The libgpiodbus shared library and the .typelib file for libgpiod GLib
bindings are currently packaged in the main libgpiod package which
should really only contain the core libgpiod shared library. Move them
to libgpiod-manager and libgpiod-glib packages respectively.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Bug-fix release addressing several issues discovered during an
AI-augmented security audit. The most severe bug was found in the C
extension code of the python bindings - which also get an update - but
there were some memory leaks and integer overflow bugs in the core C
library as well as in tools and DBus daemon.
Full changelog:
Bug fixes:
- fix buffer over-read bugs when translating uAPI structs to library types
- fix variable and argument types where necessary
- sanitize values returned by the kernel to avoid potential buffer overflows
- fix memory leaks in gpio-tools
- add missing return value checks in gpio-tools
- fix period parsing in gpio-tools
- use correct loop counter in error path in gpio-manager
Improvements:
- make tests work with newer coreutils by removing cases checking tools'
behavior on SIGINT which stopped working due to changes in behavior of the
timeout tool
Also: drop the patch that's now upstream from the recipe.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
Inherit python3targetconfig so that cross python3-config is used instead
of native.
This fixes the below error when building on a 64 bit host for a 32 bit
target:
| In file included from /workspaces/yocto/build/tmp/work/core2-32-poky-linux/libgpiod/1.6.5/recipe-sysroot-native/usr/include/python3.14/Python.h:72,
| from ../../../sources/libgpiod-1.6.5/bindings/python/gpiodmodule.c:8:
| /workspaces/yocto/build/tmp/work/core2-32-poky-linux/libgpiod/1.6.5/recipe-sysroot-native/usr/include/python3.14/pyport.h:429:2: error: #error "LONG_BIT definition appears wrong for platform (bad gcc/glibc config?)."
| 429 | #error "LONG_BIT definition appears wrong for platform (bad gcc/glibc config?)."
| | ^~~~~
Signed-off-by: Ricardo Simoes <ricardo.simoes@pt.bosch.com>
Signed-off-by: Mark Jonas <mark.jonas@de.bosch.com>
Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
openembedded-core updated to coreutils v9.10 in which commit
8c2461933411 ("timeout: honor ignored signal dispositions") changed the
behavior of timeout. It will no longer propagate SIGINT to background
shell jobs which breaks test-cases for gpio-tools which verify that they
exit correctly after receiving SIGINT. This backports the patch sent
upstream that removed the offending test-cases as we already have a
similar set of tests for SIGTERM.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a bugfix release addressing issues in tools and core library. No
API changes.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
In order to facilitate using e.g. dbus-broker as runtime D-Bus
implementation, change the Freedesktop.org specific D-Bus runtime
dependency declarations to honor oe-core's new VIRTUAL-RUNTIME_dbus
variable.
Signed-off-by: Niko Mauno <niko.mauno@vaisala.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The main set of changes in this release doesn't concern yocto - it
brings in proper sphinx docs infrastructure for publishin on
readthedocs. We still have a bunch of code fixes though and we can also
remove the patch we're carrying in the recipe as it's upstream now.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Linux kernel commit 8bd76b3d3f3a ("gpio: sim: lock up configfs that an
instantiated device depends on") revealed an issue in libgpiod bindings
tests where a GPIO simulator device is not disabled before removing its
configfs entries.
Add a patch (already submitted upstream) that fixes the observed ptest
failures.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
We already set EXTRA_OECONF in libgpiod.inc, drop the second assignment.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Make the packaging a bit more fine-grained. Create separate packages for
the libgpiosim library and gpio-manager configuration files. Make the
ptest package run-time depend on only the gpio-manager config. This way,
we don't pull the entire daemon infrastructure if all we want is to run
its test suite.
While at it: add GPIO udev rules to the gpio-manager config package.
Currently they are in the main libgpiod package where they don't really
belong.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
There's an additional, unneeded newline at the end of libgpiod.inc.
