Glass now internally uses the AllianceStationID enum, but the two
separate topics remain for simple use cases. The Invalid alliance is now
published as Red, Station 0, instead of 3.
Fixes#6488 and fixes#8725.
To aid in simplifying the packing/unpacking of Lists to protobuf, this
adds helper methods that put the iteration logic in `Protobuf.java`,
similar to arrays. This makes packing/unpacking a List a single line,
similar to other types. Useful for #8776 and #8172.
Move pybind11 over, delete the WORKSPACE files, and fix a test which
assumed WORKSPACE support was enabled.
---------
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
The latest toolchain release has my fix for remote execution. Use it.
This makes remote execution work fully in wpilib.
---------
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
Alignment code isn't needed since C++17.
Bazel tests were segfaulting and asan traced this to use-after-frees
related to some of the atomic booleans. Worker threads are now joined to
ensure the thread is completely stopped and destroyed before the atomic
boolean is freed. The primary camera calibration thread also has had its
finished flag moved to a shared_ptr to ensure it lives long enough.
#8787 Added a new compound unit that was tripping up the pybind11
stubgen when trying to mirror over to mostrobotpy land. This fixes the
error, and also cleans up some `create-imports` things.
```
[680/680] Generating semiwrap/modules/_wpimath.pyi with a custom command
FAILED: [code=1] semiwrap/modules/_wpimath.pyi
/usr/local/bin/python -m semiwrap.cmd.make_pyi wpimath._wpimath wpimath/_wpimath.pyi semiwrap/modules/_wpimath.pyi -- wpimath /tmp/build-via-sdist-2hfn0_cv/robotpy_wpimath-2027.0.0a6.post1/wpimath/__init__.py wpimath._init__wpimath semiwrap/_init__wpimath.py wpimath._wpimath semiwrap/modules/_wpimath.cpython-311-x86_64-linux-gnu.so
pybind11_stubgen - [ ERROR] In wpimath._wpimath.AntiTipping.__init__ : Invalid expression 'wpi::units::unit_t<wpi::units::unit<std::ratio<1l'
pybind11_stubgen - [ ERROR] In wpimath._wpimath.AntiTipping.getP : Invalid expression 'wpi::units::unit_t<wpi::units::unit<std::ratio<1l, 1l>, wpi::units::base_unit<std::ratio<1l, 1l>, std::ratio<0l, 1l>, std::ratio<-1l, 1l>, std::ratio<-1l, 1l>, std::ratio<0l, 1l>, std::ratio<0l, 1l>, std::ratio<0l, 1l>, std::ratio<0l, 1l>, std::ratio<0l, 1l> >, std::ratio<0l, 1l>, std::ratio<0l, 1l> >, double, wpi::units::linear_scale>'
pybind11_stubgen - [ ERROR] In wpimath._wpimath.AntiTipping.setP : Invalid expression 'wpi::units::unit_t<wpi::units::unit<std::ratio<1l'
Warning: stubgen - [WARNING] Raw C++ types/values were found in signatures extracted from docstrings.
Please check the corresponding sections of pybind11 documentation to avoid common mistakes in binding code:
- https://pybind11.readthedocs.io/en/latest/advanced/misc.html#avoiding-cpp-types-in-docstrings
- https://pybind11.readthedocs.io/en/latest/advanced/functions.html#default-arguments-revisited
pybind11_stubgen - [ INFO] Terminating due to previous errors
ninja: build stopped: subcommand failed.
```
These were caught by building on engflow. This fixes a
couple of places where paths were escaping the sandbox and resolving
back to the git repo.
---------
Signed-off-by: Austin Schuh <austin.linux@gmail.com>
This piggy-backs off #8362. `wpiformat` will handle all the non-python
files (C++ trailing spaces in yaml files, etc), and the robotpy files
(guarded by being under src/main/python or src/test/python) will be
handled by black only.
This updates to the 2027-1 toolchains. This also is the first version
with the `rules_bzlmodrio_toolchains -> wpilib_toolchains` rename, so
the surface area of the change is a little bit big.
The opencv dep has bzlmod'ified as part of this as well.
This adds a C++, Java and Python version of a Biquad filter class to
wpimath.
For testing, I took output from scipi functions and commented the inputs
I used above each test case.
---------
Co-authored-by: Drew Williams <drew.williams@capstanmedical.com>
- Fix a dependency on other test global state via Vin by calling
ResetData() to ensure consistent state
- Clamp the ground truth voltages to match sim.SetInputs()
- The C++ test was checking the noisy measurement output instead of
using sim.GetState()
- Make GetState() public in C++ and Java
The `command3` example project contains a program that could plausibly
play in the 2026 rebuilt game. It includes nested mechanisms (`Intake`
has an inner `IntakeWrist` and `IntakeRoller` and is similar to the v2
superstructure concept), swerve drive with localization and path
following (albeit stubbed for sake of example), opmodes and
opmode-scoped commands, and command-scoped triggers.
The template projects are basic skeletons. The larger template includes
a basic command that just increments and prints a counter variable every
time it runs.
The hatchbot v3 example has been refactored to be more idiomatic:
- `RobotContainer` removed
- "Subsystem" names in packages, comments, and classes has been replaced
with "Mechanism"
- Some v2-specific comments and structures have been reworded or deleted
- The Drive mechanism now provides commands for arcade drive and driving
a distance, instead of exposing public methods that write directly to
hardware (which broke encapsulation and made it possible to issue
conflicting hardware requests)
This turns the styleguide on for the non-python robotpy files.
The overwhelming amount of changes were
related to whitespace, followed by some IWYU for standard library
headers.
I tasked Claude with converting the existing C++ tests to python for
wpilib. I gave it a decent review for comparison to the existing tests,
and it seems to have covered everything. I did do some small cleanup in
a couple of places. One notable test that is missing is the LED
patterns, but that is getting handled in
[mostrobotpy](https://github.com/robotpy/mostrobotpy/pull/254) land.
Resolves#8587
This PR implements the requested anti-tipping utility and refactors the
math to correctly adhere to the NWU coordinate system.
**Key Changes:**
* Fixed the axis mapping: Positive pitch now correctly maps to a forward
tip (+X), and positive roll maps to a rightward tip (-Y).
* Inverted the proportional control logic: The correction vector now
applies a positive `kP` to drive *into* the direction of the fall to get
the wheels back under the center of gravity, rather than driving away
from it.
* Added a comprehensive JUnit test suite (`AntiTippingTest.java`) to
verify the calculated `ChassisSpeeds` correctly zero out the orthogonal
axis and provide the correct positive/negative velocity across all four
tipping directions.
Tested locally against `testDesktopJava` and passes all style/formatting
guidelines.
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
