The current DS thread model has some pretty major issues. It makes it difficult to know if all data is from the same remote packet, and if the data changes while the robot loop is running. Additionally, the DS thread is used for a few other things (MotorSafety and State Tracking for EducationalRobot). This also makes sim difficult, as user code has to wait for the thread to know it has new data.
This change completely rethinks how threading works in the driver station model.
First, the DS HAL system receives a new data callback, either from Netcomm or DriverStationSim. Inside the context of this callback, all the low latency data is read and put into a cache. Doing some investigation on the robot side, this is perfectly safe to do, and also ensures a ds packet will not be parsed before we finish reading the current packet data.
After all data is read, the cache is swapped with a 2nd buffer. This buffer just stores the data, none of the HAL DS calls read from this buffer. An event is then fired, stating there is new data ready to go.
Robot code calls HAL_UpdateDSData(). This swaps the 2nd buffer with a 3rd buffer, which always contains the current data. This data will not be updated until HAL_UpdateDSData is called again. Which solves the state problem.
The high level driver station classes have. an updateData() call, which calls HAL_UpdateDSData, and then update button state variables, then data log and update the NT FMS data table (Java also caches across the JNI boundary here, but that could trivially be removed). An extra event provider is provided, allowing other threads to know when this call has been completed.
IterativeRobotBase calls DS.updateData() at the beginning of each loop, and only once per loop. This means all commands will always have the same state.
All of this means there is no longer a DS thread. Everything happens synchronously. This means Sim and testing is easier, as you can just call DriverStationSim.NotifyNewData(), and then DriverStation.UpdateData(), and you can guarantee that all the DriverStation.*** data is up to date.
As for Motor Safety and Educational Robot State Handling, those can all be handled by their own threads. The Educational Thread only needs to run under EducationalRobot, and MotorSafety will only be started if there is a motor safety object enabled.
In some cases, knowing roborio 2 might be useful. This also creates a higher level enum that might be usable later for the discussion on more complex runtime types.
- Twine, StringRef, Format, and NativeFormatting have been removed
- Logging now uses fmtlib style formatting
- Nearly all uses of wpi::outs/errs have been replaced with fmt::print() or
std::puts()/std::fputs() (for unformatted strings).
- A wpi/fmt/raw_ostream.h header has been added to enable
fmt::print() with wpi::raw_ostream
This makes code easier to read and more consistent between C++ and Java.
Also update clang-format settings to always add a line break (even if no braces are used).
Also move some things in HAL for consistency.
WAS:
C++:
- C APIs: #include "mockdata/AccelerometerData.h"
- User side class: #include "simulation/AccelerometerSim.h"
Java:
- JNI APIs: hal.sim.mockdata.AccelerometerData (and a few classes in hal.sim)
- User side classes: hal.sim.AccelerometerSim
IS:
C++:
- C APIs: #include "hal/simulation/AccelerometerData.h"
- C++ class: #include "frc/simulation/AccelerometerSim.h"
Java:
- JNI APIs: hal.simulation.AccelerometerData
- User side class: wpilibj.simulation.AccelerometerSim
This allows high-level library classes to implement enhanced simulation
support even if no low-level corresponding simulation library exists, and
avoids the need for bit-banging complex interfaces like SPI or CAN.
Default behavior is still to run the robot main loop in the main thread.
The ability to run the robot main loop in a separate thread and add a hook
for running a different function in the main thread is needed for simulation
GUI support on some platforms.
std::scoped_lock was introduced in C++17 and is strictly better than
std::lock_guard as it supports locking any number of mutexes safely.
It's also easier to use than std::lock for locking multiple mutexes at
once.
This makes callback registration completely thread safe.
This patch also uses templates and macros to dramatically reduce the amount of
manual boilerplate.
The old headers were moved into folders because doing so avoids polluting
the system include directories.
Folder names were also normalized to lowercase.
During shared library loading, a different libLLVM can be pulled in, causing
llvm symbols from dependent libraries to resolve to that library instead of
this one. This has been seen in the wild with the Mesa OpenGL implementation
in JavaFX applications (see wpilibsuite/shuffleboard#361).
This is clearly a very breaking change. For some level of backwards
compatibility, a namespace alias from llvm to wpi is performed in the "llvm"
headers. Unfortunately, forward declarations of llvm classes will still break,
but compilers seem to generate clear error messages in those cases
("namespace alias 'llvm' not allowed here, assuming 'wpi'").
This change also moves all the wpiutil headers to a single "wpi" subdirectory
from the previously split "llvm", "support", "tcpsockets", and "udpsockets".
Shim headers will be added for backwards compatibility in a later commit.
