Superseded verification requests
These are the original nine open tasks before the operator explicitly waived further proof collection. They are preserved as historical requests, not executed tests. Current acceptance/disposition replaces their closeout gate; this file does not supply per-case results.
- [ ] 0.3 On a reserved board, opt in to the Rust probe and record 568×1232 map/unmap, distinguishable software frame, real touch/move/cancel, bounded IPC acknowledgement, cold start, RSS, CPU and touch-to-commit/present cadence against the C reference and Qt probe. Keep this task open until the physical trace and camera evidence are committed. Do not install the probe as the normal session from a build alone.
- [ ] 5.1 On glass, record bottom Home→live deck, empty deck→drawer, drawer→app, top shade→Settings, contextual Back, keyboard escape, and no permanent final bar/footer; verify `python3 tools/capture-feature.py coherent-shell --provenance real-touch --duration 60 --description 'Gesture Home drawer shade and recovery on glass' --output-dir docs/evidence/coherent-shell`, plus `flock /tmp/k230-board.lock ./tools/console.py /dev/ttyACM0 --wait=3 'systemctl is-active shell'` saved as sanitized console evidence. Inspect camera and trace separately; operator-declared provenance alone is insufficient.
- [ ] 5.2 Record app→shrinking live card→adjacent swipe/snap→expand, reverse mid-transition, upward throw and refusal/timeout; verify `python3 tools/capture-feature.py coherent-cards --provenance real-touch --duration 60 --description 'Card motion and close recovery on glass' --output-dir docs/evidence/coherent-cards` and cite the sibling live-source/focus/cost evidence. An injected gesture or endpoint screenshot does not complete this task.
- [ ] 5.3 Record real-finger drawer/Settings scrolling, icon/text/fallback readability, long-press cancellation, app/keyboard conflict and accessibility aid discovery; verify `python3 tools/capture-feature.py coherent-gestures --provenance real-touch --duration 60 --description 'Gesture ownership and drawer discovery on glass' --output-dir docs/evidence/coherent-gestures` with an outcome table and no accidental activations. Keep the old bar rollback if any mandatory route fails.
- [ ] 5.4 Record preview while typing, shade scroll/side-swipe cancellation and committed dismiss, critical recovery and failed settings action; verify `python3 tools/capture-feature.py coherent-notifications --provenance real-touch --duration 60 --description 'Shade notification and settings motion on glass' --output-dir docs/evidence/coherent-notifications` with sanitized timestamps and outcome table.
- [ ] 5.5 Measure two-app live drag, app/deck reversal, drawer rise and shade arrival with `python3 tools/shell-motion-trace.py --board --output docs/evidence/coherent-shell/motion.json` under the board reservation; correlate input→scene commit→output presentation and camera-visible movement, report p50/p95/p99, blank/missed frames, CPU/RSS, renderer and actual cadence. Record failed budgets honestly.
- [ ] 5.7 On the reserved board, record up-then-left/right in one contact, both bottom quick-switch directions, stationary hold, diagonal reversal, one-app/end-of-deck behavior, target exit, and app/keyboard conflicts. Verify `python3 tools/capture-feature.py coherent-two-axis --provenance real-touch --duration 60 --description 'Two-axis app entry and bottom quick switch on glass' --output-dir docs/evidence/coherent-two-axis` with installed revision, sanitized scene/focus observations and an outcome table. A host/QEMU pass or native screenshot alone does not complete physical tracking/feel acceptance.
- [ ] 5.8 After the coherent image and required on-glass gates pass, select its matching boot files persistently under the reserved board lock, preserving the previous boot selection and recovery image. Hash every changed boot file against the built image; use a file-only update only where the kernel/initrd/firmware match, otherwise select the complete matching bundle or flash with the existing recovery procedure. Verify a fresh ordinary boot and exact `readlink -f /run/current-system`, shell service health and selected theme restoration via `python3 tools/console.py /dev/ttyACM0 --wait=5 'readlink -f /run/current-system; systemctl is-active shell shell-ui'` plus a sanitized boot-file hash report and reviewed UI capture. A test activation, updated profile, host image build or theme receiver restart cannot complete this physical task. Progress (2026-10-10): the combined daily mainline candidate built, passed host inspection, staged with all eight normal boot files/profile preserved, temporarily booted, and passed exact runtime/client/compositor/helper plus native Help checks; see `docs/evidence/coherent-shell/combined-board-candidate-2026-10-10/README.md`. The user then accepted the requested navigation/Help checks on that exact temporary build ("it all worked land it"). The guarded matching boot selection was installed and untouched ordinary reboot passed with fresh boot ID `effcb1f4-3cdc-4647-bd7b-517d4072ae1a`, matching system/profile/client/compositor/helper, four active services, all eight boot-file hashes, retained rollback roots and reviewed panel-native Help. Task 5.8’s exact console command passed; installation/runtime reports and sanitized console output are committed in the same evidence directory. No active appearance generation/report existed before or after, so user-selected theme restoration is not asserted. This proves the bounded installation/ordinary-boot portion; the prerequisite broader on-glass gates remain open and this task stays unchecked.
- [ ] 6.4 Coordinator only: install the reviewed shell artifact on the reserved board, activate dark and light themes, capture Home/drawer/overview/Settings/shade/theme picker and perform real glass drawer/search/theme apply/return gestures. Record exact operator commands, screenshot/camera provenance, readable theme contrast and limits in `docs/evidence/shell-polish/physical/`. Physical proof remains UNVERIFIED until this exists; use `./tools/console.py /dev/ttyACM0 --wait=3 "<reviewed runtime inspection command>"` plus native `grim` and real-glass observations, not an injected event alone.