/**
 * Bring a focused item of a Marquee track back into view by SEEKING the lap, not by scrolling
 * (gh#983).
 *
 * The track travels by `transform` towards the inline start, so an item that has left the viewport
 * on that side sits in NEGATIVE overflow — which is not scrollable overflow. The browser's own
 * "scroll the focused element into view" therefore does nothing for it, and Tab through the track
 * landed on links that were partly or wholly invisible: measured 8 of 18 tab stops at 375px, one
 * at -149..-65 against a 75..1321 viewport at 1440px (WCAG 2.2 SC 2.4.11).
 *
 * The lap is seamless by construction, so any offset within it is a legitimate position: moving the
 * animation's `currentTime` moves the whole strip to where the item is visible, and resuming later
 * continues from there exactly as a pause would (G4).
 */
/** Inline-axis interval, in viewport pixels. */
export type MarqueeSpan = {
    start: number;
    end: number;
};
/**
 * The animation time at which the item sits fully inside the box, or `null` when it already does
 * (so tabbing across visible items never jolts the strip).
 *
 * It aims to CENTRE the item — clear of the `fade` mask at both edges — and settles for any fully
 * visible position when the centre is out of reach.
 *
 * It NEVER wraps into another lap. The copies look identical, but only the first holds the real,
 * focusable item; the clones are `inert`. Seeking by a whole lap therefore moves the focused item a
 * whole copy away while the strip looks unchanged — measured, the first attempt at this fix put
 * every item at exactly `centre − lap` (-139 against a 698 centre, lap 837) at 1440px. So the
 * answer stays inside the iteration the animation is in.
 *
 * `step` is how far the item moved when time advanced by a hundredth of a lap — measured rather
 * than derived, so `direction="end"` (a reversed animation) and the RTL keyframes need no case of their
 * own. A probe that crossed an iteration boundary shows a jump of nearly a whole copy the OTHER
 * way; `lap` (one copy, in px) is what tells the two apart. Only its SIGN is used. It is a hundredth
 * of a lap and not a millisecond because a slow track (99.6s a lap on the logo wall at 320px) moves
 * 0.008px in a millisecond, which measured as 0 — and a zero step meant no seek at all.
 */
export declare function marqueeRevealTime({ item, box, time, duration, lap, step, }: {
    item: MarqueeSpan;
    box: MarqueeSpan;
    time: number;
    duration: number;
    lap: number;
    step: number;
}): number | null;
/** Wire it to the live DOM. A no-op where there is no running animation (reduced motion, jsdom). */
export declare function revealInMarquee(viewport: HTMLElement, track: HTMLElement, item: Element): void;