Upstream daily notes, 2026-10-27
Written in our own words. Facts and ids only, each with its source.
Coverage matrix (scripts/kb/z2m-coverage.js, all sources)
| Branch | Total | Absent | Compacted | Wrong class | Partial | Complete |
|---|---|---|---|---|---|---|
| master (before, 26/10) | 3651 | 536 | 2301 | 3 | 157 | 654 |
| master (after, 9.0.1602 tree bf486bba98) | 3573 | 520 | 2218 | 1 | 161 | 673 |
| stable-v5 (before) | 3651 | 564 | 2361 | 9 | 128 | 589 |
| stable-v5 (after, 5.12.515) | 3573 | 544 | 2316 | 9 | 130 | 574 |
| bastien-home (before) | 3651 | 497 | 2363 | 7 | 142 | 642 |
| bastien-home (after, 1.0.217) | 3573 | 471 | 2324 | 7 | 142 | 629 |
The total fell by 78 because Johan's app (JohanBendz/com.tuya.zigbee 328d24e, v0.2.76) now lists fewer cross-product couples (1284 to 1161 Tuya couples). Nothing was removed on our side. The master gains (complete +19, wrong class 3 to 1) come from the one-couple-one-driver batches pushed on master overnight. The only wrong-class row left on master is _TZE284_ikul00sx / TS0601 in fan_light_tuya_5_levels (Z2M exposes a fan plus a light; our class is fan, the light part is a capability, so this is a matrix limitation rather than a misplacement).
Parked couples (publish budget still full)
Payload budgets at this run: master 19997/20000 with the TS0601 top-up (15498/15500 before it), stable-v5 15500/15500, bastien-home 15500/15500. The seven couples parked on 2026-10-26 (UPSTREAM_DAILY_2026-10-26.md) are still absent on all three branches and stay parked; the three wrong-class leads listed there (_TZE204_6a4vxfnv, _TZE200_edl8pz1k, _TZE284_2fnssffc) are unchanged. Not placed this run because another session is moving couples on master and the budget has no room; adding them now would only push out attested couples.
Upstream changes since 2026-10-26
| Source | Range | What matters for us |
|---|---|---|
| zigpy/zha-device-handlers | d54cf9e..2b1abce | PR #5531: Xiaomi/Aqara read responses are now decoded like reports (0xFF01 struct), but a model attribute (0x0005) seen in a read response is no longer treated as a reset-button press. Our lib/xiaomi/XiaomiSpecialHandler.js already decodes 0xFF01 from read responses; lesson: never fire button events from a read we requested. PR #5547/#5395: constant attributes are neither cached nor reported as unsupported (relevant to the P3011 unsupported guard: an attribute we answer locally must never be marked unsupported). PR #5239: Aqara E1 TRV schedule accepts 00:00. No Tuya couple added. |
| make-all/tuya-local | 8de6b3c..2454696 | New Wi-Fi config HQL DS-123 radar switch, product id k2xfhfqsqbqvbbig (DP1 switch, DP102 presence string "presence", DP103 delay 5-86400 s, plus sensitivity and alarm DPs), from PR #6274. Not in our data/wifi/tuya-local-product-schemas.json (pinned at tuya-local 95300d05); lead for the next schema refresh. Klarstein Bornholm heater config fixes (Wi-Fi). |
| SmartThingsCommunity/SmartThingsEdgeDrivers | 3c3946ec07..3a48d3d84f | Sonoff SWV2C valve fingerprint (#3171). Sonoff, not Tuya: out of scope for the matrix. |
| JohanBendz/com.tuya.zigbee | 61577d8..328d24e | Releases v0.2.60 to v0.2.76 and rain sensor _TZ3210_tgvtvdoc / TS0207 (#983). All 126 manufacturer ids added in the range are already listed on our three branches (the rain sensor is in rain_sensor on all three). Johan issues updated since 26/10 (#1542 hy369 _TZE200_ckud7u2l, #1461 NTCHT02, #1382 _TZE200_crq3r3la, #869 _TZ3210_j4pdtz9v, #1026 _TZ3000_ja5osu5g) all map to existing drivers. |
| Unchanged | zhc 7462b54 (remote HEAD checked), deCONZ, Blakadder, tinytuya, localtuya, zigbee-clusters, zigbee-OTA, Hubitat, ioBroker, Tasmota, Domoticz |
Unknown commands
No new unknown command, cluster or DP confirmed by a source this run. dlnraja/com.tuya.zigbee had no new issue since the last run (only the daily digest). Gmail diags and the forum topics are read by their own routines later today.