Remove it.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* I thought that PN in FILESEXTRAPATHS worked as everywhere else and
as described in:
https://lore.kernel.org/all/2261745fc01a4e3887ee98f98a425e88@vivavis.com/T/
* but surprisingly (to me)
FILESEXTRAPATHS:prepend := "${THISDIR}/${PN}-2.x:"
ends with effectively ${BPN}-2.x in bitbake -e (and the build doesn't fail
to fetch file://gpio-manager.init in lib32-libgpiod build
* I've tried with kirkstone and dunfell as well and it ends with
"libgpiod-2.x" for all 3 PNs:
dunfell $ grep ^FILESEXTRAPATHS= env.* | sed 's@/OE.*dunfell/@@g'
env.lib32-libgpiod:FILESEXTRAPATHS="meta-oe/meta-oe/recipes-support/libgpiod/libgpiod-2.x:__default:"
env.libgpiod:FILESEXTRAPATHS="meta-oe/meta-oe/recipes-support/libgpiod/libgpiod-2.x:__default:"
env.libgpiod-native:FILESEXTRAPATHS="meta-oe/meta-oe/recipes-support/libgpiod/libgpiod-2.x:__default:"
dunfell $ grep ^PN= env.*
env.lib32-libgpiod:PN="lib32-libgpiod"
env.libgpiod:PN="libgpiod"
env.libgpiod-native:PN="libgpiod-native"
There is the QA check for PN in SRC_URI, maybe I took it a bit
further and updated all FILESEXTRAPATHS references as well in our
layers (in 2023) even when it wasn't causing build failures, e.g.:
38d3cbf010
* so it's not so important to use BPN, but it's a bit clearer as that's
what it ends evaluated into and what default FILESPATH is using as well:
meta/classes-global/base.bbclass:FILESPATH = "${@base_set_filespath(["${FILE_DIRNAME}/${BP}", "${FILE_DIRNAME}/${BPN}", "${FILE_DIRNAME}/files"], d)}"
meta/conf/bitbake.conf:#FILESPATH = "${FILE_DIRNAME}/${PF}:${FILE_DIRNAME}/${P}:${FILE_DIRNAME}/${PN}:${FILE_DIRNAME}/${BP}:${FILE_DIRNAME}/${BPN}:${FILE_DIRNAME}/files:${FILE_DIRNAME}"
BPN was added in denzil 2009 (and FILESPATH moved to base.bbclass,
bitbake.conf has commented-out value since then):
https://git.openembedded.org/openembedded-core/commit/?h=denzil&id=94c895aad5ad286f172b04bc33ba670220d5eba8
PF was dropped in dylan 2012:
https://git.openembedded.org/openembedded-core/commit/?h=dylan&id=3efa13cd76bbd5611805021945fc9def88d9fd93
P and PN were dropped shortly after in dylan 2012:
https://git.openembedded.org/openembedded-core/commit/?h=dylan&id=fb359583b659cda643973fa285002aaffb729a51
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The oe-core bbclass for GObject introspection sets GI_DATA_ENABLED if
the relevant distro feature is enabled. That means that it's up to the
inheriting recipe to disable it if needed.
For libgpiod: we only want to enable it if any of the GLib based
packages is being generated (GLib bindings, D-Bus daemon or ptest).
This also effectively reverts commit 78a332939f8b (libgpiod_2.2: depend
on glib-2.0) which indiscriminately pulls in GLib for all packages
generated by this recipe which feels like shooting a fly with a cannon.
Fixes: 9626e8a2c8f2 ("libgpiod: update to v2.2")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* the default from bitbake.conf:
S = "${WORKDIR}/${BP}"
works better, because it points to correct directory even with e.g. when
building lib32-libgpiod with multilib
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
In my setup a build fails with
| checking for gobject-introspection... configure: error: gobject-introspection-1.0 is not installed
| NOTE: The following config.log files may provide further information.
| NOTE: /var/home/krm/build/oe-core_master/build/tmp/work/cortexa55-tdx-linux/libgpiod/2.2/build/config.log
| ERROR: configure failed
From config.log this seems to be related to glib-2.0 not found:
| configure:22484: $PKG_CONFIG --exists --print-errors "gobject-introspection-1.0"
| Package glib-2.0 was not found in the pkg-config search path.
| Perhaps you should add the directory containing `glib-2.0.pc'
| to the PKG_CONFIG_PATH environment variable
| Package 'glib-2.0', required by 'gobject-introspection-1.0', not found
| configure:22487: $? = 1
| configure:22491: error: gobject-introspection-1.0 is not installed
Note that glib-2.0.pc is present in recipes-sysroot-native but not
in recipes-sysroot.
