Implementation lane: Reborn product and battery

Dated implementation receipt. The campaign identities, test counts and pending actions below belong to the recorded source cut. The newest sealed integrated candidate is Baseline02. UART01 hardware now proves bounded C1/C2/C3. Baseline02 new-image/automatic cold boot and unrelated release gates remain open.

Entry: Linux 9ac2c0b7018ce707ec15f8845daf58ce38ec18a0, Reborn eb64b8b21058eacb1ef183f7570cca8100cbbb46. Existing dirty Linux documentation was preserved. No device writes, flash, push or Git identity changes occurred.

Priority 8 — Product v2

Starting state: Product v2 and the new first-complete-frame splash handoff were already implemented. EQ existed in FFmpeg but had no product control; sleep had no typed UI contract; diagnostic export meant private state backup. Codec selection required Stopped although the ordinary UI only exposed Pause.

Implementation:

• Audio → Equalizer uses existing DSP/limiter with five default bands,

  ±12 dB/1 dB wheel editing, Select apply, Back cancel and Reset to Flat. Existing custom bands are preserved up to the engine's eight-band limit. Reset Settings now rebuilds playback when disabling an active EQ.

• Audio → Output Details exposes existing observed source bits/rate,

  decoder/DSP precision and opened PCM format/rate through Diagnostics. No native high-resolution claim is inferred from source or container width.

• System/Quick Settings → Sleep consumes typed power.sleep and

  y2-platform sleep request, with explicit requested/refused/sleeping/ restoring/restored/restore-failed states. Active playback is refused with actionable wording. Ordinary deep sleep still returns physical_qualification_required from the platform; owner qualification retains its existing explicit path. Same-boot/kernel-completion checks protect a restored claim. Short Power still means screen off, preserving playback and volume controls.

• Bluetooth presents preference separately from actual negotiated codec.

  Paused/stopped playback permits selection; reconfiguration and the platform PCM lease still exclude active streams. SBC quality and LDAC quality/ABR controls reflect supported runtime settings; requested/effective settings and pending restart remain distinct. Quality labels include both LDAC rate families. CLI help and parser tests cover every viable Classic preference. Final integration review fixed initial read-only capability discovery (audio lane), saved-preference application once per observed peer connection, and SBC-only quality visibility. Selection waits for paused/stopped playback and a non-running PCM; PCM replacement cannot trigger repeated Auto selection. Accepted CLI preferences also persist to ordinary player settings.

• Diagnostics exposes a separate redacted report export, immediately shows

  its archive path/retrieval instructions, and labels state export private. Failed operations cannot display a successful export message. Opening the result retains the previous page/focus, so Back returns to Diagnostics.

• Estimated SOC is named on the Battery page. Long modal lists now account for

  description height, fixing focus rows outside the panel.

• Ready-frame splash, shutdown save/audio stop/SQLite close/black frame/

  backlight off/ack sequence and input duplicate/acceleration/direction-reset foundations remain unchanged and included in current software validation.

Files: Reborn app/reborn/src/{main,ctl}.rs, preview fixtures; crates/reborn-core/src/{lib,platform}.rs; platform {client,power}.rs; UI {lib,pages,screens,components,diagnostics,tests}.rs; docs/ui/REBORN-FEATURE-COMPLETION.md and platform boundary documentation. No additional direct Y2 filesystem/shell access exists outside reborn-platform.

Validation: final current-source fmt check and strict workspace clippy passed. The combined workspace passed 238 tests, zero failures/ignored, including codec discovery/history/reconnect and complete product regressions. A concurrent fork/CLOEXEC fixture race in the PCM lease test was corrected by using the same bounded reacquire wait already used earlier in that test; production lock semantics remain unchanged. Host preview generated 69 480×360 states; new EQ/picker/sleep/LDAC cases were visually inspected. Output is Y2Reborn/out/feature-completion-previews/; final logs are Y2Reborn/out/feature-completion-final-{fmt,clippy,workspace}.log.

Remaining physical proof: actual sleep restoration, new-control wheel and playback behavior on the panel, codec peers/qualities/ABR, audio output, boot/shutdown video and diagnostic retrieval in the one integrated run. Distribution: optional codec enablement is the private integration profile; product controls do not resolve redistribution/certification questions. Final lane state: PARTIAL / source implemented and software validated, with full sleep and high-resolution capabilities still platform/physical gates.

Priority 4 — Battery/charging

Starting state already includes 70/450/650 mA source ceilings, 4.175 V CV, voltage-limited termination/hold/recharge, watchdog ownership, SOC estimate, warning/critical policy, and a debounced 3.4 V shutdown floor. Reborn already saves session, stops audio, closes SQLite and acknowledges the bounded platform shutdown; the coordinator syncs and requests ordinary init shutdown. These paths were reviewed and preserved rather than changing proven electrical policy without new evidence.

Exact source findings:

• Retained Y2 fgauge_read_current at 0xc04caa04 and

  fgauge_read_columb at 0xc04caa2c return zero immediately and never fill the output pointer (20-byte functions). These are stubs, not zero measured current or a coulomb counter. Same functions in the exact MT6582 BSP also return STATUS_OK without producing data.

• Stock temperature conversion assumes a 10 kohm NTC, 16.9 kohm pull-up,

  27 kohm pull-down, 1.8 V. Stock public temperature is constant 25 C. Existing physical BATON raw10387–10388/570.6mV did not establish NTC wiring.

• Stock charger-current computation subtracts ISENSE/BATSNS and assumes

  68 milliohm. Existing paired read result (-0.081mV mean,1.638mV standard deviation with charging disabled) cannot validate resistance, channel offset, or net pack-current routing. It is not a calibrated current path.

• Existing raw IIO BATON1/ISENSE/BATSNS and die thermal acquisition already

  provide safe instrumentation. A PMIC die temperature is not pack temperature.

Primary sources freshly fetched: MT6582 battery_meter_hal.c, MT6323 AUXADC implementation. Local copies under out/feature-completion-battery-research/, SHA256 respectively f0abff44297fa5613cec5e8e74cd0bfe60b67348156e1df6bf36d5b5b453cc0c and 196b4605c4d367189f6b7b4a6263d617c0c90dfade649020a4e7139acdf27a98. Y2-specific retained evidence: ADC result, stock reconstruction, charger policy provenance.

Implementation: truthful estimated-SOC presentation only; no fabricated sensor conversion or unproven register/voltage changes. Linux targeted validation: power contract 8 PASS, SOC 5 PASS, charger 4 PASS. Existing broader power/charger tests remain part of parent integrated validation.

Remaining exact blocker: pack/sense schematic or continuity/resistor evidence, reference current/temperature calibration and actual charge/discharge topology. The existing tools/development/sample-battery-adc.sh records bounded paired ADC observations; add its results to the integrated harness, not a separate flash workflow. Owner equipment is required to turn those voltages into proven pack current/temperature. Fuel gauge/coulomb support is not established, not proved physically impossible. No new image alone can establish wiring.

Final lane state: conservative charging/low-battery source retained and tested; measurement additions BLOCKED_BY_HARDWARE_PROOF. Full discharge reserve, charging/hold/recharge and low-voltage graceful shutdown remain physical proof.