Kirin CURRENT_NOW misread: display floor for current + plausibility floor for charge power (#152 v1)

[almothafar]

Fixes the user-visible fallout of the Kirin/HiSilicon CURRENT_NOW unit bug (#152): the fuel gauge reports mA instead of µA, so real readings arrive 1000× too small and slipped through every gate that only guarded against too-large values.

v1a — current display floor (BatteryRateTracker)

• New MIN_DISPLAY_CURRENT_MA = 10; the hasCurrent gate in computeRate now requires the rounded magnitude to reach it (was >= 1).
• Kills "Discharging −1 mA" / "+1 mA" on both surfaces (ongoing notification's mA fallback and the details table) — the misread values cap out at ~1–3 "mA".
• On genuine-µA devices, sub-10 mA draw only happens in deep sleep; nothing informative is lost.

v1b — charge power plausibility floor (ChargeSpeed)

• New MIN_PLAUSIBLE_MW = 100 (0.1 W), mirror of the existing MAX_PLAUSIBLE_MW guard: a phone that reports CHARGING is never genuinely taking in less.
• Kirin's fake ~3–4 mW estimate now reads as unknown instead of a known Trickle, which:
  • stops the wrong "Slow charging" tier in the charge-connected announcement and the ongoing notification (Redesign charger notification: charging speed (tier + wattage), wired/wireless, and 3 notification modes #122/Charging: show tier + wattage (not %/h) in the ongoing notification while charging (shrunk scope) #125) — they fall back to the plain label;
  • stops the guaranteed false slow-charge alarm (Charging: slow / bad-cable warning #123/Slow-charge warning: detect frayed cable / dying charger (#123) #150) — SlowChargeDetector's powerKnown gate keeps it asleep.

Tests

• Boundary tests at the display floor (9.4 mA hidden / 9.5 mA shown) plus the Kirin raw −800 case with the level-over-time rate still shown.
• Boundary tests at the power floor (99 mW unknown / 100 mW known) plus the Kirin ~3.8 mW case end-to-end through fromMeasurements.
• ./gradlew :app:testDebugUnitTest and :app:lintDebug pass.

On-device verification (Mate 10 Pro)

• No ±1/±2 mA in the notification or the table across several plug/unplug cycles (per the 2026-07-16 field observation, check both surfaces)
• Notification falls back to the plain status label during warm-up
• No "Charging slowly" warning during a normal wired charge

v2 (unit auto-calibration, showing the real current on Kirin) follows as a stacked PR.

Part of #152 — v1 of the two-stage fix; the stacked v2 PR closes it.

🤖 Generated with Claude Code