Adding glib-2.0 to depends allows successful build of libgpiod 2.
Fixes: 117d09ce43c7 ("libgpiod: update to v2.2")
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a big update for libgpiod bringing in the D-Bus daemon and its
command-line client. This is why the patch is pretty hefty. The majority
of the line count comes from the removed patch that is now upstream but
the v2.2 recipe also now contains a lot of bits and pieces that make the
D-Bus daemon work both with system as well as sysvinit.
The project now packages the systemd service and udev rules that allow
the manager to run under its own user in a well sandboxed environment
but the sysvsinit script is carried in the recipe as I don't want to
put support for legacy systems in libgpiod repo.
The patch also updates the ptests for libgpiod to run three new
test-suites: one for GLib bindings, one for the command-line D-Bus
client and finally one testing some daemon corner-cases not supported by
the client.
Build tested with various config options with systemd and sysvinit (the
latter with glibc and musl) on 32- and 64-bit architectures.
The complete changelog since v2.1 (taken verbatim from the NEWS file) is
below:
libgpiod v2.2
=============
New features:
- add GObject bindings with introspection
- add a D-Bus interface to libgpiod together with a daemon implementing it and
a command-line client
- split out the common shell test code for gpio-tools into a reusable test
harness for GPIO command-line tools
- add minutes as a new supported time unit for tools and allow longer periods
for timeouts, line holding, etc.
- add a script for generating sdist and wheels for python bindings
- migrate C++ tests to using Catch2 v3
Improvements:
- relicense C++ bindings under LGPL-2.1-or-later in order to make the project
more attractive to users with GPL-3.0 restrictions
- remove dependency on grep from tools' tests
- make shell scripts pass shellcheck treewide
- use ppoll() in tools to actually achieve the advertised microsecond
granularity of timeouts
- documentation improvements
- improve typing info in python bindings
- improve __repr__() implementations in python bindings
- make reconfiguring lines more flexible in python bindings by relaxing the
requirement to carry the ordering of the config entries from the request
- support casting of line.Value to bool in python bindings
- various new test-cases for improved coverage treewide
- align the test cases with the current kernel requirement for specifying
direction explicitly when reconfiguring lines
Bug fixes:
- fix C++ bindings build using slibtool
- accept the new style automatic GPIO chip labels from gpio-sim in bash tests
- fix C++ tests with recent kernels which introduced stricter reconfigure
behavior
- fix a use-after-free bug in python bindings
- fix passing the event clock property to line requests in python bindings
- fix a memory leak in tools
- make sure the string buffers in line-info and chip-info are big enough to not
truncate the strings they hold below the size accepted by the kernel
- remove buggy and unnecessary flags sanitization from line-config
- fix python bindings installation with Makefile build
- sanitize the return values of GPIO ioctl()s which in some cases may be
erroneously positive
- fix requesting lines by name with multiple entries in python bindings
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Bug fixes:
- fix C++ tests with recent kernels which introduced stricter reconfigure
behavior
- fix a use-after-free bug in python bindings
- fix passing the event clock property to line requests in python bindings
- fix a memory leak in tools
- make sure the string buffers in line-info and chip-info are big enough to not
truncate the strings they hold below the size accepted by the kernel
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Catch2 recipe is moving to version 3 of the Catch2 API which will make
the C++ tests in libgpiod uncompatible. libgpiod v2 as ported to the new
interface but v1.6.x will not be so just disable building the C++ tests.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This release contains a build fix for environments that require libgen.h
to be included for basename() and don't pull it implicitly with any
other header.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a small bugfix release that fixes a build issue with slibtool
(not relevant to meta-oe ATM), makes the licensing of C++ bindings less
restrictive and preemptively fixes an issue that will be triggered with
linux v6.9 when running gpio-tools ptest suite.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The bash in ptest RDEPENDS would cause following QA error:
ERROR: QA Issue: lib32-libgpiod package lib32-libgpiod-ptest-dev -
suspicious values 'bash-dev' in RRECOMMENDS [multilib]
Add a ptest enable dependency to avoid this QA error happens.
