TL;DR: kill https://github.com/wpilibsuite/bzlmodRio-libssh, update
versions, and make things easier for a script to update.
The main motivation behind this was that libssh was out of date with the
gradle version. Using these "external" dependencies can cause agita as
we churn through toolchain updates. Unlike the toolchains and opencv,
which can be used by external users (i.e. a all bazel robot, if a vendor
wanted to use bazel) a C++ wrapper around libssh maven deps almost
certainly wouldn't be pulled in, much like ceres and mrclib.
As a result of vendor'ing libssh, and knowing some potential pitfalls
and annoyances, this PR does the following:
- Vendor libssh, akin to how mrclib is pulled in
- Moves the vendor'd ceres to `shared/bazel/thirdparty`, and refactors
it to better match mrclib. To me it doesn't necessarily belong in its
`thirdparty` location because it is bazel only.
- Due to the refactoring, libssh and ceres can now be pulled into
`MODULE.bazel` instead of the `WORKSPACE.bzlmod` helper. This is good
prep for a potential upgrade to killing `--enable_workspace` / bazel 9
- Write a python script that can deal with the integrity / sha256 of the
http_archives. I suggested this be added to davo's mrclib PR, but this
one is a little bit more robust and will actually edit the files in
place. This makes upgrading versions substantially easier.
- Upgrade libssh version to what gradle is using, and mrclib to the
version in #8858. These changes have been tested against that PR.
This helps reduce lock contention on Synchronization Signal objects if
the user does not intend to use this signaling path.
Use this option on the RobotBase MultiSubscriber to all topics.
There were a few issues with the generated classes.
Python had a __getattr__ forwarder that we defintely did not want.
C++ was using a struct with constexpr values, and not an enum class for
button types.
In all 3, the forwarders from gamepad didn't exist on the generated
types.
This fixes all 3.
---------
Co-authored-by: David Vo <auscompgeek@users.noreply.github.com>
The `RobotState::GetRobotMode()` API returns this enum class, but
`DriverStationSim` was using the `HAL_RobotMode` enum instead. This
commonizes the two APIs to return the same `RobotMode` enum class.
This difference in the APIs also affected Python, as the `hal.RobotMode`
and `hal._wpiHal._RobotMode` types are not compatible with each other.
Store DriverStation-owned GenericHID and Gamepad instances in Java and
C++, and expose the cached objects to Python bindings.
Move hand-written command gamepad and joystick wrappers to compose
cached CommandGenericHID instances plus typed HID wrappers, including a
Python CommandGamepad.
This will let us remove UserControls, while helping ensure that we don't
have state smashing between GenericHID objects.
Another bonus is without inheritance, intellisense will no longer show a
bunch of annoying methods, and instead just what actually exists.
---------
Co-authored-by: Peter Johnson <johnson.peter@gmail.com>
@AustinSchuh's `rules_cc` patches have all been merged upstream. This
updates `rules_cc` to include the relevant patches, removing the need to
patch it here.
This also squashes this warning on every build:
> WARNING: For repository 'rules_cc', the root module requires module
version rules_cc@0.2.13, but got rules_cc@0.2.14 in the resolved
dependency graph. Please update the version in your MODULE.bazel or set
--check_direct_dependencies=off
This fixes simulation WaitForProgramStart(true) potentially advancing
due to a stale or pre-start notifier created before SetProgramStarted()
was called.
Using MrcLib on the robot is going to be the plan for the future, to
make things easier.
MrcLib is how sim is supported going forward. The desktop version of
mrclib can act as a robot server.
This is set up where the mrclib interface is in shared code. On robot,
that is the only backend used. On desktop, a default sim backend is
used. However, the sim plugin can switch that to the real robot backend,
so the robot code will exactly look like a real robot.
Fix SetPeriod() valuePos handling when an ID moves between period queues
In SendValue(), don't append older values, and fix the total size
accounting for the underflow case when a smaller value replaces a larger
value.
Add comprehensive unit tests.
This is increasingly difficult to maintain, and has very limited
benefit. Modern coprocessors with enough horsepower to run Java
applications can use the Gradle or Bazel build systems instead.
Useful for catching bad behavior when doing math or configuring ratios
with integers. Errors can be turned off with
`@SuppressWarnings("IntegerDivision")`
```java
double kV = 12 / 6000; // error: integer division in a floating-point context
double kV = 12. / 6000; // OK
double kV = 12 / 6000.; // OK
int kV = 12 / 6000; // OK, but evaluates to 0
double ratio = 42 / 12; // error: integer division in a floating-point context
@SuppressWarnings("IntegerDivision")
double ratio = 42 / 12; // OK, evaluates to 2
```
```
Execution failed for task ':developerRobot:compileJava'.
> Compilation failed; see the compiler output below.
/Users/sam/code/wpilib/allwpilib/developerRobot/src/main/java/wpilib/robot/Robot.java:10: error: integer division in a floating-point context
double kV = 12 / 6000;
^
1 error
```
Also adds the compiler plugin as a dependency to the developerRobot
project for dev testing
The original implementation was `m_gyroOffset + (rotation -
m_pose.Rotation())`, which means the first `RotateBy` should be using
`-m_pose.Rotation()` (see `ResetPose()`). Java is already correct. This
also adds new tests to catch this particular error.
When the bazel remote cache was switched from buildbuddy to a self
hosted server in #8342, I asked that the buildbuddy hooks be remain so
that I could still use their caching service for local builds.
The downside of this was that my forks builds aren't leveraging
buildbuddy, so if I'm fiddling with something heavy like a wpimath
robotpy thing, my CI builds never update a cache and never are warm when
I push fixups.
This PR combines the `setup-bazel-remote` and `setup-build-buddy`
actions which set up the bazel remote cache. Rather than having two
different version, the correct one will be choosen in the following
order:
1. Use wpi's server with write access if the `bazel_remote` information
is set (This basically would only happen on main branch in
`wpilibsuite/allwpilib` since secrets aren't accessible from builds
originating in forks)
2. Use buildbuddy if the key it is present (This would work for my fork
builds)
3. Fall back to the readonly version of wpi's server
As seen
[here](https://github.com/bzlmodRio/allwpilib/actions/runs/25777428163/job/75712863120#step:7:28)
the build in my fork will run with buildbuddy, and my PR's build here
should fall back to readonly mode.
This also updates the bazel scripts to behave more like the C++ and Java
examples, and updates the copybara scripts to be able to sync up
`mostrobotpy`
These two functions were changed to return new objects instead of
modifying in-place, so the `const` qualifier should be added. Otherwise,
users with a const reference to a `SwerveModuleVelocity` need to make a
temporary copy first.
`added ^ removed` (XOR) is incorrect for the update guard.
Both `added` and `removed` can be true simultaneously (e.g. subscription
options change), and XOR would incorrectly skip the period update. Fixed
to `added || removed`.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
This addresses a race condition caused by TimedRobot and other
frameworks not creating their first notifier alarm until after signaling
program start.
Default this to true because it's the most common desired use case when
combined with TimedRobot.
The FTC DS had a deadband. Additionally, the FRC one had an implicit
deadband due to the only 8 bit resolution. We need a deadband by default
now with the high resolution gamepads.
