android: the media mirror grows a queue window the card never shows

The Kotlin side of the mirror has always known about exactly one track, and
media3's `SimpleBasePlayer.getState()` — the thing an Android Auto session is
built on — wants a playlist and an index into it. So the data lands a round
before the migration that needs it, where it cannot break anything:
`NowPlaying` and `NowPlayingArgs` carry a bounded `queue` of at most
twenty-one entries, the current item and ten either side, each with its id,
title, artist, album, duration and art key, plus `queueIndex`, where the
current one sits within that window rather than within the whole queue.
`src/mobile/media.svelte.js` fills them from `player.queue` and `player.qi`;
`MediaService` ignores every bit of it, so the card is byte for byte the card
it was. Ten either side because one covers the card's own prev and next and
nothing else while a queue list — which is what Auto asks for — wants a
screenful plus a scroll both ways, and it is a cap rather than a length: a
two-thousand-track queue crosses the bridge as twenty-one entries exactly
like a twenty-one-track one, measured at two kilobytes of JSON for twenty-one
short titles and six with long ones and real cover paths, against the thirty
to eighty of base64 cover that already ride the same call once per record.

Nothing renders it, which is the whole difficulty in checking it — and both
hops of this bridge fail silently, because serde drops an unknown field and
tauri's Jackson mapper has `FAIL_ON_UNKNOWN_PROPERTIES` off, so a key spelled
snake_case where camelCase was wanted arrives as nothing at all with no error
anywhere. `MediaSessionPlugin.update` logs one line per update, and the phone
walk holds that line against `__ss.card` — the mirror media.svelte.js now
keeps of what it last sent, the shape xfernote has kept of the shade since it
was written — field for field, through a next and a previous. The art-key
count is compared as well as the size and the index, since that is the only
camelCase field an entry has. The check arms at queue index one rather than
at wherever the walk left off, which buys a row on either side — `prev()`
seeks to zero at the top of a queue and `next()` stops at the end — a track
that has only just started, so `prev()` is still a step rather than the seek
it becomes past five seconds, and a window whose own index is the queue
index, because nothing is clamped off the front while the index is under ten.
Each of the three readings has to be a fresh one at both ends, the push
counter up and a new line in the log, since a `prev()` that lands back where
it started reads exactly like a stale line otherwise; and a queue too short
to arm, or repeat-one, under which `next()` seeks instead of stepping, fails
the check rather than skipping it.

The plugin's `consumer-rules.pro` keeps the new `@InvokeArg` class by name
beside the one it already keeps; without that line a release build would fill
`NowPlayingArgs` and leave every queue entry null, quietly. Rust and Kotlin
both changed, so this one cannot go down `dev.sh` — it needs the full
`deploy.sh` detour. And everything the walk already asserts about the card is
untouched, which is the real check on this round.

ANDROID.md

@@ -857,6 +857,50 @@ with a `ClassNotFoundException`. `src-tauri/android/build.sh` exports it;
 after adding a plugin, touch `src-tauri/build.rs` too, so the app's build
 script re-runs and rewrites the wiring.
 
+Every update also carries a **queue window** that nothing reads. `NowPlaying`
+(Rust) and `NowPlayingArgs` (Kotlin) hold a `queue` of at most twenty-one
+entries — the current item and ten either side, each with its id, title,
+artist, album, duration and art key — plus `queueIndex`, where the current one
+sits *within that window* rather than within the whole queue.
+`src/mobile/media.svelte.js` fills both from `player.queue` and `player.qi`;
+`MediaService` ignores them completely, so the card is byte for byte the card
+it was. The reason is media3: `SimpleBasePlayer.getState()` wants a playlist
+and an index into it, and the data that migration needs should have been
+running on the phone for a while before anything depends on it. Ten either
+side because one covers the card's own prev and next and nothing else, while a
+queue *list* — which is what Auto asks for — wants a screenful plus a scroll
+in both directions; and it is a cap, not a length, so a two-thousand-track
+queue crosses the bridge as twenty-one entries exactly like a twenty-one-track
+one does. Measured at two kilobytes of JSON for twenty-one short titles and
+six with long ones and real Android cover paths, against the thirty to eighty
+kilobytes of base64 cover that already ride the same call once per record —
+and an update is a track change, a transport change or the thirty-second
+top-up, never a position tick.
+
+Because it is on no screen, the walk reads it out of `logcat`:
+`MediaSessionPlugin.update` logs one `media window n=… i=… art=… at=…` line
+per update under the tag `SoulStream`, and the phone walk's `queue-window`
+check holds that against `__ss.card` — the mirror `media.svelte.js` keeps of
+what it last sent, beside xfernote's — field for field, through a next and a
+previous. Field for field is deliberate: a key spelled snake_case where the
+bridge wants camelCase arrives as nothing at all on both hops, with no error
+anywhere, which is why the art-key count is compared and not just the size and
+the index.
+
+Three things in that check are load-bearing and easy to undo by accident. It
+arms at queue index **one** rather than wherever the walk left off, which buys
+a row on either side (`prev()` seeks to zero at the top of a queue and
+`next()` stops at the end), a track that has just started (`prev()` is a seek
+again past five seconds), and the fact that while the queue index is at most
+ten nothing is clamped off the front of the window — so the window's index IS
+the queue index, which is what lets the logged index be compared against the
+queue's directly. It requires each reading to be a FRESH one at both ends, the
+push counter up and a new line in the log, because a `prev()` that lands back
+on the row it started from reads exactly like a stale line otherwise. And it
+FAILS rather than skips when it cannot be armed — a queue under three rows, or
+repeat-one, under which `next()` seeks instead of stepping — with the reason
+in the message.
+
 ## What a screen reader says
 
