoS via TKEY name cache exhaustion
An unauthenticated user can repeatedly register names TKEY
names, which floods a cache causing legitimate TKEYs to be
expunged. This can practically block the use DNS TSIG
signing.
https://www.samba.org/samba/security/CVE-2026-58218.html
o CVE-2026-58221:
Samba AD authenticated LDAP access domain takeover
Samba AD low-privilege authenticated LDAP access allows
modifications to internal LDB special DNs, which permits a
domain takeover.
https://www.samba.org/samba/security/CVE-2026-58221.html
o CVE-2026-58222:
Samba AD LDAP Compare filter injection and trusted-request
confusion disclose protected attributes
An ordinary authenticated domain user can bypass access
checks and query confidential Active Directory attributes
(such as KDS root keys) via LDAP Compare requests. Due to a
filter injection flaw and trusted execution context, the
LDAP Compare operation can be turned into a
protected-attribute disclosure oracle.
https://www.samba.org/samba/security/CVE-2026-58222.html
o CVE-2026-58224:
The CTDB protocol has bounds checking issues
CTDB fails to do integrity checking of received packets.
This includes failure to check field lengths against packet
lengths when unmarshalling packets.
https://www.samba.org/samba/security/CVE-2026-58224.html
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit b5e6133eab)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
The --with-systemd option controls only if systemd units are
installed, so there is no build time dependency. Disable systemd
unconditionally for the host package because Buildroot cannot and
should not install systemd units on the host.
Fixes host-p11-kit install error:
install: reguläre Datei '/usr/lib/systemd/user/p11-kit-server.socket' kann nicht angelegt werden: Keine Berechtigung
install: reguläre Datei '/usr/lib/systemd/user/p11-kit-server.service' kann nicht angelegt werden: Keine Berechtigung
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit a0cadb4552)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Changelog since 1.0.6:
This is a bugfix release, fixing modest security vulnerabilities
discovered by A. Ramos.
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
(cherry picked from commit 81f7c82740)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
Changelog since 1.0.5:
Fix various minor issues, like theoretical
integer overflows for extreme inputs or NULL
dereferences if the application passes NULL
where it should not.
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 6d8a1bb4c8)
Signed-off-by: Titouan Christophe <titouan.christophe@mind.be>
https://github.com/strukturag/libheif/releases/tag/v1.23.1
Fixes the following CVEs:
CVE-2026-62289 (GHSA-jc8f-p23p-5hjg)
Integer underflow in Fraction constructor via double clap transform
application
CVE-2026-62291 (GHSA-xpw3-9rhw-482x)
Heap out of bounds write in libheif uncompressed encoder when writing
images with mismatched auxiliary alpha dimensions
CVE-2026-62292 (GHSA-73p7-m7gg-w2jv)
Out-of-bounds read in uncompressed unci tile range slicing
CVE-2026-62377 (GHSA-9ww4-9v47-m7pj)
Reachable assertion in HeifContext::get_track() aborts on a valid-but-
empty HEIF sequence file
(GHSA-46rp-pcq2-rpmr)
Heap out-of-bounds write in the uncompressed encoder for RRGGBB images
with interleaved bit-depth ≤ 8
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit bcb48623fa)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://downloads.isc.org/isc/bind9/9.20.26/doc/arm/html/notes.html#notes-for-bind-9-20-26https://downloads.isc.org/isc/bind9/9.20.26/doc/arm/html/changelog.htmlhttps://seclists.org/oss-sec/2026/q3/208
Fixes
CVE-2026-10723: Incorrect acceptance of NSEC3 records
CVE-2026-10822: Key Record using PRIVATEDNS algorithm may lead to unexpected exit
CVE-2026-11331: Potential wildcard CNAME RPZ policy bypass
CVE-2026-11605: Unnecessary validation of DNSSEC signed records
CVE-2026-11622: Potential memory usage beyond configured limits
CVE-2026-11721: Cache poisoning possible with label count discrepancy, RRSIG, and wildcards
CVE-2026-12617: Record ordering based unexpected exit with CNAME or DNAME
CVE-2026-13204: Unexpected exit in certain situations with NSEC and NSEC3 both present
CVE-2026-13321: DNSSEC Validation Bypass via Out-of-Zone NSEC Next Field
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
[Julien: update pgp key id in hash file comment]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 8c96a8981f)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
A few headers outside libiberty use LGPL-2.1+, too, and other headers
and some scripts use GPL-2.0+.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 95b77f7a71)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
BSD-0-Clause is not a recognized SPDX license identifier. 0BSD is used
in the source SPDX-License-Identifier headers, and the associated text
[1] matches the COPYING.0BSD file.
FSFUL and FSFULLR are the "few permissive licenses" of certain build
system files mentioned in COPYING.
Move "Public Domain" to the end of the list because it applies only to
a few files.