Signed-off-by: Xiangyu Chen <xiangyu.chen@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a bug-fix release addressing a couple issues in core libgpiod.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a small release bringing in a bunch of improvements and a couple
new interfaces. Details are available in NEWS.
We need to move the code in recipes around a bit as the tests now use
shunit2.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a small bug-fix release addressing a segfault, a memory
corruption bug and a couple minor issues.
This also removes the patch that's already upstream.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
No longer used in generating packages
Also creates a possible confusion with the recipe maintainer
name.
Signed-off-by: Michael Opdenacker <michael.opdenacker@bootlin.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The current use of RDEPENDS to add a dependency on bats results
in the QA warning/error
lib32-libgpiod package lib32-libgpiod-ptest-dev - suspicious values
'bats-dev' in RRECOMMENDS [multilib]
when building lib32-libgpiod with ptest not enabled. We add the
dependency only if ptest is enabled.
Signed-off-by: Joe Slater <joe.slater@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Look for level transitions when testing toggling
values because using fixed delays to assume value
changes is not reliable.
Signed-off-by: Joe Slater <joe.slater@windriver.com>
Reviewed-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
While RECIPE_MAINTAINER doesn't seem to be used anywhere, let's add it
in hopes that it will at least make people Cc me on any recipe patches
in the future.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Look for level transitions when testing toggling
values because using fixed delays to assume value
changes is not reliable.
Signed-off-by: Joe Slater <joe.slater@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
There's no need to put the test executables into ${bindir} as they're
already installed into ${PTEST_PATH}. This just pollutes $PATH.
Future libgpiod releases will not install test binaries at all, leaving
it to the library users. For now: remove them manually after
installation.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Now that python3-gpiod fetches the sources from pypi, we no longer need
to provide the SRC_URI in a separate include. Fold it into the
libgpiod.inc used locally only.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This header is used for building tests. While libgpiod can access it
from the source tree, the python bindings that live in a separate recipe
must be able to use it from its recipe sysroot.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a bug-fix release addressing bugs discovered since the release
of v2.0.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
In order to avoid having to rename the directories containing ptest
scripts on every update, use a more generic, per-major-release name and
extend the FILESEXTRAPATHS.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Assigning a value to S is not necessary and it will in fact break the
recipe when updating the version so drop it.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
We will fail to install or run all the tests if we don't enable all the
features exposed by the libgpiod project.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The final v2.0 release is out so update the recipe.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
I decided to make one more RC before the final release as there was one
late minor change to C++ interface (marking public classes as final).
Next week libgpiod v2.0 should be finally carved in stone.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a recipe for libgpiod v2 release candidate. Let's put common
bits and pieces into an .inc file and the source fetching into a
separate include that will be later reused by the python bindings that
for v2 will be moved into a separate recipe in meta-python.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This is a bug-fix release bringing in several fixes for build and
testing issues.
Remove the patch that's already present in this release.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
It seems that automake's AM_PATH_PYTHON() macro no longer sets the
$PYTHON variable. This means the python bindings build for libgpiod has
been broken for some time. As we know which version we're using -
python3 - let's hardcode it in configure.ac. For now add a patch that's
been merged upstream and once a new bugfix release is out, we'll remove
it.
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Allow the ptest packages to be disabled even if ptest is enabled in
DISTRO_FEATURES.
Signed-off-by: Alex Kiernan <alex.kiernan@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This reverts and reworks commit e75cc87c4f944dff766c426fafafd48c378544fe
These tools are needed by test-suites, not the package itself.
I do not want bash (gpl3) or python (big/lot of dependencies)
in my release image when I have ptest enabled in my distro for
sdk/testing image.
Signed-off-by: Peter Marko <peter.marko@siemens.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
