Each time a new pipeline is triggered, some jobs may fail due to
temporary issue with a Gitlab runner (network, power supply, docker or
maintainance).
Most of the problems are "runner system failure" [1] and require to
retart each failed jobs manually by maintainers to complete the
pipeline with only real failures if any.
The "retry" keyword allows to configure how many times a job is retried
if it fails. "retry:when" allows to retry a failed job only on
specific failure types like "runner_system_failure".
While at it, retry a job if it failed due to a timeout failure (this
timeout means that the job was pending for more than 24h) [2].
Such timeout failures occur on pipelines testing each Buildroot's
defconfig since there is not enough gitlab runner available to build
all of them within 24h.
Retry only jobs that are more likely to wait for a runner
(generate-gitlab-ci-yml, runtime_test_base, defconfig_base and test_pkg).
[1] https://gitlab.com/buildroot.org/buildroot/-/jobs/4936949397 (runner system failure)
[2] https://gitlab.com/buildroot.org/buildroot/-/jobs/4936949530 (timeout failure or the job got stuck)
https://docs.gitlab.com/ee/ci/yaml/#retrywhen
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Arnout Vandecappelle <arnout@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit e0166ecba0)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
While switching ATF to github, the BR2_TARGET_UBOOT_CUSTOM_GIT=y
symbol was not removed. Since then this defconfig fail to build
in gitlab-ci due to invalid defconfig check.
WARN: defconfig ./configs/freescale_imx6qsabresd_defconfig can't be used:
Missing: BR2_TARGET_UBOOT_CUSTOM_GIT=y
[1] dd42b159a5
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/4889436612
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 0ee8ef970b)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bugfix release, fixing a number of regressions in 2.0.16
From the changelog
(https://github.com/eclipse/mosquitto/blob/master/ChangeLog.txt)
2.0.17 - 2023-08-22
===================
Broker:
- Fix `max_queued_messages 0` stopping clients from receiving messages.
Closes#2879.
- Fix `max_inflight_messages` not being set correctly. Closes#2876.
Apps:
- Fix `mosquitto_passwd -U` backup file creation. Closes#2873.
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit c11950fe4a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add local patch to allow to override CFLAGS and undefine
CONFIG_LITTLE_ENDIAN by default and use the correct endianness according
to target architecture.
Fixes:
http://autobuild.buildroot.net/results/fe67db3884573ef750eda9d0dccd5f97b3ae698e
Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 624814f4ec)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
See release announce:
https://lists.gnu.org/archive/html/screen-users/2023-08/msg00000.html
Fixes:
CVE-2023-24626: https://www.cve.org/CVERecord?id=CVE-2023-24626
Note: Buildroot installs screen as setuid, so the described scenario
in CVE applies.
This commit also rebases all patches on this release. Patch were
regenerated with 'git format-patch -N', so patch file name changed in
this process. The file .checkpackageignore is also updated accordingly.
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 96a54b0907)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit 0a8ef2f3f7 bumped the headers
version requirements, but did not update the associated comment.
Remove the comment entirely, as it does not apply anymore.
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit bdaade3e34)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This package is affected by gcc bug 43744 and I have not found a work
around for it(i.e. the common -O0 we use or other), so let's disable it if
gcc has such bug.
Fixes:
http://autobuild.buildroot.net/results/ab289769c5fea435934ed260d38e0a4fdd2ba72d
Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 6b3d0c5adb)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
In commit 30997eaa65438a2ce726ad8a204ac5a36363f5c8 a mistake
was made. Guard the definition correctly.