[1] https://spdx.org/licenses/0BSD.html
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 15e2433e1f)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The package is part of Buildroot and thus under GPL-2.0-or-later
unless stated otherwise, as described in COPYING.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 0f80126bfe)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The package is part of Buildroot and thus under GPL-2.0-or-later
unless stated otherwise, as described in COPYING.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit c19a7fad93)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The package is part of Buildroot and thus under GPL-2.0-or-later
unless stated otherwise, as described in COPYING.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 9b8a2d8633)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The package is part of Buildroot and thus under GPL-2.0-or-later
unless stated otherwise, as described in COPYING.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 770b684426)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
All these packages are part of Buildroot and thus under
GPL-2.0-or-later unless stated otherwise, as described in COPYING.
package/skeleton-custom is excluded because it provides a way to copy
skeleton data from any location chosen by the user, the license of
which is unknown to Buildroot.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 927c575964)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://matt.ucc.asn.au/dropbear/CHANGES
- Security: Fix a use-after-free in X11 forwarding that could possibly lead
to memory corruption. This is vulnerable to authenticated users if X11
forwarding is enabled. By default X11 forwarding is not built.
In 2026.89 the server is running as the authenticated user for X11
forwarding, in earlier versions it runs as root.
This removes X11 "single connection" which has probably never been used.
Reported by @peter-pe
882f83806d
Added patch to fix build errors which would be introduced by this bump:
86baa66fc9
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 0c8bd5739c)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://gitlab.com/libssh/libssh-mirror/-/tags/libssh-0.11.5https://www.libssh.org/2026/07/21/libssh-0-12-1-and-0-11-5-security-release
s/
Fixes the following security problems:
CVE-2026-15370: Stack buffer overflow in SFTP server longname construction
CVE-2026-59843: Denial of service via zero advertised channel packet size
CVE-2026-59844: Denial of service via oversized SFTP read length
CVE-2026-59845: Denial of service via unchecked ProxyCommand fork() failure
CVE-2026-59846: Information disclosure via ProxyCommand %r username expansion
CVE-2026-59847: Integrity downgrade via OpenSSL AES-GCM tag verification
CVE-2026-59848: Denial of service via SFTP responses with unknown request IDs
CVE-2026-59849: Denial of service via automatic certificate authentication loop
CVE-2026-59850: Use-after-free via data callbacks on closed channels
Zero-initialize every ssh_string
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 93e7a673e8)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
"bzip2-1.0.4" is not a valid SPDX license identifier. The license text
in archival/libarchive/bz/LICENSE in the busybox source matches the
text associated with the bzip2-1.0.6 identifier [1].
[1] https://spdx.org/licenses/bzip2-1.0.6.html
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit f9687af626)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The Nettle manual (nettle.html in the source archive) states:
> Nettle is dual licenced under the GNU General Public License version
> 2 or later, and the GNU Lesser General Public License version 3 or
> later. When using Nettle, you must comply fully with all conditions
> of at least one of these licenses. [...]
This means "or" is the correct description. Ellipsis covers a note
that a few files are licensed more permissively.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 9a77d5547a)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
All these bundled subprojects explicitly give "Apache-2.0 OR
LGPL-2.1-or-later" in their README.md files.
Signed-off-by: Fiona Klute (Othermo GmbH) <fiona.klute@gmx.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 0c2b14a8b5)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
The patch 0002-Makefile.am-fix-build-without-makeinfo.patch is not longer
needed since it builds even if 'makeinfo' is not installed, so it's removed.
For release annoucne, see:
https://lists.gnu.org/archive/html/info-gnu/2025-12/msg00002.html
Changelog:
Changes are detailed in the git repo logs:
$ git clone git://git.sv.gnu.org/guile.git
$ git whatchanged
Signed-off-by: Pedro Aguilar <paguilar@paguilar.org>
[Julien:
- remove .checkpackageignore entry to fix check-package error
- add link to release announce
]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 643f480f43)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
When building pahole using a modern GCC (e.g. when building
host-pahole), discarded-qualifiers and unused-but-set-variable warnings
are produced. In builds that don't set CMAKE_BUILD_TYPE=Release, -Werror
gets set and causes a build failure.
Fix this by backporting an unreleased upstream patch (for
discarded-qualifiers) and adding another pending one for
unused-but-set-variable (and, arguably, a correctness issue).
Fixes: https://autobuild.buildroot.org/results/ba7/ba79fb9d08b2c6ec573b79fdbb6b4880bf603a57
Signed-off-by: Florian Larysch <fl@n621.de>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 8419f0f0c7)
[Thomas: backport to v1.28]
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Buildroot commit fabcd7c8cc bumped the
package from 28.0 to 30.0 which includes upstream commit
a01cb6e63f
that was first included in version 29.0 and added among others
-fstack-protector-all to cxx_flags guarded with a new configure option
ENABLE_HARDENING which default to ON.
This commit causes build errors with toolchains with SSP support:
[ 44%] Linking CXX executable ../bin/bitcoin
/home/autobuild/autobuild/instance-40/output-1/host/lib/gcc/i686-buildroot-linux-musl/15.2.0/../../../../i686-buildroot-linux-musl/bin/ld:
CMakeFiles/bitcoin.dir/bitcoin.cpp.o:
in function `fs::PathToString(fs::path const&)':
bitcoin.cpp:(.text+0x4c):
undefined reference to `__stack_chk_fail_local'
Upstream later removed the forementioned configure option with commit
00ba3ba303
that was first included in version 30.0 so we need to remove
-fstack-protector-all from CMakeLists.txt.
