Hint where a CAN daisy chain is broken, and report PD connection

[nlaverdure]

Summary

When a CAN cable or connector fails, every device past the break drops at once, and the Driver Station fills with disconnect alerts that don't say where to look. This PR gives each device a CAN index, its position along the daisy chain from the SystemCore. When exactly one break explains the pattern, a single HIGH alert names the spot, for example:

CAN chain break on SC1 (order traced 2026-10-03 A.Student): #0–#4 respond, #5–#11 don't. Check #4 BackLeft turn (ID 11)'s outgoing connector, the cable, and #5 BackRight turn (ID 19)'s incoming connector.

It also gives the REV PDH a connection signal and a PD/disconnected alert.

How it works

• Topology in Constants. Each bus declares a CANChain, which names the bus and its port, then one line per device:

  public static final CANChain CHAIN = new CANChain("SC1", CANPort.CAN_S1);
  public static final CANChain.Device BACK_LEFT_DRIVE = CHAIN.add(10, "BackLeft drive");
  ...

  • Java runs static initializers in the order written (JLS §12.4.2), so line order = wiring order, and a device's position is its CAN index. Each device is declared once, so none can be left out of the chain.
  • The chain freezes when the monitor first reads it, and an add after that throws.
  • Each constant is a CANChain.Device carrying its bus, ID and label, so the bus is stated once: by which bus class the device is declared in. Phoenix's SwerveModuleConstants hold only IDs, so DriveConstants builds each module from a ModuleDevices record and returns it from DriveConstants.devices(module). ModuleIOTalonFXBase, GyroIOBoron and LoggedPowerDistribution take their bus from their devices.
  • CHAIN_ORDER_TRACED (date and name) turns on a bus's hint. Both buses start untraced, because the current order is the old constant order, not the wiring. The CANBusPorts Javadoc explains how to trace a bus.
• Connection sources. Each subsystem reports every CAN device it owns, on any bus, in one canConnections() map keyed by CANChain.Device (which carries the bus, since a CAN ID is only unique within a bus): Drive for the module devices and the gyro, LoggedPowerDistribution for the PD. Each bus's LoggedCANBus (built from its chain, new LoggedCANBus(SC1.CHAIN, SC1.CHAIN_ORDER_TRACED)) owns its monitor. Its monitorChain(...) takes every subsystem's map and keeps the entries on its own port. A device reported twice turns that bus's check off with one DS warning.
• Break detection (CANChain, static and pure): a clean split with devices 0..k-1 up and k..n-1 down, with at least 2 down. With one device down, a failed device and its last cable look the same, and that device's own alert already names it.
• Hold time (in CANChainMonitor): the break must hold for 2.5 s. Redux's isConnected() waits 2 s while the Phoenix flags drop after 0.5 s, so a shorter hold could briefly name a link one step too far along the chain.
• Monitor (CANChainMonitor, owned by each bus's LoggedCANBus, which updates it at the end of log()):
  • Logs the raw index every cycle (CANBus/<bus>/ChainBreakIndex), so flickers show up in post-match review.
  • Turns off with one DS warning if the chain and its connection sources don't match one to one.
  • Reads only logged inputs, so replay reproduces the alert.
• PD connection: WPILib has no connected check. On SystemCore, though, a PDH read with no status frame in the last 40 ms times out, and the HAL then returns exactly 0.0 V (REVPDH.cpp:465-467, allwpilib v2027.0.0-alpha-7). A responding PDH can't read 0 V, because it powers the SystemCore running the code. So voltage > 0 (debounced 0.5 s) is logged as the PD's connected input, and it:
  • raises PD/disconnected;
  • skips the other PD reads while the PDH is missing, since each one sends a DS error.

Verified

• Unit tests: 30 new tests (CANChainTest, CANChainMonitorTest, DriveConstantsTest) cover:

  • every split case;
  • the hint text;
  • validation (missing, unknown and duplicate IDs);
  • chain order and freezing;
  • sorting connection sources by bus, and rejecting a device reported twice;
  • each module's IDs coming from its devices, equal to main's values, all on CAN_S1;
  • the hold time;
  • the slow-device skew scenario.

  ./gradlew build runs 218 tests. The only failure is VisionFilterTest > yawConsistency, which fails the same way on main-2027-alpha7.

• ID refactor: a script compared every device ID with the int constants on main-2027-alpha7 (all 14 match, none missing), and DriveConstantsTest pins each module's drive, turn and encoder IDs. A sim run with Phoenix sim devices built all 12 module devices from the device lookup.

• Sim, with an injected break at SC1 index 5 (temporary, not committed):

  • The raw index was 5 from the first loop.
  • The alert named BackLeft turn (ID 11) at index 4 and BackRight turn (ID 19) at index 5 2.501 s later, listed above the per-module disconnect alerts.
  • SC0 (only the sim gyro down) stayed silent.

• Sim, committed code: no chain entries and no alerts. /PD/Connected is true, and the PD outputs are unchanged.

Robot session (disabled, on blocks)

• Power off. Measure CAN-H to CAN-L at each bus's SystemCore end and expect about 60 Ω, which confirms termination at both ends.
• Trace SC0 and SC1, reorder each bus's device lines to match, and set CHAIN_ORDER_TRACED.
• Unplug one mid-chain SC1 connector: the devices before it stay connected, and the hint names the right link. This tests the premise that the SystemCore side of a break keeps working.
• Unplug at the SystemCore: the "no device responds" hint appears.
• Unplug the PDH's CAN: /PD/Connected goes false within about 0.6 s, PD/disconnected appears, and the DS console stops flooding.
• Reseat everything: the alerts clear. Boot twice from power-off: no chain alert.
• Note /SystemStats/Network/CAN<n>/FD for both buses. CAN FD is less tolerant of a missing terminator.

Relation to #54

This PR is independent of #54. Both touch Robot.java and LoggedCANBus.java, so whichever merges second gets rebased onto main-2027-alpha7.

🤖 Generated with Claude Code

[nlaverdure]

Superseded by PR 57 (#57), which merges this PR's CAN chain hint and PD connection signal with the device and CAN bus alerts from PR 54, simplified per the 2026-09-27 review.