Home settles onto a whole, existing page (board)

Change: openspec/changes/home-pages-settle-after-a-cancelled-swipe.

Before the fix (system yvh5nh6n…, 2026-10-15)

• At 22:15 the operator’s two-finger trackpad presses on page 2 each ended in touch-cancel in the shell log, and a grim at that time showed Home frozen about 44% of the way between pages 1 and 2. None of their presses there reached a widget.
• At 22:29 a scripted right-click and badge click (tools/board-pointer.py) removed the only widget on page 2. before-fix-blank-vanished-page-1280x800-grim.png: Home stayed on the deleted page, blank and with no page dots, while still in rearrange mode.

After the fix (system 7csn4grb…, deployed live 2026-10-15 22:40)

Battery - Details was put back alone on page 2 by writing ~shell/.local/state/k230-shell/home.json before the switch restarted the shell. Then:

ssh k230 "$PY - 1280x800 'wait 300; swipe 1000 520 300 520 250; wait 1200; rclick 300 300; wait 800'" < tools/board-pointer.py
grim -> page2-right-click-1280x800-grim.png
ssh k230 "$PY - 1280x800 'wait 300; click 46 136; wait 1500'" < tools/board-pointer.py
grim -> after-removing-last-page-widget-1280x800-grim.png

The log shows home-widget-secondary-click (22:41:06), then pointer-down, pointer-up and home-layout-changed (22:41:14). home.json no longer contains battery_details. The second capture shows Home back on page 1 (Clock and Video) in rearrange mode. With one page left, no page dots are drawn. Evidence class: injected pointer + grim.

The cancelled-trackpad-pan path was not injected: tools/board-pointer.py drives a pointer, not the two-finger trackpad adapter. It is covered by host tests and by the operator’s task 2.2.