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 3f46db39e6)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Release notes: https://www.samba.org/samba/history/samba-4.18.6.html
Includes a mitigation for CVE-2007-4559.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 2235ab809f)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes CVE-2023-34241 (see [0] for details)
[0] https://github.com/OpenPrinting/cups/releases/tag/v2.4.6
The number of changes between 2.4.4 and 2.4.6 is really small, and
limited to bug fixes:
9d614a4b3184205294c55355a1d2eb54d4532ccd (tag: v2.4.6) Update CHANGES.md
6f6da74ec284e28c156f0b9f62f3bd610e61aa78 Fix use-after-free in cupsdAcceptClient() (fixes CVE-2023-34241)
3f12185ca9cbb5350a6370d6046066907b8abc12 Merge pull request #735 from AtariDreams/Fixer
f5281777c80cdf820a2a71c9e7f08b91f0e11160 Fix compilation on older macOS versions
ee82c5b18409def3ec1424ce2eb343aabb0ff0d1 Merge pull request #730 from zdohnal/cupssinglefile_24x
1504527b2415a4b67b0e3e17593b053f3628746f cups/ppd-cache.c: Put cupsSingleFile into generated PPD
3be1d5da8fe9ee13aab5ee6ecc11b2f9387821a6 Prepare files for next release
c1f54ec966ccc5d5564eed95dcb540842af7b5ca (tag: v2.4.5) cups/cups.h: Update for 2.4.5
70dba05b7511a96476ea0ef8fe1d92c6500c6e61 Finish hotfix release 2.4.5
87f5cb7d8f0da8fa2835bb0aa3ca48b5e5a66a3f Merge pull request #727 from AtariDreams/hotfix
61aa0b259183fe59124566f08ecf649bb806cd24 Regression: Certificate data is corrupted during base64 conversion
7362f41c45d834564f876ffac536f59eece843ec Prepare files for next release
Signed-off-by: Clement Ramirez <ramirez.clement3@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit db46428e0a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Updated license hash due to copyright year bump:
52a3f2a546
Needed for freeswitch bump to 1.10.10:
7c1faeff48
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 171977f4bf)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Bump the kernel to the latest release of 5.4 kernel series
to fix an issue with gcc 12.
arch/mips/jz4740/setup.c: In function 'plat_mem_setup':
arch/mips/jz4740/setup.c:64:25: error: comparison between two arrays [-Werror=array-compare]
64 | if (__dtb_start != __dtb_end)
| ^~
arch/mips/jz4740/setup.c:64:25: note: use '&__dtb_start[0] != &__dtb_end[0]' to compare the addresses
CC fs/debugfs/file.o
cc1: all warnings being treated as errors
Fixed in v5.4.195:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=a60def7568216684f62b117c199429c4a9a89570
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/4889436573
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 486a3ba4c3)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
objtool built by the kernel requires libelf
ldd output/build/linux-6.1.24/tools/objtool/objtool
linux-vdso.so.1
libelf.so.1 => output/host/lib/libelf.so.1
While updating the kernel [1] we forgot to select
BR2_LINUX_KERNEL_NEEDS_HOST_LIBELF to provide Buildroot's host-libelf.
Using host-libelf avoid linking with libelf installed on the host or
failing to build objtool if libelf is not installed.
[1] d45538f2e7
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/4889436869https://gitlab.com/buildroot.org/buildroot/-/jobs/4889436872
Signed-off-by: Romain Naour <romain.naour@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit b661f635b5)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This is the follow-up commit to the 4de60e4 which disables
sourcehighlight dependency for the target GDB package.
Do the same for the host variant to be sure that this
won't be picked up from the host system.
Signed-off-by: Bartosz Bilas <b.bilas@grinn-global.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 9ba399a3dd)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
nftables provides python bindings; it uses setuptools to install them.
We currently install those bindings by telling the nftables buildsystem,
autotools, to install the python bindings.
However, we do not pass any of the environment variables that are needed
for setuptools packages. When host-python-setuptools is installed before
nftables is built [0], this breaks the system at runtime, as the
bindings are not installed; only the egg is, resulting in runtime errors
like:
# python -c 'import nftables'
Traceback (most recent call last):
File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'nftables'
Upstream has been doing some changes on their python handling, but it is
not in a released version yet, and we can't backport those changes
either, due to other big changes.
