Follow-up to the etm1 fixes, driven by live debugging against a real ABB
Terra AC 7 kW behind a Waveshare RS485->TCP gateway (<ip-lan>, slave 2):
- etm3: disable TCP auto-discovery (createMethods = "user" only). Discovery
blind-scans the LAN and SIGSEGVs nymead as soon as any Modbus device answers
:502 with an exception response. The crash is in the generated reachability
handler (abbterramodbustcpconnection.cpp:2062 dereferences the check-reachable
reply after it is freed); it could not be fixed from the discovery code alone.
A gateway's slave id can't be discovered anyway, so manual add is required.
- etm2: still defer/guard AbbTerraTcpDiscovery::cleanupConnection() teardown
(kept for correctness even though discovery is now off by default).
- etm4: setupTcpThing() never called connectDevice(), so manual setup hung until
timeout. Start the connection when the network monitor reports the device.
- etm5: decode the serial-number and firmware register fields as BCD per ABB
spec §5.1 (0x46 = week 46, not 70; 0x22 = 22 kW). The binary reading rejected
real chargers ("does not match the ABB Terra AC Modbus register map").
Verified end to end: charger adds cleanly (TACW7-4-4624-S0034), states poll and
actions execute, no SIGSEGV. The underlying discovery/generated-code crash is
tracked separately.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two coupled fixes so the ABB Terra AC can be added over Modbus TCP:
- Restore checkReachableRegister ("serialNumber") in the register JSON.
The libnymea-modbus 1.15 code generator *requires* it — without it the
connection classes (abbterramodbustcpconnection.h) are never generated and
the plugin fails to build. Reverts the false lead in 4fa06df. The register
is valid: mbpoll confirms holding reg 16384 (uint64) responds at slave 2.
- Fix a use-after-free in AbbTerraTcpDiscovery::cleanupConnection() that
SIGSEGV'd nymead during network discovery whenever any Modbus device on the
LAN answered on :502 with an exception (e.g. "illegal data address"). The
method is invoked from several of the connection's own signals; it now
bails out on the second (re-entrant) call and disconnects the connection's
signals before disconnectDevice(), so the teardown can no longer re-enter
while the QModbusReply "finished" handler is still dereferencing the reply.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The status block (12 registers × 2 words = 24 words) was read as a
single Modbus request, producing a 53-byte response that overflows the
RT38B01 433 MHz radio module's 31-byte frame limit.
Dissolve the status block into individual registers with
readSchedule:update so queuedRequests sends them one at a time
(max 9 bytes per response). The deviceInfo init block (8 words → 21
bytes) is left untouched.
Bump to 1.15.0+etm6.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
reachableChanged(true) never fires within the nymea setup timeout when
the RTU link is slow (e.g. 433 MHz radio relay). Add a fallback:
trigger initialize() immediately if master is already connected
(QTimer::singleShot 500 ms), and also on master::connectedChanged.
The existing reachableChanged path is kept for TCP-style RTU adapters.
Bump to 1.15.0+etm5.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Diagnostic tcpdump : le plugin ouvrait/fermait la socket Modbus TCP à
chaque cycle de poll, générant 6-7 paquets SYN consécutifs qui saturaient
l'ESP32 WiFi de la borne (timeouts en boucle). mbpoll avec une connexion
persistante = 0 % d'échec.
Corrections :
- connectDevice() : guard "si déjà connecté → no-op" (élimine le warning
"already in ConnectedState" et les FIN/RESET parasites).
- onConnectionStateChanged(true) : appelle initialize() directement au lieu
de setReachable(true) — évite la ré-entrée initialize() via reachableChanged.
- initialize() : setReachable(true) uniquement au succès, après lecture
MinIntensity/MaxIntensity ; guard || m_updating ajouté.
- update() : guard || m_initializing ajouté (pas de poll concurrent avec init).
- Timer : if (!master->update()) master->connectDevice() — le socket reste
ouvert tant que la connexion TCP est active ; connectDevice() n'est appelé
que quand le TCP est réellement coupé.
- initializationFinished(thing) : master->update() immédiat après reconnect.
- reachableChanged handler : suppression de l'appel à initialize() (déplacé).
Résultat attendu tcpdump : 1 seul SYN à l'établissement, puis uniquement
des échanges Modbus (P. length 12/13) toutes les 30 s, aucun FIN/RESET.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Decode SlaveError (reg 0x0BC5) with full table (codes 0-10) into new
slaveErrorMessage state; code 04 (WiFi reconnect) also surfaces in
statusMessage for immediate SAV visibility.
- Add L1/L2/L3 phase powers (regs 0x0BD9-0x0BDB, confirmed hw) as
diagnostic states polled every 30 s.
- Add RegChargeTime (0x0BC6) to enum for completeness (not yet exposed).