 Every control on the phone is named by what it does, in the app's own
@@ -1851,12 +1895,13 @@ read over real IPC; a cold artist page, its ⋯, the whole-discography page and
 its A–Z; the first-run preview with `setupWrites` at 0; eleven raw backend
 strings through `__ss.say`; and a refused transfer's sentence.
 
-Then, after the step loop, seven things the mock cannot do at all: a
+Then, after the step loop, eight things the mock cannot do at all: a
 force-stop and the paused restore that follows it (waited out past the
 thirteen-second FGS-kill window), the position restored, playback resumed, the
 system media session and its ownership, media keys, ten seconds in the
-background with the position still advancing, and SoulStream's own mixer
+background with the position still advancing, SoulStream's own mixer
-underruns.
+underruns, and the queue window that rides along with every card update and
+that nothing on the card reads.
 
 One block in the phone walk pulls bytes, and only when the app is signed in:
 the trailing network block types `--query` into Search, streams a peer's copy

crates/tauri-plugin-media-session/android/consumer-rules.pro

@@ -7,8 +7,9 @@
 # FindClass), constructed through "(Landroid/app/Activity;)V", and its two
 # @Command methods — update and clear — are dispatched by NAME through
 # PluginHandle's commands map. NowPlayingArgs is filled by Jackson field by
-# field. Rename any of it and the app still builds, installs and plays; the
+# field, and QueueItemArgs — one per entry of the queue window it carries —
-# lock-screen card simply never appears.
+# the same way. Rename any of it and the app still builds, installs and
+# plays; the lock-screen card simply never appears.
 #
 # The other half of this plugin is in the app-wide file: the card's buttons
 # come back as the plugin event "action", which the WebView subscribes to
@@ -20,3 +21,4 @@
 # them by name. Not repeated here: one place decides each of them.
 -keep class com.soulstream.mediasession.MediaSessionPlugin { *; }
 -keep class com.soulstream.mediasession.NowPlayingArgs { *; }
+-keep class com.soulstream.mediasession.QueueItemArgs { *; }

crates/tauri-plugin-media-session/android/src/main/java/com/soulstream/mediasession/MediaSessionPlugin.kt

@@ -28,6 +28,32 @@ class NowPlayingArgs {
     var art: String? = null
     var hasNext: Boolean = true
     var hasPrev: Boolean = true
+    /** the queue around the current item — it and up to ten either side.
+     *  NOTHING READS IT: MediaService ignores both of these on purpose, so
+     *  the card is byte for byte the card it was. media3's
+     *  SimpleBasePlayer.getState() wants a playlist and an index into it,
+     *  and this is that data arriving a round before the migration does.
+     *  Nullable because a non-null Kotlin property whose setter Jackson
+     *  hands a null throws, and the only thing between here and the JS that
+     *  builds it is two JSON hops. */
+    var queue: List<QueueItemArgs>? = null
+    /** where the current item sits in `queue` — not in the whole queue;
+     *  null when the window is empty */
+    var queueIndex: Int? = null
+}
+
+/** One entry of the queue window. Flat and small: it crosses the bridge on
+ *  every update, so it holds what a media3 MediaItem is built from and no
+ *  more. `artKey` names the cover the way NowPlayingArgs.artKey does; the
+ *  bytes never ride along here. */
+@InvokeArg
+class QueueItemArgs {
+    var id: String? = null
+    var title: String? = null
+    var artist: String? = null
+    var album: String? = null
+    var duration: Double? = null
+    var artKey: String? = null
 }
 