Instead, we split the pyhon bindings to their own package.
For legacy handling, we make that new package default to y, so that
existing (def)config still work. The only novelty is that it can be
disabled now.
Many thanks to Julien for testing and finding the offending dependency,
to James for suggesting the package split, and to Adam for, well,
trigerring the issue in the first place! ;-p
Note: a git bisect of the issue turns up 72 candidates for the breakage,
all around the time we dropped python2 support in early 2022; the last
known-good commit is 55df30f8b1 (package/zfs: drop python2 support)
and the first known-bad commit is 697acda00d (package/pkg-python: drop
python2 host/setuptools support); everything in-between does not
configure (package/python/Config.in.host still sourced but already
removed), or does not build (host-python still in the dependency chain
but already removed), so had to be skipped during the bisect.
[0] This can happen when another python package using setuptools is
built before nftables. However, with PPD, this never happens because
host-python-setuptools is never in the dependency chain of nftables.
Reported-by: Julien Olivain <ju.o@free.fr>
Tested-by: Julien Olivain <ju.o@free.fr>
Suggested-by: James Hilliard <james.hilliard1@gmail.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
Cc: Adam Duskett <aduskett@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit eb74998125)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fixes the following security issues:
- CVE-2023-28366: Fix memory leak in broker when clients send multiple QoS 2
messages with the same message ID, but then never respond to the PUBREC
commands.
- CVE-2023-0809: Fix excessive memory being allocated based on malicious
initial packets that are not CONNECT packets.
- CVE-2023-3592: Fix memory leak when clients send v5 CONNECT packets with a
will message that contains invalid property types.
- Broker will now reject Will messages that attempt to publish to $CONTROL/.
- Broker now validates usernames provided in a TLS certificate or TLS-PSK
identity are valid UTF-8.
- Fix potential crash when loading invalid persistence file.
- Library will no longer allow single level wildcard certificates, e.g. *.com
For more details, see the changelog:
https://mosquitto.org/ChangeLog.txt
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 9f1ca943d5)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Commit e3322e2c7 "lsof: add license info" added the package license
info. At that time, lsof was at version 4.85. The package was not
including a dedicated license file. Instead, the license text was
copied in many files (source files, readmes, ...). It was then decided
to use the small source file "dialects/linux/dproto.h".
Recently, lsof added a COPYING license file. See [1]. This commit was
first included in lsof version 4.97.0. We can now use this file.
The license file hash is updated due to formatting changes. As a side
note, the copyright year in source file was 1997 (see [2]), whereas in
the new license file it is 2002. Source files have different copyright
dates, and can be as old as 1994, for example in [3]. The rest of the
license text remains unchanged.
[1] 62dab61cae
[2] https://github.com/lsof-org/lsof/blob/4.98.0/dialects/linux/dproto.h#L9
[3] https://github.com/lsof-org/lsof/blob/4.98.0/arg.c#L7
Signed-off-by: Julien Olivain <ju.o@free.fr>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 64490ea84c)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The compiler-rt project intructs to build out of tree:
https://compiler-rt.llvm.org/
Otherwise, the dependency chain declared in the include/CMakelLists.txt
causes a circular dependency of source files upon themselves:
make[4]: Circular include/sanitizer/allocator_interface.h <- include/sanitizer/allocator_interface.h dependency dropped.
make[4]: Circular include/sanitizer/asan_interface.h <- include/sanitizer/asan_interface.h dependency dropped.
make[4]: Circular include/sanitizer/common_interface_defs.h <- include/sanitizer/common_interface_defs.h dependency dropped.
[--snip 19 other files--]
This is because include/CMakeLists.txt unconctional declares
dependencies in this manner:
set(SANITIZER_HEADERS
sanitizer/allocator_interface.h
sanitizer/asan_interface.h
sanitizer/common_interface_defs.h
[...]