Fixes:
https://autobuild.buildroot.net/results/093/093253c79fed1c5b93decc3124e9ef4622168efe/
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit ed51b4b691)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Boot an armv5 image running the mdnsd responder alongside the mquery
client, then check that service discovery works: mquery browses for the
bundled _http._tcp service and mdnsd, on the same host, answers over the
loopback of the eth0 multicast group.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
[Julien: add "mdnsd -v" invocation to show version]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 4cd5559163)
[Thomas: remove last test that requires v1.1]
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
https://github.com/tuxera/ntfs-3g/wiki/NTFS-3G-Release-Historyhttps://seclists.org/oss-sec/2026/q3/152
Multiple vulnerabilities have been discovered in ntfs-3g.
A new version 2026.7.7 is now available at https://github.com/tuxera/ntfs-3g
(ntfscat) Fix heap memory corruption when processing a corrupt or maliciously crafted filesystem. (CVE-2026-42616)
Fix heap memory corruption when copying index data from root to an index block in a corrupt or maliciously crafted filesystem. (CVE-2026-42617)
Fix single-byte heap buffer overflow when decompressing maliciously crafted compressed file data. (CVE-2026-42618)
Fix heap buffer overflow when copying the tail data of an index block to a freshly allocated block. (CVE-2026-46569)
Fix out-of-bounds read when processing symlink reparse data in a corrupt or maliciously crafted filesystem. (CVE-2026-46571)
Fix heap memory corruption for maliciously crafted or corrupt index data descending to an out-of-bounds tree depth. (CVE-2026-46570)
Fix heap buffer overflow for maliciously crafted or corrupt index data during a node split. (CVE-2026-46572)
Fix heap buffer overflow when building inherited ACL data. (CVE-2026-56135)
Fix out of bounds access when clearing an index root in maliciously crafted or corrupt index data. (CVE-2026-56136)
Switched to sha256 tarball hash provided by upstream.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 26811cb110)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
During its configure step, libgpg-error needs awk:
configure -> src/gen-lock-obj.sh -> objdump + awk on test binary
This call is used on the output from objdump on a test binary to
retrieve the size of some data structure, and generate accordingly an
internal header (lock-obj-pub.native.h). However, libgpg-error buildroot
package does not depend explicitely on host-gawk: if host-gawk is not
part of the build, or if it is built later than libgpg-error,
libgpg-error autotool tooling will eventually pick the build machine
awk. On top of this issue, despite the configure.ac suggesting that it
supports different implementations of awk, libgpg-error configure step
will not parse correctly the output from objdump when using mawk,
leading to a wrong lock-obj-pub.native.h header being generated:
With gawk:
typedef struct
{
long _vers;
union {
volatile char _priv[24];
long _x_align;
long *_xp_align;
} u;
} gpgrt_lock_t;
#define GPGRT_LOCK_INITIALIZER {1,{{0,0,0,0,0,0,0,0, \
0,0,0,0,0,0,0,0, \
0,0,0,0,0,0,0,0}}}
With mawk:
typedef struct
{
long _vers;
union {
volatile char _priv[0];
long _x_align;
long *_xp_align;
} u;
} gpgrt_lock_t;
#define GPGRT_LOCK_INITIALIZER {1,{{}}}
This issue has been observed for example on Debian 13 with mawk
installed: it fails to parse mtx_size, leading to a size 0, and so
generating crashes at runtime because of the invalid size embedded in
libgpg-error.
Side note: the lock-obj-pub.native.h header file is correctly
generated, when using mawk 1.3.4 20200120 from Debian 12.
The described issue happen with mawk 1.3.4 20250131 from Debian 13.
libgpg-error should be fixed upstream to properly handle this data
structure size guessing even when using mawk, but anyway the
corresponding buildroot package should not randomly use the build
machine host tooling or buildroot-provided host tooling depending on
whether host-gawk has been selected and built before it.
Enforce an explicit dependency on host-gawk for libgpg-error to make
sure that it systematically uses the buildroot-provided awk.
Co-developped-by: Bernard Gautier <bernard.gautier@nav-timing.safrangroup.com>
Signed-off-by: Bernard Gautier <bernard.gautier@nav-timing.safrangroup.com>
Co-developped-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
[Julien: add the side note that mawk from debian 12 works]
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 21bf78411e)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
Buildroot commit 2774502344 bumped the
package from 4.2.2 to 6.2.3. Upstream added the usage of
std::scoped_lock in version 6.2.0:
a760b3278a
causing a build error detected by the Gitlab pipelines for the
bootlin-aarch64-glibc-old defconfig:
../src/inc/upnpapi.h:132:22: error:
‘scoped_lock’ is not a member of ‘std’
scoped_lock was introduced in gcc 7.1:
https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2017
so we raise the minimum required gcc version accordingly.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Julien Olivain <ju.o@free.fr>
(cherry picked from commit 5f543cfcda)
Signed-off-by: Thomas Perale <thomas.perale@mind.be>
