Changelog:
- add support for free-threaded (no-GIL) CPython
- generate wheels for CPython 3.14t (free-threaded)
- drop support for Python 3.9
- migrate the C extension and test modules to multi-phase module
initialization
- release the GIL in chip_get_info()
- fix heap-buffer overflow bugs when setting/getting line values
- check for errors after PyIter_Next() when setting output values
- check for errors after PyIter_Next() when reading request values
- fix reference count imbalances (missing Py_None and line-name
increments, erroneous decref on PyList_SetItem() failure)
- allow closing a Chip multiple times without raising an error
- allow releasing a LineRequest multiple times without raising an error
- remove a duplicated edge-detection setting
- add a CHANGELOG.md to track changes between releases
- copy the license file into source and binary distributions
- require setuptools v77 or newer
- remove wheel from the build dependencies
- add multi-threaded tests
- add a dedicated lint dependency group and update linter configuration
- apply linter recommendations and add type annotations across examples,
setup and helper scripts
- modernize the C extension using upgrade_pythoncapi.py suggestions
- minor internal cleanups and performance improvements
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Port the project from autotools to meson. All build options are
preserved with the same semantics.
The goal of this change is to produce the same set of files as the
current autotools approach: libraries with correct SONAME/version,
headers, pkg-config files, GObject introspection data, Python extension
modules, D-Bus interface and configuration files. This is done to enable
a one-liner test command running all available test suites.
The only thing we drop is the `make docs` target as it's rarely used and
docs can be generated easily with sphinx directly. They are also not
required for testing.
The Rust workspace is built via a custom target that invokes cargo and
uses SYSTEM_DEPS_* environment variables to point the libgpiod-sys and
gpiosim-sys crates at the locally built C libraries.
In addition to porting the build files to meson, update the docs, remove
unneded items from .gitignore files (or remove them entirely if they're
no longer needed) and update the shunit2 test scripts to no longer care
about source directories.
Tested-by: Erik Wierich <erik@riscstar.com>
Tested-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://patch.msgid.link/20260428-meson-build-v3-1-8da6976fb0a2@oss.qualcomm.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Previously, tp_new and tp_init functions were defined for Request with
the tp_init function raising a NotImplementedError to prevent the object
from being initialized (__init__, the stage after __new__).
Similar results can be achieved by removing the tp_new and tp_init
functions which raise a slightly different error:
TypeError: cannot create 'gpiod._ext.Request' instances
As this is an internal only detail, there should be impact to users.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://patch.msgid.link/20260423222125.29097-4-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
The PyObject_CallMethod() function jumps through hoops to find the method
to call and to build a stack (if necessary) to call a given method.
In the case of finalizing the request object, that overhead can be
avoided by simply sharing a function with `request_release()` to close the
underlying gpiod_line_request object.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://patch.msgid.link/20260423222125.29097-3-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
The PyObject_CallMethod() function jumps through hoops to find the method
to call and to build a stack (if necessary) to call a given method.
In the case of finalizing the chip object, that overhead can be avoided
by simply sharing a function with `chip_close()` to close the underlying
gpiod_chip object.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://patch.msgid.link/20260423222125.29097-2-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
The C extension methods request_set_values() and request_get_values()
iterate over a caller-supplied sequence and write offsets/values into
pre-allocated buffers self->offsets and self->values, without checking
the write index.
If the user tries to read values for more lines than they previously
requested, we overflow the buffers.
Add appropriate bounds checks and implement test cases for this issue.
Fixes: b7ba732e6a ("treewide: libgpiod v2 implementation")
Reported-by: 4fqr <4fqr@proton.me>
Tested-by: Vincent Fazio <vfazio@xes-inc.com>
Reviewed-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://patch.msgid.link/20260407-treewide-fixes-v1-1-66c9744a56a3@oss.qualcomm.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
For lines with no name, chip_get_line_name() returns Py_None but does
not increment its reference count. Do it enough times and the following
error will occur with cpython before v3.12:
Fatal Python error: none_dealloc: deallocating None: bug likely caused by a refcount error in a C extension
Add the missing Py_INCREF(). It's safe to do even on python >= v3.12 as
Py_INCREF() has no effect on immortal objects in more recent versions.