)
set(COMPILER_RT_HEADERS
${SANITIZER_HEADERS}
[...])
set(output_dir ${COMPILER_RT_OUTPUT_DIR}/include)
foreach( f ${COMPILER_RT_HEADERS} )
set( src ${CMAKE_CURRENT_SOURCE_DIR}/${f} )
set( dst ${output_dir}/${f} )
add_custom_command(OUTPUT ${dst}
[...]
)
[...]
endforeach( f )
The froeach() loop creates dependency rules between the files in
output_dir and CMAKE_CURRENT_SOURCE_DIR, without provision for the
case they are both the same directories, thus in-source builds are
not supported.
With the Makefiles backend, this only triggers the above warning
from make, because make arbitrarily breaks circluar dependencies, and
in this case it makes it work. But when we switch to the ninja backend,
this is going to be a hard error.
Anyway, ninja or make, compiler-rt does not support in-source builds.
Signed-off-by: Thomas Devoogdt <thomas@devoogdt.com>
[yann.morin.1998@free.fr: drastically expand commit log]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit dedb686dd2)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Since it was introduced in commit d4ecefa3cc (package/perftest: new
package), perftest was missing the reverse dependency on the kernel
headers version inherited from rdma-core.
In commit 3279d8b9c4 (package/rdma-core: needs NPTL), the dependency
of rdma-core on NTPL was not propagated to its reverse dependencies,
notably perftests.
Fixes:
- http://autobuild.buildroot.net/results/00c/00ce19a1589c6028ee77c333ed9107e2c8c8d6de
- d4ecefa3cc
- 3279d8b9c4
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
[yann.morin.1998@free.fr:
- also add misisng headers dependency
- expand commit log
]
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 00cb5f5e4a)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fix the following build failure:
>>> toolchain-external-codescape-img-mips 2018.09-02 Copying gdbserver
Could not find gdbserver in external toolchain
Fixes:
- http://autobuild.buildroot.org/results/b0786965e0b249c8168df855682e54cfe95fa0cc
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 76af307f01)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Fix the following build failure:
>>> toolchain-external-codescape-mti-mips 2018.09-02 Copying gdbserver
Could not find gdbserver in external toolchain
Fixes:
- http://autobuild.buildroot.org/results/ead304285a9cf32bf6b9dcc7fc30acece25ec680
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit f7ebab67ab)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Add a custom case to make sure that a random configuration with an empty
configuration file for ubi doesn't fail.
ubinize: error!: no sections found the ini-file "/home/buildroot/autobuild/instance-2/output-1/build/ubinize.cfg"
Fixes:
- http://autobuild.buildroot.org/results/f678c17cc4df06fb2737467e769cd8f72a3ea420
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 283e6859f1)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Use full relative pathes for rootfs and kernel, like in other
readme's.
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 0a5bb69982)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Urcu does not like -O0 it needs gcc optimizations.
Force -O1 in this case. Inspired by package fwts.
The issue got introduced in commit 2f8f69012e:
before this commit, xfsprogs builds fine at -O0, after this commit, it fails
to build at -O0. As this commit went into 2022.02, the fix needs to be
backported to stable branches.
Fixes:
- http://autobuild.buildroot.net/results/3bd/3bd7de784953ab1fa70bf5572eccd7b10e4a0e17
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 2bd93a30a6)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
This was previously only defined for aarch64, define the missing
stuff for all architectures.
This is required for the zeek package.
Fixes:
- http://autobuild.buildroot.net/results/f32/f32777b27c6b4f5a41c255df53502164d21b0fd4
Signed-off-by: Waldemar Brodkorb <wbx@openadk.org>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 55b47a4060)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
gcc 12 added a warning that triggers on access to low addresses. Add a
patch to allow access since this is normal for low level code.
Rebase our existing patch on top. While at it, add also a proper
Upstream tag.
