fix: a palm on the glass no longer drives the pen's stroke (#338)

[difficcd]

The other half of #338 — "펜이 뚝뚝 끊김 (태블릿 환경에서)". Found by reading the pointer handlers rather than on the device, and the asymmetry it turns on is specific enough to be worth stating plainly.

Cause

A PC has one pointer. So every pointer event arriving at the canvas belonged to the stroke being drawn, and no handler ever had to ask which pointer it came from. A tablet has several: the pen, and the palm resting on the glass beside it.

startDraw refuses to begin a gesture from a touch — the palm rejection that has always been there:

if (e.pointerType === 'touch') return;

But nothing stopped a touch from feeding the gesture already running:

handler	guard it had	what the palm could do
onDraw	if (!gesture.drawing.current) return;	its moves were appended to the pen's stroke as points — a jump across the drawing
stopDraw	() => { — took no event, so it could check nothing	lifting off committed the pen's stroke mid-line

After that early commit the pen kept moving over a gesture that had ended, drawing nothing until it was lifted and put back down. A palm that rests and shifts while you write does this repeatedly: a line in pieces, on a tablet and nowhere else.

Change

The gesture already knew whose pointer it was — pointerId was set by begin() and read by nobody. It is now read, through one predicate in useGesture.ts that both handlers share:

export function pointerIsOwner(active, ownerId, eventId) {
    if (!active || ownerId === null || eventId === null || eventId === undefined) return true;
    return eventId === ownerId;
}

Two things deliberately pass through:

• Events with no pointer to name. onPointerLeaveCanvas ends a stroke without an event; refusing those would stop leaving the canvas from stopping the stroke.
• Everything while nothing is in flight. onDraw does its hover work through the same guard, and refusing a pointer there would leave the cursor stale over a selection's resize handles.

stopDraw's parameter is optional for the same reason.

Tests

test/hooks/gesturePointer.test.js, four cases — the owner check, the hover passthrough, the no-event passthrough, and:

• pointer id 0 is a real pointer, not a missing one — the guard cannot test eventId for truthiness, because a first pointer is legitimately id 0 on some browsers and !0 would hand every stroke to whichever pointer came second. That is the shape this fix would most plausibly come back in.

1284 tests pass; npm run check green (460 helpers indexed).

To try

On the tablet, draw a long line with the pen while resting your hand on the screen — which is the natural way to draw and the condition this needs. The line should stay one unbroken stroke. Dragging a finger across the canvas mid-stroke should now do nothing to it.

This is the stroke break-up half of #338. The pencil-pressure half is #360.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
easy-mv-maker	Ready	Preview	Oct 1, 2026 6:26am UTC