Fixes: 0c0993569c ("bindings: python: ext: add ability to query line name")
Reported-by: Pierre Gramme <pgramme@gmail.com>
Closes: https://github.com/brgl/libgpiod/issues/166
Link: https://lore.kernel.org/r/20260113092852.42573-1-bartosz.golaszewski@oss.qualcomm.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Changelog:
- make request and reconfigure behavior for lines requested by names
consistent with that of lines requested by offsets
- fix parsing non-tuple keys when requesting a line
- disable UP045 linter check
- use quoted types when casting
- generate wheels for CPython 3.14
- issue a warning when reconfiguring a line not in the original request
- update dependencies for wheel generation
- update licensing information
- coding style fixes
- minor performance improvements
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Previously, lines requested by name could also be referred to by offset
when calling LineRequest.reconfigure_lines. However, lines requested by
offset could not be reconfigured by name if they had one.
Now, the name map is populated if the line has a name regardless of if
it's requested by name or offset.
To make behavior as consistent with earlier versions as possible, a line
requested by name always returns the first line on the chip with that
name. If there are duplicate line names and lines are requested by
offset, the first line in the request is mapped to the name. A line
requested by name always overrides a name mapped via an offset request.
If lines 0-3 all have the same name and a config object looks like:
{("foo", 1, 2): None}
or
{(1, "foo", 2): None}
Then lines 0-2 are requested and line 0 can be reconfigured by name.
If the config object looks like:
{(1, 2): None}
Then lines 1 and 2 are requested and line 1 can be reconfigured by name.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20251009130516.3729433-9-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Previously, calling `reconfigure_lines` with line identifiers that were
not included in the original request produced different behaviors
depending on the identifier type.
If the identifier was an integer, it would be silently ignored.
If the identifier was a string, it would raise a `ValueError`.
To be consistent, invalid line names are now silently ignored.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20251009130516.3729433-7-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Previously, a check was performed to prevent requests that included
duplicative lines.
While this provided quick feedback on erroneous requests, it penalized
callers that had a good config due to the over head of iterating the
config multiple times.
Now, duplicate check is performed as the config object is iterated.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20251009130516.3729433-5-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Previously, an offset map was maintained to lookup the name used to
request a line so that a list could be generated that reflected the
identifier used to request a given line.
For example, if a config looked like {0: None, ("foo", "bar"): None}, the
LineRequest.lines property would contain [0, "foo", "bar"].
Now, the line identifier is tracked in request order as the config
object is iterated to avoid maintaining an offset map.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20251009130516.3729433-4-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Previously, Chip.request_lines and LineRequest.reconfigure_lines
iterated over a config object but did so in subtly different ways.
Create a generator that iterates over a config object and returns line +
settings pairs. Update the aforementioned functions to use the generator
to consolidate and condense code.
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20251009130516.3729433-2-vfazio@gmail.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
When `chip.request_lines` was modified to allow `Iterable`s instead of
only `tuple`s, the code that checked for duplicate line entries in the
`config` argument was not updated to account for the expanded types.
If the `config` argument had a key that was not a `tuple`, `str`, or
`int`, a `TypeError` would be raised when resolving the line offset.
Refactor the code the resolves IDs to offsets into a separate function
to make the logic a bit clearer and to account for the widened types.
Fixes: 8f62e6c453 ("bindings: python: loosen type requirements in public API")
Closes: https://github.com/brgl/libgpiod/issues/148
Signed-off-by: Vincent Fazio <vfazio@gmail.com>
Link: https://lore.kernel.org/r/20250902123905.6491-1-vfazio@gmail.com
[Bartosz: don't pass the only argument by name in the call to
_resolve_config_keys_to_offsets()]
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Minor gpiod package release.