- Collapse debian/changelog to single 1.15.0+etm1 entry (new plugin).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The Trydan ESP32 WiFi stack saturates under nymea's periodic ICMP pings
(observed: latency > 14 s, Modbus timeouts, charger drops off).
Changes:
- NetworkDeviceMonitor removed entirely from setupThing() and thingRemoved().
The charger is expected to have a DHCP-reserved IP (documented in code).
Address comes from trydanThingAddressParamTypeId param — already the correct
path after the previous setup fix.
- Connectivity now derives from Modbus poll success/failure only (reachable()
from TrydanModbusTcpMaster), NOT from network ping. A ping response does
not imply Modbus usability on this hardware.
- postSetupThing() timer (30 s): if not reachable, calls connectDevice() for
reconnect; if reachable, calls update(). The 30 s period IS the backoff —
no aggressive retry loop.
- Modbus timeout: 2 s → 5 s (tolerates residual WiFi latency spikes).
- Modbus retries per read: 1 → 0 (abort fast on first timeout; full poll
sequence already aborts at first error via doNextRead).
- k_errorLimit: 5 → 3 (3 × 5 s = 15 s before marking unreachable).
- New state "statusMessage" (QString): set on disconnect with timestamp +
cause ("timeout Modbus — vérifier signal WiFi de la borne"); cleared on
successful poll. Visible in nymea-app; helps SAV without SSH access.
- "networkdevice" removed from JSON interfaces (contract broken without monitor).
Invariants preserved: Big/Big float32 decode, no block read, PauseState+Lock
mirror, PauseDynamic conflict management, writeCompleted address filtering.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Bug: ModbusTcpMaster: device ':502' ConnectionError when adding a trydan
by manual IP (createMethod "user") or after nymead restart before the
NetworkDeviceMonitor completes its ARP scan.
Root cause: TrydanModbusTcpMaster was constructed with
monitor->networkDeviceInfo().address(), which is null at first
registerMonitor() — the monitor only populates its cache after an
asynchronous ARP/mDNS resolution. connectDevice() then ran against
QHostAddress() → ":502".
Fix:
- Read the initial IP from the stored address param (always set at
discovery or user-entry) and use it as the fallback when the monitor
has not yet resolved the address.
- Guard the networkDeviceInfoChanged callback: only call setHostAddress()
when the new address is non-null (the monitor emits transient null info
during its internal refresh cycles).
Order is now: create master with valid address → wire callbacks →
connectDevice(). The monitor update path remains for DHCP renewals.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Build fixes found during first local compilation (qmake6 + g++ -Werror):
1. Logging category: nymea-plugininfocompiler generates 'dcV2C' (uppercase C)
from the JSON vendor name "V2C". All three source files used 'dcV2c'
(lowercase c) — corrected to 'dcV2C' in integrationpluginv2c.cpp,
trydanmodbustcpmaster.cpp, v2ctcpdiscovery.cpp.
2. modbus.pri: hardcoded include(/usr/include/nymea-modbus/modbus-tool.pri)
fails when libnymea-modbus-dev is not system-installed. Changed to a
conditional: checks $$[QT_INSTALL_PREFIX] first, then NYMEA_MODBUS_PATH
override, with a graceful fallback (warning + manual C++17 CONFIG flag).
The v2c plugin has MODBUS_CONNECTIONS empty so the tool loop is a no-op
either way; the fallback is safe.
3. debian/: add powersync-plugin-v2c package entry (control, .install,
changelog entry 1.15.0+etm4).
Build verified: libnymea_integrationpluginv2c.so links against libnymea.so.1,
libnymea-modbus.so.1, Qt6Network/SerialBus/Core, 0 compiler warnings
(with -Werror -Wall -Wextra -std=c++17).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
qMin(32, m_maxIntensity) gives 0 when firmware returns 0 for MaxIntensity
(aberrant value). maxIntensity()==0 then causes every setMaxChargingCurrent
call to be clamped to 0, falling into the <6 A path which pauses charging
silently — the user sees the action succeed but the charger stops.
qBound(6, val, 32) is correct in all cases:
- firmware reports 16 A (mono Trydan) → 16 (real installation limit kept)
- firmware reports 0 (aberrant) → 6 (safe floor, does not claim 32 A)
- firmware reports 65535 (aberrant) → 32 (IEC absolute ceiling)
The real installation limit still comes from MaxIntensity read on the device;
the bounds [6, 32] are only IEC guardrails, not hardcoded defaults.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Étape 1 complete: V2C Trydan nymea plugin (Modbus TCP, status Partial).
Plugin (IntegrationPluginV2c):
- discoverThings: delegates to V2cTcpDiscovery, maps results to
ThingDescriptors; createMethod "user" bypasses discovery.
- setupThing: registers a NetworkDeviceMonitor (IP tracking on DHCP
renewal), creates TrydanModbusTcpMaster, wires reachableChanged →
initialize() → state update on initializationFinished(), then starts
periodic polls on a 30s PluginTimer.
