evidence · 7.0 KiB

trackpad.md

Committed at docs/evidence/proposal-closeout/2026-10-01/trackpad.md. This is what grounds the requirements that cite it.

Open raw source

the-touchscreen-becomes-an-hdmi-trackpad: operator acceptance

Recorded 2026-10-01 from the project operator’s conversation. This is a physical operator report about the installed device, not an agent-run test, new camera capture, injected-event trial or quantitative measurement.

launch splash works. mouse nav and hdmi trackpad works. clock works fine don’t worry about digging in further i don’t think

The operator explicitly requests closeout, trusts their reports and waives additional difficult captures/retesting. Existing source, host/QEMU and board evidence remains unchanged and retains its original provenance. No new firmware, kernel, userspace deployment or board action was performed here.

Closeout worktree: /mnt/MediaVolume/home/jadams/src/gitlab.daringbit.com/josh/tdisplay-k230/.scratch/coordinated-work/accepted-trackpad-closeout; branch closeout/accepted-trackpad; base c2ba4e0870dad01300e1ebd52ffe2d7383063db5. Owned paths: this change and its archive, its affected capability specs, this evidence report and this card’s entry in docs/work-board-status.json. No board/serial reservation. The coherent closure build was serialized under /tmp/k230-nix-build.lock; other closeouts need no build reservation.

Disposition

Individual four-finger recognition and other unenumerated contact-count cases were not separately reported in this closeout. Their source/routing proof remains, but physical per-case claims stay UNVERIFIED. The operator accepts overall navigation and requests no further digging.

Prior committed evidence

Superseded verification protocol

The following are historical requests, not tests performed on 2026-10-01. Their disposition above replaces capture-only acceptance barriers or transfers unperformed measurements, rather than falsely checking a test as executed.

3.2 With an HDMI monitor and the panel dark, drag one finger across
      the touchscreen glass and confirm the pointer moves on the monitor;
      tap once and confirm a left click; two-finger-tap and confirm a
      right click; two-finger drag and confirm scrolling; a pinch gesture
      and confirm zoom (in whichever Wayland client is focused). Capture a
      photograph or screen-recording of the monitor showing the response,
      per AGENTS.md's evidence-class distinctions (a console transcript
      alone does not prove pixels/pointer motion reached the monitor).
      Commit under `docs/evidence/the-touchscreen-becomes-an-hdmi-trackpad/board/`.
3.3 Reboot back to the panel DTB (or, if no-reboot switching exists
      by then, disconnect HDMI) and confirm over the console that
      `k230-touch-trackpad` logs `mode -> DirectTouch` and that the
      coordinator's direct absolute touch mapping still works exactly as
      before this change (an `evtest`/touch check per the existing
      `display/touch` evidence pattern). This is the proof that trackpad
      mode never regresses panel-mode touch.
3.4 Resolve this change's `specs/display/touch/spec.md`
      `<!-- UNVERIFIED -->` marker against the outcome of 3.0–3.3: either
      remove it with the board evidence committed, or restate the
      requirement against whatever was actually observed (including a
      documented shared-GPIO23/24 interaction with the LT9611 bridge, if
      one is found, or a still-unresolved hang if 3.0 does not pass clean).
3.5 On the installed HDMI shell, use the real glass trackpad to click
      a Home/launcher icon and confirm its app opens. Commit a photograph or
      recording plus the launcher/Wayland log under this change's board
      evidence. Injected pointer input can independently test dispatch and
      app presentation, but does not satisfy this physical input gate.
3.6 With an app open on HDMI, pinch four real fingers inward on the
      board's touchscreen while its built-in display is inactive; confirm
      app overview opens. Record the physical observation and console log
      under this change's board evidence. Config parsing/IPC acceptance does
      not prove libinput recognized a real four-finger pinch.
5.6 Capture actual glass top/bottom two-finger swipes, card browsing/reversal/hold/fling, close/recovery, ordinary app scroll/pinch and pointer click. Use `python3 tools/capture-feature.py hdmi-trackpad-gestures --provenance real-touch --duration 30 --description 'HDMI two-finger shell gestures and ordinary app input' --output-dir docs/evidence/the-touchscreen-becomes-an-hdmi-trackpad/gesture-ux`; commit reviewed camera/native/console provenance and operator feedback. Leave physical feel open without that observation.
6.5 Capture real two-finger opening/content close/scroll/reversal/card browsing and three-finger keyboard gestures with `python3 tools/capture-feature.py hdmi-trackpad-gestures --provenance real-touch --duration 30 --description 'Uniform HDMI shell gestures' --output-dir docs/evidence/the-touchscreen-becomes-an-hdmi-trackpad/uniform-gestures`; retain physical gates without that observation.