Changelog:
- add fileno() for Chip and LineRequest
- relax the verbosity adjustment in build_tests.sh
- fix potential PyDict_Next() crash
- fix tests failing after a kernel fix to gpio-sim
- various documentation fixes
- rework the code treewide to make it conformant with mypy and ruff
linter recommendations
- fix requesting lines by name with multiple entries
- fix a use-after-free bug in python bindings
- fix passing the event clock property to line requests in python
bindings
- improve the output of LineSettings.__repr__()
- import gpiod attributes in external module
- specify a tar extract filter for python >= 3.12
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Certain polling APIs in the standard library - most notably: the
select() function and the poll class - allow to poll any object that
implements the fileno() method returning the underlying file descriptor
number.
Implement fileno() for Chip and LineRequest which allows users to do:
rd, _, _ = select([chip/request], [], [], 1)
where rd will contain the actual object passed to select which makes for
easier reading of events afterwards.
Reviewed-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://lore.kernel.org/r/20241211-python-fileno-v2-1-84c642572485@linaro.org
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
The module is now called _internal.py so update the Makefile. Without
this, we're no longer able to generate the distro tarball.
Fixes: b306ce694a ("bindings: python: make internal a private submodule")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Since their inclusion in Python 3.6, f-strings have become the preferred
way to format strings with variable values as they are generally more
readable as the value substitution is in place and doesn't have to be
parsed from the list or arguments to `.format()`.
Where it does not impact readability (when the line is <120 characters),
swap usage of `.format()` to an f-string.
For lines that are not converted, inform the linter to ignore attempts
to upgrade those instances to f-strings [0]
[0]: https://docs.astral.sh/ruff/rules/f-string/
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://lore.kernel.org/r/20241114145116.2123714-15-vfazio@xes-inc.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
The `values` argument of `_ext.Request.get_values` uses a preallocated
`list[None]` as a buffer that is populated with `Value`s by the external
module that are then returned from the function.
Use `cast` to inform the type checker it's a `list[Value]` despite how
it's allocated.
Also, as `lines` is typed as an `Iterable`, there is no guarantee it has
a `__len__` method. Instead, use the size of the `offsets` array to
allocate the buffer.
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://lore.kernel.org/r/20241114145116.2123714-14-vfazio@xes-inc.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Since `LineRequest._req` is typed to be `Optional` and because type
checkers may not be able to infer that an object is not `None` from an
earlier call (such as `_check_released`) it is necessary to inform type
checkers of the state of the object to silence union-attr [0] errors.
Instead of littering the code with "# type: ignore" comments, use casts
to inform type checkers that objects are not `None`.
Using `assert` is another option, however this duplicates the logic in
`_check_released` so is redundant at best and, at worst, is not a safe
replacement as `assert` can be elided in optimized Python environments
and these checks need to be runtime enforced.
[0]: https://mypy.readthedocs.io/en/stable/error_code_list.html#check-that-attribute-exists-in-each-union-item-union-attr
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://lore.kernel.org/r/20241114145116.2123714-12-vfazio@xes-inc.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Since `Chip._chip` is typed to be `Optional` and because type checkers
may not be able to infer that an object is not `None` from an earlier
call (such as `_check_closed`) it is necessary to inform type checkers
of the state of the object to silence union-attr [0] errors.
Instead of littering the code with "# type: ignore" comments, use casts
to inform type checkers that objects are not `None`.
Using `assert` is another option, however this duplicates the logic in
`_check_closed` so is redundant at best and, at worst, is not a safe
replacement as `assert` can be elided in optimized Python environments
and these checks need to be runtime enforced.
[0]: https://mypy.readthedocs.io/en/stable/error_code_list.html#check-that-attribute-exists-in-each-union-item-union-attr
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Link: https://lore.kernel.org/r/20241114145116.2123714-10-vfazio@xes-inc.com
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
