Master publish unblock + issue #560 (TRV06) + bjawzodf/TY0201 + MemwarnGuard (2026-10-18, own words)
A1 — Auto-Fix + Publish red since 16/10
- Step "Scaffold missing driver.flow.compose.json (P2518 self-heal)" runs
scripts/gates/native-matrix-gate.js. - It counted 258 mandatory non-native clusters against a baseline of 255. The five new entries were the Tuya private cluster 0xE000 (57344) in
zigbee.endpoints[*].clustersofwall_remote_2_gang(ep1, ep2) andwall_remote_3_gang(ep1, ep2, ep3). - They arrived with the TS0042/TS0043 "interview shape" commits (afba3bd4b0, b09b3902ba). The later peer-compose commits (4a24c4d558, f5ed9b65bf) did not touch clusters.
- Fix (additive for behaviour, no couple / capability / binding touched): 0xE000 leaves the mandatory manifest clusters; the runtime listener overlay stays in
lib/devices/WallSceneRemoteHybridInit.js(E000 cluster + BoundCluster + raw-frame interceptor).data/native-matrix-baseline.jsonis untouched (shrink-only). Tests p2827/p2828 now assert 0/1/6 native + "no 0xE000 in the manifest".
A2 — issue #560 TRV06 _TZE200_rxq4iti9 / TS0601 (profile me167), reporter artisare, tested on 9.0.1512
Issue: https://github.com/dlnraja/com.tuya.zigbee/issues/560
- Battery: Z2M
TS0601_thermostat_3(zigbee-herdsman-converters, Koen Kanters and contributors) lists_TZE200_rxq4iti9(whitelabel EARU TRV06) and maps DP35 througherrorOrBatteryLow(value 1 = low battery, also published aserror). There is no battery percentage. A user write-up in Koenkk/zigbee2mqtt#25093 (MichaelHainbach, 2026-10-11, on the cousin_TZE284_o3x45p96) says bit 0 of DP35 is the low-battery flag and that the valve only sends DP35 when it is queried or its fault state changes. Another commenter (Jonstep101010, same issue) listed DP102 "Battery Low" from a Tuya DP table of that cousin: UNCONFIRMED for rxq4iti9, and DP102 is alreadyeco_tempin our me167 map, so it is NOT mapped. - Change: me167 DP35 transform is now bit 0;
alarm_batteryis added only when a DP35 report really arrives (no phantom alarm on silent valves).measure_batterystays unset (nothing real is reported). - Humidity: driver-wide capability list unchanged (other TRVs keep it);
measure_humidityis removed at init for the exact mfr_TZE200_rxq4iti9only, and never when a humidity value was actually stored. - Flow loader:
triggerDPReceivedused to retry the same failing card lookup on every DP (the poison-cache reset re-picked the device card). It now tries the device card, then the app card, remembers the working one, and goes quiet after both fail (one log line). The app.js FLOW-GUARD wrapper logs a missing card id once per method+id, then skips quietly (home_mode_changed / capability_value_changed_generic / device_interaction).
A3 — _TZ3000_bjawzodf / TY0201 (Wilhelm Dall suggestion, 2026-10-17)
- Source: Z2M
WSD500A(TS0201 +_TZ3000_bjawzodf, temperature + humidity + battery) and our DB entry P2266 (Temu round LCD). - The exact couple is listed by one driver only:
lcdtemphumidsensor(class sensor). No switch driver lists it. - Why it still paired as a socket:
device_plug_smart(class socket) has an EMPTYmanufacturerNameand productIds that include TY0201/TS0201 (plus other sensor pids), so Homey offers it for any manufacturer. Removing those pids is a removal and the file carries a 2026-08-21 "do not strip" note: left for the owner (see report). Meanwhilelib/utils/ClimateSensorClassHeal.jsfixes the class of an already paired device in place (socket/other -> sensor, one-shot, exact couple only) fromdevice_plug_smartandbutton_wireless_plug.
A4 — Universal Tuya 9.0.1506 crash (build 3590)
MemwarnGuard.handle called process.memoryUsage() unguarded; it throws ENOENT (uv_resident_set_memory) when the process is in trouble. Now wrapped (falls back to v8 heap stats, then -1), the memwarn listener also catches any handler failure, and the same unguarded call was wrapped in PerformanceOptimizer, DiagnosticReportExport and AutonomousVerificationEngine.