kerfmaster pin list ยท UI / workflow decisions
questionUnresolved, waiting on Aristide or Jordan.
Content last changed 2026-11-06 โ computed from the item itself, not typed.
Not reproduced headless (2026-11-03, scratch :8479, 1600x900, default + new layout, fixture nest): pan inside/off-sheet, drag over the right panel, shift-click drag, double-click -- no page selection, no band. Unconfirmed lead: Chrome's text-selection highlight over the collapsed Dimension strip (sideways text); the canvas pointerdown never cancels the default. Needs from Jordan when it recurs: which drag (pan, hand editor, other), panel layout, window size, whether one click on empty space clears it, whether resize clears it. Next attempt: part-list nest with the Nests chip row + his panel layout. Fix only after repro. 2026-11-03 lead (replaces the text-selection guess, which a resize would not clear): the drawing canvas is only re-measured by resize() on window resize / applyWorkspace. If the canvas area narrows on some other path (a panel appearing on a tab switch), the backing store keeps the old wider size squeezed into the narrower box: draw() clears only the new width, so the right-most strip keeps old frames (trail/stripes), and mouse positions no longer match drawn handles (start-point drag misses). Window resize re-measures = cleared. Repro plan: headless, switch tabs/panels and compare canvas.width/dpr to its on-screen width after each step; a mismatch = reproduced. Fix candidate (only after repro, own branch): canvas watches its own size (ResizeObserver -> resize()); screen-only. Jordan 2026-11-03 follow-up: no right panel opened or changed width; each time the large display looked COMPRESSED TO THE LEFT (fits a squeezed canvas). Repro attempt itself on hold: 'wait until we lock it in'.