 /**
@@ -52,6 +78,20 @@ class MediaSessionPlugin(private val activity: Activity) : Plugin(activity) {
     @Command
     fun update(invoke: Invoke) {
         val args = invoke.parseArgs(NowPlayingArgs::class.java)
+        /* The window is on no screen anywhere — MediaService ignores it and
+           the card cannot show it — so logcat is the only place on the
+           device it can be read back, and tests/mobile/smoke.mjs reads it
+           from exactly here. One line per update, and an update is a track
+           change, a transport change or the thirty-second top-up, never a
+           position tick. `at=` goes LAST because a title may hold anything,
+           spaces and `=` included. */
+        val window = args.queue ?: emptyList()
+        android.util.Log.i(
+            "SoulStream",
+            "media window n=" + window.size + " i=" + (args.queueIndex ?: -1) +
+                " art=" + window.count { !it.artKey.isNullOrEmpty() } +
+                " at=" + (window.getOrNull(args.queueIndex ?: -1)?.title ?: ""),
+        )
         ensureNotificationPermission()
         MediaHub.update(activity.applicationContext, args)
         invoke.resolve()

crates/tauri-plugin-media-session/src/lib.rs

@@ -41,12 +41,50 @@ pub struct NowPlaying {
     pub has_next: bool,
     #[serde(default = "yes")]
     pub has_prev: bool,
+    /// The queue around what is playing: the current item and up to ten
+    /// either side, rebuilt on every update.
+    ///
+    /// NOTHING READS IT YET, and that is the point — `MediaService` ignores
+    /// both of these fields, so the card is exactly the card it was. media3's
+    /// `SimpleBasePlayer.getState()` wants a playlist and an index into it,
+    /// and when that migration comes the data it needs should have been
+    /// running on the phone for a while rather than have been written the
+    /// same day.
+    ///
+    /// A CAP, not a length: a two-thousand-track queue crosses the bridge as
+    /// twenty-one entries exactly like a twenty-one-track one does.
+    #[serde(default)]
+    pub queue: Vec<QueueItem>,
+    /// Where the current item sits IN `queue` — not in the whole queue.
+    /// `None` when the window is empty.
+    pub queue_index: Option<u32>,
 }
 
 fn yes() -> bool {
     true
 }
 
+/// One entry of the queue window. Deliberately flat and small: this crosses
+/// two hops of IPC on every update — the WebView to Rust, Rust to Kotlin —
+/// so it carries what a media3 `MediaItem` is built from and nothing else.
+/// `art_key` is the same kind of key `NowPlaying::art_key` is, the cover's
+/// path, so whatever decodes a bitmap for the card can key the window's
+/// covers off the same thing; no bytes ride along here.
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+#[serde(rename_all = "camelCase")]
+pub struct QueueItem {
+    /// The track's own id as a string — a library track's is a number, a
+    /// catalog track's is `cat:<n>`, a playlist row's is `pl<n>`. NOT a
+    /// unique key: one queue may hold the same track twice, and a window
+    /// entry is identified by where it sits, not by this.
+    pub id: Option<String>,
+    pub title: Option<String>,
+    pub artist: Option<String>,
+    pub album: Option<String>,
+    pub duration: Option<f64>,
+    pub art_key: Option<String>,
+}
+
 #[cfg(mobile)]
 pub struct MediaSession<R: Runtime>(tauri::plugin::PluginHandle<R>);
 

