POWER-FINAL-29 — finish-stage and discharge-estimate corrections

Status: INSTALLED, FULL CHARGE QUALIFIED (2026-10-14, archive 127, boot cac3e209). Package out/y2linux-platform-v1-power-final-candidate-29/ (fallback = 28), receipts out/deep-sleep-final/build29-receipts/. BOOTIMG 3f82ae593a936067730cd735c780cf42d483457ab8f2591590748448cc6efc07, Y2ROOT 5d9f298bb8432a0bcf13acb1ae2498e9d95252764f4f461003d9a0bc2b717f52. Identity: build Y2LINUX-PLATFORM-V1-POWER-FINAL-29, kernel 6.18.56-y2linux-platform-v1-power-final-29, root 2026.08-platform-v1.59, release 1.0.0-platform-v1-power-final-candidate.29; Linux f9ce653 (changes f2c9e0b), Reborn 1c9edd4. Contains 27 and 28.

28 on the device (archive `20261014/126`)

Boot 8f43c71e started at 89 % from the power-on OCV (27's loaded-sample method: ≈81 %); the restart (129e5611) restored the saved 85 %. One short sleep: pmic_sleep 1/1, last_changed 0x3ff, vsys_refusals 0 (27's VSYS step works). A second request was refused by the known wake_event_in_progress race. Charge (charge-log.jsonl): 450 mA CC for 15 min to 4.175 V, finish stage entered without fault, then 70 mA: the pack fell back to 4.09 V and stayed flat (ISENSE est. ≈90 mA, screen off, idle, awake on USB). The 3-h top-off budget would have ended it as a timeout fault. On battery the display drifted 89 → 85 in 3.5 min toward the loaded-voltage SOC.

What 29 changes

Finish stage steps 450/650 → 200 → 70 mA at successive 4.175 V samples; at 70 mA it completes on 4.175 V, the stock
taper rule, or a plateau (no 5 mV rise in 10 min). Full still needs the ≥4.07 V rest. The 3-h top-off budget completes a finishing cycle (stock BAT_TopOffModeAction); 24-h and precharge timeouts stay faults. charging_state shows finish as the current cap, plateau_uv, topoff_done.
On battery with samples ≤60 s apart, the estimate follows stock oam_run (model current (OCV(SOC) − V)/R(V)
integrated against 1300 mAh); after longer gaps (suspend) the rested voltage is used as before.

Tests

test_charger (steps, plateau from the archive-126 trace, rise restarts the window, top-off budget completes only a finishing cycle), test_battery_soc (archive-126 drift: stays ≥88 instead of 85; sustained load still discharges; suspend gap uses the rested voltage). Kernel warnings unchanged (47). validate PASS, ELF 354, installed ARM, privacy, inventory 108, container suite 708 OK.

Owner checks after the flash

A full charge on the PC with SSH: finish 200000 → 70000 → Full, full_rest_uv, no fault.
After Full: 100 on the charger; unplug: 100 → 99 → 98 without a quick sag while using it.
Restart: the percentage stays.

On the device (archive `20261014/127`, `charge-log.jsonl`)

Boot restored the saved 94 % (power-on OCV 4.094 V). Charge with SSH, no fault at any point: 450 mA → 4.175 V → 200 mA (28 min, sense drop ≈46 mV, voltage rising) → 4.175 V → 70 mA → voltage flat 4.13–4.15 V → plateau completion after 10 min (status=3) → rest ≥4.07 V for 60 s → Full (status=4). The owner unplugged once 1 s after the first plateau hold (cycle reset by design), the replugged cycle completed at 3676 s: full_rest_uv 4128442 (26/27: ≈4.09 V), recharge_uv 4098442, display 100, anchor_percent 97.68, full_latched true. Owner observation: while charging the display stayed at 97–98 until Full (stock 99 cap / estimate), then 100.