The existing code takes 750ms to start per resolver. This is too long,
especially in mrccomm where we start 4 on initialize, and 3 any time the
team number changes.
Solve this by handling the async setup more correctly. Assuming it
returns pending, we're assuming the announcer is good. Then on stop(),
we cancel the register in case its still pending, wait for the callback,
and then deregister.
Easier then the last one that put everything in a sub namespace. By
prefixing the name less things break, and intellisense will be less
confusing to new users during the transition.
Fixes local clients receiving inconsistent data updates. Data was only
flushed to local clients on incoming websocket data (or when explicitly
flushed), so local clients on a quiet server would stop receiving
updates.
Revert some changes from #7997.
Instead of relying on periodic sends, always immediately send local data
updates to the wire for lower latency. This means that on local clients,
the periodic update rate and sendAll settings will be ignored, as all
data updates will be sent immediately.
Used Gemini to convert the C++ tests in wpimath/.../kinematics to
pytest. For ease of use required a better constructor for
`MecanumDriveWheelPositions`
I did this conversion months ago (before robotpy landed and before the
reorg), so new tests might be missing. At least its better than nothing
🤷
Broke the huge pr that does almost everything into parts so its easier
to review.
---------
Co-authored-by: David Vo <auscompgeek@users.noreply.github.com>
This anecdotally fixes a problem that occurs when building with bzlmod.
At least for me this fixes the problem, and now that there is a function
for it it can be easier to deal with any other special casing that needs
to happen (windows, remote execution, etc). It is still a bit naive, but
should fix the major problem at the moment.
Semiwrap / meson / robotpy define `NDEBUG` when building their software
in all modes, while `allwplib` only does it when building debug. This
causes the size of `DenseMap` to differ between the shared libraries
built here, and the extension modules built in `mostrobotpy`, causing
segfaults when you try to execute code that uses `DenseMap`. This is not
a problem with the robotpy code in `allwpilib`, because bazel uses the
exact same compiler flags when building the shared libraries and
pybind11 extensions.
wpical was unable to use wpimath and its dependent libraries because
Ceres was compiled with a different version of Eigen. Now that the Ceres
build has been redone and shipped in #8151, we can now use wpimath and
our C++ apriltag wrapper in wpical, allowing for major refactors. This
includes:
* Using `to_json` and `from_json` specializations to concisely serialize
and deserialize all JSON files instead of manually handling JSON.
* Removal of the `Fieldmap` and `Pose` classes, which were duplicates of
the `AprilTagFieldLayout` and `Pose3d` classes respectively.
* Using `AprilTagDetector` instead of the raw libapriltag library.
* Using `Pose3d` instead of raw Eigen matrices.
In addition, several other refactors were made to make the code more
readable and to fix several UX issues and crashes. This includes:
* Eagerly parsing every JSON file when selected by the user. This means
JSON files are only parsed once on selection, instead of every time a
downstream function wants to use the data. This also means invalid JSON
can be detected upfront and a specific error shown immediately instead
of a catch all error when trying to calibrate.
* Using `std::optional` to indicate a calibration failed instead of
status codes.
* Processing videos on separate threads to not block the UI thread and
take advantage of parallelization for camera calibration. (2x speedup on
my laptop)
* Removing the OpenCV calibration option, since mrcal should be better
in every scenario.
* Showing a progress bar for camera calibration.
* Breaking up the massive `DisplayGui` function into separate functions
which contain code for different popups. This also allowed for better
organization and scoping of static variables.
* Renaming variables to make their purpose more clear.
* Displaying the tags present in a field layout when trying to combine
multiple field layouts.
Fixes#7722.
Resync with `mostrobotpy`
This mostly involves the big "ignore almost everything in the HAL
project" and some fixups for the Addressable LED classes.
Required two small hand fixes to get it building over here with bazel,
and with more compiler warnings on.
I also manually zeroed out the `repo_url` field in the toml files to
avoid unnecessary churn whenever it goes from a release build to a
development build. I already did this with `version` field in there, and
will do a follow up PR that updates the copybara script to do it
automatically.
---------
Co-authored-by: Default email <default@default.com>
This adds a type caster for `WPI_String` so that pybind11 can more
easily auto-convert between that and strings. This helps remove the need
to do things like
[this](f1d77244c3)
in the opmodes fixup
This is a high churn change whenever `mostrobotpy` switches from
pointing to `release-2027` to `development-2027`. The value isn't used
by the bazel build scripts, so there is no reason to have a zero impact
change on ~11 files muddying up the synchronization diffs.
In the `allwpilib -> mostrobotpy` direction, we update [this
](https://github.com/robotpy/mostrobotpy/blob/main/rdev.toml) config
file and run
[this](https://github.com/robotpy/mostrobotpy/blob/main/rdev.sh) script
which updates it in that repo during.
I manually did this change in #8503, but this will do it automatically
in the future
I tried to sync `mostrobotpy` with `allwpilib` and was getting a
compilation error I had not seen before when it tried to do the stub
generation (which `allwpilib` does not do).
Luckily, I was able to debug it here by writing some unit tests (i.e.,
having Gemini convert the C++ tests into python) that failed in a
similar way. The main problem was needing to write a custom constructor
for the class and adding a `force_type_casters`. I used `ChassisSpeed`
as my main example, but I did not copy all of the other custom code like
overriding the index operator, `__repr__` operator, feet helpers, etc. I
can gladly add those in.
In the future, we should start enforce a policy that if you add a C++ or
Java unit test, you also have to add a python test. That developer might
have gotten more stuck on the minutia of how to fix it, but this problem
would have at least been caught earlier before it landed.
This makes error messages point directly at the variable use, instead of
on the enclosing AST node:
```
error: `outerCoroutine` may not be in scope
outerCoroutine.yield()
^
error: `outerCoroutine` may not be in scope
consume(x, outerCoroutine);
^
```
instead of
```
error: `outerCoroutine` may not be in scope
outerCoroutine.yield()
^
error: `outerCoroutine` may not be in scope
consume(x, outerCoroutine);
^
```
Sequential group builder had an inverted if condition causing NPEs
Parallel group builder was building parallel groups using an old ctor
signature and creating groups that didn't match what was specified in
the builder
Test coverage has been added for both builder types
Supports Doxygen 1.14, and should be better with Doxygen 1.15 that we
use, compared to the current version.
Improves, but does not fix#8486. The sidebar now covers the search
results, rather then both bleeding through
<img width="1206" height="1969" alt="image"
src="https://github.com/user-attachments/assets/e8c0cefd-a72a-4c41-a5bf-c191752250f4"
/>
We need to wait, or otherwise OpModeRobot will immediately reinstantiate
and re-run the opmode, which is generally undesirable (e.g. for
autonomous).
Fixes#8475.
This got missed in the reorg, and these values aren't actually used for
anything when building in `allwpilb`, but we might as well fix them here
to make the copybara process easier.
