Raise WPILib alerts for Phoenix device and CAN bus health

[nlaverdure]

Summary

Phoenix 6 sends its automatic alerts straight to MrcLib, so they never reach WPILib's alert registry and AdvantageKit doesn't log them (#47). This PR covers the conditions that affect this robot with WPILib Alerts, which AdvantageKit logs under /RealOutputs/Alerts. It also adds the CAN bus fields from #50.

Every alert is computed from logged inputs, so replaying a log shows the same alerts.

What changed, and why it differs from #52's written tasks

#52 task	This PR	Why
Config failure	HIGH alert per device, for drive setup, turn config, and CANcoder read and write	As written. A failed CANcoder read also skips the write and doesn't save a turn zero to Preferences, because the config then holds defaults.
CANivore disconnect	Bus Status and CAN states on the SystemCore ports: HIGH for ErrorPassive, BusOff, Stopped, a failed status read, a rising bus-off or restart count, or a stalled reader. MEDIUM for ErrorWarning.	The robot has no CANivore. REC/TEC > 0 was rejected as a trigger: 6328's committed alert with that trigger can only fire in the first 0.5 s (Robot.java@3ea1eb0:195,255-257), so there's no field evidence it stays quiet on a healthy bus.
Firmware mismatch	HIGH alert when setControl returns FirmwareTooOld or ApiTooOld. Versions are logged for reference.	Phoenix blocks output only on a compliancy mismatch, and setControl returns that verdict (ParentDevice.setControlPrivate). Pinning versions wouldn't catch it.
Disk space	No alert. Phoenix's .hoot logging is behind FeatureFlags.HOOT_LOGGING_ENABLED, off by default	The .hoot files are rarely used. 6328 and Spectrum both turn off auto-logging. The pre-match display already warns when the USB drive is low.

Also added:

• A CANcoder-disconnected alert.
• Drive/ConstructMs, which measures boot cost. The upper bound with SC1 dead is about 18–19 s (derived from the timeouts, not measured).

#50: CAN bus fields

• LoggedCANBus now logs BusErrorCount, ArbitrationLostCount, RestartCount, State, Status and SampleCount.
• It reads CANBus.getStatus() on a background thread every 400 ms, and only in REAL mode. CTRE's Javadoc says the call "can block for up to 0.001 seconds."
• Log Phoenix CAN bus health fields that AdvantageKit SystemStats doesn't cover #50 stays open. Removing the fields that duplicate AdvantageKit's SystemStats needs the robot comparison below. Sim can't do it: Phoenix sim reports fixed values on one simulated network, and SystemStats shows the interfaces as down.

Verified

• Unit tests (pure logic, no HAL): 23 new tests in CANBusHealthTest, PhoenixFirmwareTest and PhoenixUtilTest, all passing. ./gradlew build runs 211 tests. One fails, VisionFilterTest > yawConsistency > Rejects ambiguous PnP solution with wrong yaw, and it fails the same way on main-2027-alpha7.
• Sim, with ModuleIOSimTalonFX and injected failures (temporary, not committed):
  • Hoot flag: on, Phoenix wrote a 446 KB .hoot file; off, it wrote none. Both runs had 12 Phoenix devices.
  • Config failure: DriveInitStatus = ConfigFailed raised "Drive motor setup failed on module FrontLeft (ConfigFailed)."
  • CANcoder read failure: the alert appeared and stayed on. That module's turn-zero Preference was not saved again. A module with a good read had its deleted key saved in the same run, as a positive control.
  • Firmware decode: 0x1A460000 → 26.70.0.0, matching getVersionMajor/Minor/Bugfix/Build() on sim devices.
  • Default sim: no alerts and no .hoot files.

Robot session (can't be done in sim)

• /RealOutputs/LoggedRobot/UserCodeMS p50/p99 compared with a pre-change log.
• CANBus/SC*/SampleCount rises about 2.5 per second.
• Log Phoenix CAN bus health fields that AdvantageKit SystemStats doesn't cover #50: compare CANBus/SC<n>/BusUtilization (0–1) with SystemStats/Network/CAN<n>/Utilization (percent), and BusErrorCount with RX/Errors and TX/Errors, to check bus numbering and overlap.
• Record Drive/ConstructMs with a healthy bus and with SC1 unplugged at boot.
• Unplug one CANcoder at boot: the config and disconnected alerts appear, and no zero is saved.
• Unplug SC1 at the SystemCore: State reaches ErrorPassive or BusOff, and the HIGH CAN alert appears within about 1 s.
• Firmware strings match Tuner X, and there's no *FirmwareBlocked alert.
• A new /U/logs/session_N contains only .wpilog files.
• SC0 (no Phoenix devices) reports OK/ErrorActive. Record State/TEC/REC with the PD and gyro unplugged.
• Count CAN alert activations over a full healthy session, including boot.

Closes #52
Refs #50

🤖 Generated with Claude Code

[nlaverdure]

Superseded by PR 57 (#57), which merges this PR's device and CAN bus alerts with the CAN chain hint from PR 55, simplified per the 2026-09-27 review.