src/mobile/Mobile.svelte

@@ -47,7 +47,7 @@
     import { loadPrefs } from "./prefs.svelte.js";
     import { setup, loadSetup, maybeOpenFirstRun, openSetup, finishSetup, setupState } from "./setup.svelte.js";
     import { catalog, artistKey } from "./catalog.svelte.js";
-    import { bindMedia, syncMedia } from "./media.svelte.js";
+    import { bindMedia, syncMedia, mediacard } from "./media.svelte.js";
     import { slsk, bindSlsk, connectStored, refreshRooms, unreadTotal } from "./slsk.svelte.js";
     import { net, bindNet } from "./net.svelte.js";
     import { resolver, streamSong, streamCatalogAlbum } from "./resolve.svelte.js";
@@ -323,7 +323,7 @@
         requestAnimationFrame(() => requestAnimationFrame(() => { clearTimeout(sealTimer); sealBoot(); }));
         /* the probe handle — the desktop has __soulstream for the same reason:
            tests/mobile/smoke.mjs drives the real app through it */
-        window.__ss = { nav, top, go, push, back, openSheet, closeSheet, openHalf, closeHalf, player, lib, art, catalog, home, loadHome, lyr, loadLyrics, slsk, boot, net, resolver, streamSong, streamCatalogAlbum, sources, api, transfers, xfn: xfernote, pls, notice, importer, artTally, imp: { startFolder, startFile }, setup, openSetup, setupState, wifi: { downloadsAllowed, cellular }, next, prev, toggle, seekSecs, playQueueAt, patch, contacts, loadContacts, say, refreshRooms };
+        window.__ss = { nav, top, go, push, back, openSheet, closeSheet, openHalf, closeHalf, player, lib, art, catalog, home, loadHome, lyr, loadLyrics, slsk, boot, net, resolver, streamSong, streamCatalogAlbum, sources, api, transfers, xfn: xfernote, card: mediacard, pls, notice, importer, artTally, imp: { startFolder, startFile }, setup, openSetup, setupState, wifi: { downloadsAllowed, cellular }, next, prev, toggle, seekSecs, playQueueAt, patch, contacts, loadContacts, say, refreshRooms };
         return () => { if (unlisten) unlisten(); };
     });
 </script>

src/mobile/media.svelte.js

@@ -8,6 +8,54 @@ import { invoke, addPluginListener } from "@tauri-apps/api/core";
 import { player, hooks, toggle, next, prev, seekSecs } from "./player.svelte.js";
 import { art } from "./art.svelte.js";
 
+/* THE QUEUE WINDOW. Every update carries a bounded slice of the queue —
+   the current item and QUEUE_WINDOW either side — because media3's
+   SimpleBasePlayer.getState() wants a playlist and an index into it, and
+   the migration to it should find this data already running rather than
+   written the same day. Nothing reads it yet: MediaService ignores both
+   fields and the card is exactly the card it was.
+
+   TEN either side, twenty-one entries at most. One would cover the card's
+   own prev and next and nothing else; a queue LIST — which is what Auto
+   asks for — wants a screenful plus a scroll in either direction, and a
+   car's list pages about five rows, so ten is two of those each way. It is
+   a CAP, not a length: a two-thousand-track queue crosses the bridge as
+   twenty-one entries exactly like a twenty-one-track one does. Measured at
+   two kilobytes of JSON for twenty-one short titles and six with long ones
+   and real Android cover paths — against the thirty to eighty kilobytes of
+   base64 cover that already ride this same call once per record, and an
+   update is a track change, a transport change or the thirty-second top-up,
+   never a position tick. */
+const QUEUE_WINDOW = 10;
+
+/** A queue entry as the window carries it: what a media3 MediaItem is built
+ *  from, flat, with the cover NAMED rather than carried. The artist falls
+ *  back to the record's the way the bar and the sheet's rows do. */
+function windowItem(e) {
+    return {
+        id: e?.t?.id != null ? String(e.t.id) : null,
+        title: e?.t?.title || "",
+        artist: e?.t?.artist || e?.album?.artist || "",
+        album: e?.album?.title || "",
+        duration: e?.t?.duration_secs ?? null,
+        artKey: e?.album?.art_path || null,
+    };
+}
+
+/* What the card was last told about the queue around it. A PLAIN object,
+   not $state — nothing renders it, and push() must never make the bar's own
+   effects re-run — exactly as xfernote keeps its own mirror of the shade.
+   The walk reads it through __ss.card and holds it against what the plugin
+   logs on the far side of the bridge: a lock-screen card is not a thing a
+   page can be asked about, and this window is not even on the card. */
+export const mediacard = {
+    updates: 0,
+    window: 0,
+    index: -1,
+    art: 0,
+    at: "",
+};
+
 let lastArtKey = null;
 let timer = null;
 let bound = false;
@@ -17,6 +65,10 @@ async function push() {
     const t = player.track;
     if (!t) {
         lastArtKey = null;
+        mediacard.window = 0;
+        mediacard.index = -1;
+        mediacard.art = 0;
+        mediacard.at = "";
         await invoke("plugin:media-session|clear").catch(() => {});
         return;
     }
@@ -34,11 +86,27 @@ async function push() {
         hasNext: player.qi + 1 < player.queue.length || player.repeat > 0,
         hasPrev: true,
     };
+    /* the window, clamped at both ends. `at` is recomputed rather than
+       assumed to be QUEUE_WINDOW: at the top of the queue the slice starts
+       at 0 and the current item is the first entry, not the eleventh. */
+    const lo = Math.max(0, player.qi - QUEUE_WINDOW);
+    const win = player.qi < 0 ? [] : player.queue.slice(lo, player.qi + QUEUE_WINDOW + 1);
+    const at = player.qi - lo;
+    state.queue = win.map(windowItem);
+    state.queueIndex = at >= 0 && at < win.length ? at : null;
     /* the cover's bytes ride along once per record, and only once they exist */
     if (key && key !== lastArtKey && url && url.startsWith("data:image")) {
         state.art = url.slice(url.indexOf(",") + 1);
         lastArtKey = key;
     }
+    /* what was SENT, recorded before the call and whether or not it lands —
+       a desktop build and the mock both reject it, and neither is a reason
+       to forget what the window looked like */
+    mediacard.updates++;
+    mediacard.window = state.queue.length;
+    mediacard.index = state.queueIndex ?? -1;
+    mediacard.art = state.queue.filter((q) => !!q.artKey).length;
+    mediacard.at = state.queueIndex == null ? "" : (state.queue[state.queueIndex].title || "");
     await invoke("plugin:media-session|update", { state }).catch(() => {});
 }
 