Fixes:
https://gitlab.com/buildroot.org/buildroot/-/jobs/4795673785
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
(cherry picked from commit 651429d0b8)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
The dependency is not provided by Buildroot, and it can potentially be
found by gdb's build system from the host system, which results in the
following build errors:
../../gdb/source-cache.c:37:10: fatal error: srchilite/sourcehighlight.h: No such file or directory
37 | #include <srchilite/sourcehighlight.h>
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
make[4]: *** [Makefile:1645: source-cache.o] Error 1
make[4]: *** Waiting for unfinished jobs....
Signed-off-by: Bartosz Bilas <b.bilas@grinn-global.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 4de60e41c3)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
screenfetch requires an "awk" implementation, but upstream would
only use an executable called "awk": pick up upstream patch that
can autodetect awk implementation, so that "mawk" can be used.
Signed-off-by: Nicolas Boichat <drinkcat@google.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit ff6d6c79d3)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Most of the users use HDMI console on RPi and with increasing popularity
of systemd they face a trouble.
systemd doesn't use /etc/inittab, enable getty@tty1.service instead.
Signed-off-by: Alexey Roslyakov <alexey.roslyakov@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit c415820662)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
xlib_libXpresent is needed with x11 to avoid the following build failure
raised since bump to version 0.35.1 in commit
61ba55e9cc and
3d459832a8:
Checking for X11: no ('x11 >= 1.0.0 xscrnsaver >= 1.0.0 xext >= 1.0.0 xinerama >= 1.0.0 xpresent >= 1.0.0 xrandr >= 1.2.0' not found)
You manually enabled the feature 'x11', but the autodetection check failed.
Fixes:
- http://autobuild.buildroot.org/results/1aace6210ba966bb4979d935c08507b2a34e0460
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit a0fdd17ec2)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Needed for mpv since version 0.35.0:
3d459832a8
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
[Thomas: improved thanks to feedback from a similar patch submitted by
Bernd Kuhls.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 3bdc3f58cc)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Until now, the Python build system was building a _dbm.cpython-*.so
native module when BR2_PACKAGE_PYTHON3_BERKELEYDB was enabled, but
since the bump of Python to 3.11.x, it was no longer building this
native module.
Investigation this, we found out that Python 3.11 changed how libdb
was detected [1] and that check now fails:
quoting configure log:
checking for libdb... no
quoting python config.log:
conftest.c:(.text.startup+0x8): undefined reference to `__db_ndbm_open'
However, in fact it turns that this problem is not new in Python
3.11. In Python 3.10, the build system was always producing the native
module, but it was in fact not working at runtime:
>>> from _dbm import *
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ImportError: /usr/lib/python3.10/lib-dynload/_dbm.cpython-310-arm-linux-gnueabihf.so: undefined symbol: __db_ndbm_delete
It was not really visible because when one does "import dbm", it
actually tries several "backends", including ndbm (which relies on
_dbm above), and if it doesn't work, it falls back silently on a pure
Python implementation.
So the issue was never noticed, but has already been there,
potentially forever.
In order for this _dbm native module to be built (Python >= 3.11) or
to work (Python < 3.11), the BerkeleyDB library need to be built with
its so-called "dbm" interface, which we do by selecting select
BR2_PACKAGE_BERKELEYDB_DBM.
And now:
>>> import _dbm
>>>
[1] d2340ef257/configure.ac (L4002)
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
[Thomas: did more research to have a better explanation of what is
happening, and realize the problem is not related to Python 3.11]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit b84ffd85e2)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
Berkeley DB support in python depends on dbm, when missing python fails
to detect libdb:
d2340ef257/configure.ac (L4002)
quoting python configure log:
checking for libdb... no
quoting python config.log:
conftest.c:(.text.startup+0x8): undefined reference to `__db_ndbm_open'
This patch provides a configure option for the historic dbm interface
to be used by the python package.
Signed-off-by: Bernd Kuhls <bernd@kuhls.net>
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
(cherry picked from commit 244a8c52ef)
Signed-off-by: Peter Korsgaard <peter@korsgaard.com>