- updateThingStates: ChargeState 0/1/2 → pluggedIn/charging;
power = (PauseState==0 AND Lock==0) per evcc trydan.go Enabled().
- handleDynamicConflict: if Dynamic==1 (V2C internal PV optimizer),
write PauseDynamic=1 to suppress it. NEVER write Dynamic=0 (silences
ChargePower telemetry). Writes only on state transition to minimize
flash wear. cf. github.com/evcc-io/evcc/issues/28047.
- Action "power": writes PauseState then Lock (always in mirror), then
fire-and-forgets PauseDynamic if Dynamic==1. Action handlers filter
writeCompleted by register address to avoid reacting to concurrent
background writes.
- Action "maxChargingCurrent": clamps to [minIntensity, maxIntensity],
writes Intensity. Below 6A → pause instead of invalid setpoint.
JSON (integrationpluginv2c.json):
- ThingClass "trydan" implementing evcharger + connectable + networkdevice.
- settingsTypes: suspendInternalOptimizer (bool, default true) — allows
the user to disable Dynamic conflict management if they want the V2C
PID to remain active alongside the HEMS.
- State "chargeEnergy" exposed as diagnostic only (NOT sessionEnergy) —
firmware reliability on this register is unvalidated; cf. evcc #28047.
Build (v2c.pro):
- MODBUS_CONNECTIONS intentionally empty: the modbus-tool code generator
assumes non-overlapping block reads; leaving it empty skips generation
while still linking nymea-modbus via the PKGCONFIG in modbus.pri.
VendorId 56c3e7bb… is new (no existing V2C vendor in the repo).
PluginId f0692725… is new (generated for this plugin).
PORTING_STATUS_modbus.md documents register sources, the 4 items that
require validation on real hardware, and the "no beta matrix" constraint.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Architecture choice: Modbus TCP has no network discovery mechanism, so
discovery piggybacks on nymea's NetworkDeviceDiscovery (ARP/mDNS scan)
and then probes each host via GET /RealTimeData (V2C Trydan HTTP REST API).
The presence of a "Paused" field in the JSON response is used as a Trydan
fingerprint — it matches the charger model and is absent on generic devices.
cf. evcc charger/trydan.go realTimeData struct.
Probes run concurrently (one QNetworkReply per host) but the sequence
terminates via two independent paths that could both fire:
1. The NetworkDeviceDiscoveryReply::finished callback arms a 3-second
QTimer::singleShot to give outstanding probes time to complete.
2. Each probe completion path checks (m_scanFinished && pendingProbes==0)
and calls finishDiscovery() early if all probes are already done.
A m_finished bool guard prevents double-emission of discoveryFinished()
if both paths fire in quick succession (e.g. all probes complete exactly
as the timer fires).
Manual-entry (createMethod "user") skips discovery entirely; setupThing()
is called directly with the user-supplied IP.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Architecture choice: separate the register map / business logic from the
physical transport. TrydanModbusMaster (abstract QObject) holds:
- the complete ReadRegister / WriteRegister enums with addresses
- float32 Big/Big decode helpers (memcpy pattern, matches pymodbus
BinaryPayloadDecoder byteorder=Endian.Big wordorder=Endian.Big)
- all last-polled cached values accessible via simple getters
TrydanModbusTcpMaster (concrete, Étape 1) wraps libnymea-modbus
ModbusTcpMaster. Each register read is a distinct FC3(addr, 2)
transaction via a recursive runReadSequence/doNextRead pattern —
never a block read, because the V2C firmware's register windows
overlap (0x0BC2 and 0x0BC3 each span 2 registers, so a range read
starting at 0x0BC2 for ≥3 registers returns garbage for the second
value). Writes use FC6 (single uint16, not float).
writeCompleted(quint16 address, bool success) carries the originating
register address so that concurrent background writes (e.g. PauseDynamic
from the conflict manager firing during an action) do not interfere with
action handlers waiting on a specific register's acknowledgement.
0x177E (Dynamic) is intentionally absent from WriteRegister: writing it
to 0 silences ChargePower telemetry even though the charger keeps running.
cf. evcc charger/trydan.go and github.com/evcc-io/evcc/issues/28047.
Étape 2 (RTU) will add TrydanModbusRtuMaster on the same interface;
no changes to plugin or register logic will be required.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- abbterra-registers.json : suppression des 4 registres WO
(chargingCurrentLimitCommand, socketLockCommand, startStopChargingSession,
communicationTimeoutCommand) incompatibles avec le generateur libnymea-modbus 1.15
- integrationpluginabbterra.cpp/.h : suppression de executeAction + applyTimeoutSetting
- integrationpluginabbterra.json : power + maxChargingCurrent passes en lecture seule
(writable/displayNameAction retires)
La borne sera reconnue et lira les mesures ; le pilotage sera rebranche
quand le generateur 1.16 sera disponible.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