tests/mobile/smoke.mjs

@@ -2485,6 +2485,23 @@ function mediaSessionState() {
     } catch { return { owner: "", state: "" }; }
 }
 
+/* The queue window the phone was last handed, as MediaSessionPlugin logged
+ *  it. Round forty-six put a bounded slice of the queue into every card
+ *  update — the current item and ten either side — and left MediaService
+ *  ignoring all of it, so there is nothing on any screen to read it back
+ *  from: logcat is the only place on the device it exists, one line per
+ *  update, and the LAST line is the state now. `at=` is last in the line
+ *  because a title may hold anything, spaces and `=` included. */
+function mediaWindow() {
+    try {
+        const out = adb("logcat", "-d", "-s", "SoulStream:D").toString();
+        const lines = out.split(/\r?\n/).filter((l) => l.includes("media window "));
+        const m = (lines[lines.length - 1] || "").match(/media window n=(\d+) i=(-?\d+) art=(\d+) at=(.*)$/);
+        if (!m) return { n: -1, i: -1, art: -1, at: "", lines: lines.length };
+        return { n: +m[1], i: +m[2], art: +m[3], at: m[4].trim(), lines: lines.length };
+    } catch (e) { return { n: -1, i: -1, art: -1, at: "ERR " + e.message, lines: 0 }; }
+}
+
 /* ── the phone, turned over: the two settings that do it ───────────────
  *  `settings get system accelerometer_rotation` and `user_rotation` are the
  *  only two things on the device this walk ever writes, and both are READ
@@ -3003,6 +3020,90 @@ async function main() {
             if (!keysOk) problems.push("EXPECT media-keys " + JSON.stringify({ keyPaused, keyPlaying }));
             console.log((keysOk ? "  ok  " : " FAIL ") + "media-keys".padEnd(16) + " " + JSON.stringify({ paused: keyPaused?.status, playing: keyPlaying?.status }));
 
+            /* ── the queue window, across the bridge ───────────────────
+               Round forty-six put a bounded window of the queue into every
+               card update — the current item and ten either side, each with
+               its id, title, artist, album, duration and art key, plus where
+               the current one sits among them — and left MediaService
+               ignoring every bit of it, because media3's
+               SimpleBasePlayer.getState() will want a playlist and an index,
+               and that data should be a round old before anything depends on
+               it. Which means every check above this one is blind to it: a
+               window that arrived empty, an index off by one, or a field the
+               bridge dropped because a key was spelled the wrong way would
+               each look exactly like success.
+
+               So: what the phone SENT — __ss.card, the mirror
+               media.svelte.js keeps beside xfernote's — against what the
+               plugin LOGGED on the far side, field for field, through a next
+               and a previous. Field for field is the point: a camelCase key
+               spelled snake_case arrives as nothing at all on both hops of
+               this bridge, with no error anywhere, so `art` (the only
+               camelCase field an entry has) is compared as well as the size
+               and the index.
+
+               ARMED TO A KNOWN ROW, index one, rather than to wherever the
+               walk left off. Three things fall out of that and every one of
+               them is load-bearing: there is a row on either side, so next()
+               steps instead of stopping at the end and prev() steps instead
+               of seeking to zero at the top; the track starts at zero, so
+               prev() is still a step two and a half seconds later rather
+               than the seek it becomes past five, and no track ends
+               underneath the three readings; and while the queue index is
+               under ten the window's own index IS the queue index, because
+               the slice starts at zero and nothing is clamped off the front
+               — which is what makes `got.i` comparable to `base` below
+               without recomputing the arithmetic here. Nothing is added,
+               removed or reordered: this is the queue the walk has been
+               playing all along, stepped one row forward and one row back
+               inside itself.
+
+               Not skipped when it cannot be armed: a queue shorter than
+               three rows, or repeat-one — under which next() seeks to zero
+               instead of stepping — FAILS, with `armed` in the message
+               saying which. A check that passes by not running is the thing
+               this walk is most careful never to be. */
+            const CARD = `(function () { var s = window.__ss; if (!s || !s.card || !s.player) return null; var c = s.card, p = s.player; return { u: c.updates, n: c.window, i: c.index, art: c.art, at: c.at, qi: p.qi, title: p.track ? p.track.title : '' }; })()`;
+            const armed = await cdp.evaluate(`(function () { var p = window.__ss && __ss.player; return p ? { n: p.queue.length, qi: p.qi, rep: p.repeat } : null; })()`);
+            const base = 1;
+            const room = (armed?.n ?? 0) >= 3 && armed.rep !== 2;
+            const wins = [];
+            if (room) {
+                await cdp.evaluate(`__ss.playQueueAt(${base})`);
+                await sleep(2500);
+                wins.push(["here", await cdp.evaluate(CARD), mediaWindow()]);
+                await cdp.evaluate(`__ss.next()`);
+                await sleep(2500);
+                wins.push(["next", await cdp.evaluate(CARD), mediaWindow()]);
+                await cdp.evaluate(`__ss.prev()`);
+                await sleep(2500);
+                wins.push(["prev", await cdp.evaluate(CARD), mediaWindow()]);
+            }
+            /* each reading: the window SENT is a real one — at least two
+               entries, the index inside it, and the entry AT the index is the
+               track on the bar, which is what an off-by-one fails even when
+               the counts agree — and what the plugin LOGGED is that same
+               window, field for field. */
+            const sane = ([, sent, got]) => !!sent && sent.n >= 2 && sent.i >= 0 && sent.i < sent.n
+                && !!sent.at && sent.at === sent.title
+                && got.n === sent.n && got.i === sent.i && got.art === sent.art
+                && got.at === String(sent.at).trim();
+            const winOk = room && wins.length === 3 && wins.every(sane)
+                /* and it MOVED. The queue index steps forward and back; the
+                   LOGGED window index steps with it, which is the assertion
+                   an off-by-one or a frozen index dies on. And each reading
+                   is a FRESH one at both ends — the phone pushed again
+                   (`u`) and the plugin logged again (`lines`) — because a
+                   prev() that lands back on the row it started from reads
+                   identically to a stale line otherwise. */
+                && wins[0][1].qi === base && wins[1][1].qi === base + 1 && wins[2][1].qi === base
+                && wins[0][2].i === base && wins[1][2].i === base + 1 && wins[2][2].i === base
+                && wins[1][1].u > wins[0][1].u && wins[2][1].u > wins[1][1].u
+                && wins[1][2].lines > wins[0][2].lines && wins[2][2].lines > wins[1][2].lines;
+            if (!winOk) problems.push("EXPECT queue-window " + JSON.stringify({ armed, base, wins }));
+            console.log((winOk ? "  ok  " : " FAIL ") + "queue-window".padEnd(16) + " "
+                + JSON.stringify({ armed, seen: wins.map(([k, sent, got]) => [k, sent && sent.qi, got.n, got.i, got.art, got.at]) }));
+
             const bgBefore = await cdp.evaluate(PLAYER);
             const posBefore = bgBefore?.pos ?? 0;
             cdp.close();