@torrent-tv/proxy 2.87.21 → 2.88.1

package/test/subtitle-cue-walk.test.js

@@ -1,370 +1,370 @@
-/**
- * @file What comes out of a real cluster walk, end to end.
- *
- * This is the check the walk never had. Everything around it was covered — the
- * numbering a track is asked for by, the framing one cue's bytes carry, the
- * thread a pull is answered on — and the walk itself, six hundred lines of it,
- * had none: a file is built here with real clusters carrying real blocks, and
- * the cues that come back out are stated exactly. Two field defects landed on
- * this path in one week; a move of this code without a check that would catch a
- * regression is a change nobody can verify.
- *
- * The four things asserted are the four the walk exists for: the text of each
- * cue with its times, the found-order cursor a browser follows, the rule that
- * only already-downloaded clusters are read, and the rule that one walk fills
- * every track of the file.
- */
-
-import test from "node:test";
-import assert from "node:assert/strict";
-import { Readable } from "node:stream";
-
-import { cuesHeldFor, warmSubtitleCues, forgetSubtitles } from "../services/media/SubtitleCues.js";
-import { heldFileOver } from "./helpers/held-file.js";
-
-const ID_EBML = 0x1a45dfa3;
-const ID_SEGMENT = 0x18538067;
-const ID_SEEK_HEAD = 0x114d9b74;
-const ID_SEEK = 0x4dbb;
-const ID_SEEK_ID = 0x53ab;
-const ID_SEEK_POSITION = 0x53ac;
-const ID_INFO = 0x1549a966;
-const ID_TIMESTAMP_SCALE = 0x2ad7b1;
-const ID_TRACKS = 0x1654ae6b;
-const ID_TRACK_ENTRY = 0xae;
-const ID_TRACK_NUMBER = 0xd7;
-const ID_TRACK_TYPE = 0x83;
-const ID_CODEC_ID = 0x86;
-const ID_LANGUAGE = 0x22b59c;
-const ID_CUES = 0x1c53bb6b;
-const ID_CUE_POINT = 0xbb;
-const ID_CUE_TIME = 0xb3;
-const ID_CUE_TRACK_POSITIONS = 0xb7;
-const ID_CUE_TRACK = 0xf7;
-const ID_CUE_CLUSTER_POSITION = 0xf1;
-const ID_CLUSTER = 0x1f43b675;
-const ID_TIMESTAMP = 0xe7;
-const ID_BLOCK_GROUP = 0xa0;
-const ID_BLOCK = 0xa1;
-const ID_BLOCK_DURATION = 0x9b;
-
-function idBytes(id) {
-  const bytes = [];
-  let rest = id;
-  while (rest > 0) {
-    bytes.unshift(rest & 0xff);
-    rest = Math.floor(rest / 256);
-  }
-  return Buffer.from(bytes);
-}
-
-/** A four-byte size: a valid variable-length integer whatever the payload. */
-function sizeBytes(size) {
-  const buffer = Buffer.alloc(4);
-  buffer.writeUInt32BE(size, 0);
-  buffer[0] |= 0x10;
-  return buffer;
-}
-
-function element(id, payload) {
-  return Buffer.concat([idBytes(id), sizeBytes(payload.length), payload]);
-}
-
-function uintElement(id, value) {
-  const bytes = [];
-  let rest = value;
-  do {
-    bytes.unshift(rest & 0xff);
-    rest = Math.floor(rest / 256);
-  } while (rest > 0);
-  return element(id, Buffer.from(bytes));
-}
-
-function stringElement(id, value) {
-  return element(id, Buffer.from(value, "utf8"));
-}
-
-/** A fixed four bytes, so a table measured twice comes out the same length. */
-function uint32Element(id, value) {
-  const payload = Buffer.alloc(4);
-  payload.writeUInt32BE(value, 0);
-  return element(id, payload);
-}
-
-function trackEntry({ number, type, codecId, language }) {
-  return element(ID_TRACK_ENTRY, Buffer.concat([
-    uintElement(ID_TRACK_NUMBER, number),
-    uintElement(ID_TRACK_TYPE, type),
-    stringElement(ID_CODEC_ID, codecId),
-    stringElement(ID_LANGUAGE, language)
-  ]));
-}
-
-/**
- * One cue, as a BlockGroup so it can carry its own duration.
- *
- * @param {{ track: number, relativeTicks: number, durationTicks: number, text: string }} params
- * @returns {Buffer}
- */
-function cueBlock({ track, relativeTicks, durationTicks, text }) {
-  const header = Buffer.alloc(4);
-  header[0] = 0x80 | track; // track numbers under 128 are one byte, marker set
-  header.writeInt16BE(relativeTicks, 1);
-  header[3] = 0;
-  const block = element(ID_BLOCK, Buffer.concat([header, Buffer.from(text, "utf8")]));
-  return element(ID_BLOCK_GROUP, Buffer.concat([block, uintElement(ID_BLOCK_DURATION, durationTicks)]));
-}
-
-/**
- * A file with two text tracks and two clusters, each carrying a line for each
- * track.
- *
- * Track 2 is `S_TEXT/UTF8`, whose block payload IS the text. Track 3 is
- * `S_TEXT/ASS`, whose payload is the dialogue row without its `Dialogue:`
- * header — eight comma-separated fields and then the words, which is what
- * Matroska states and what the framing rule has to strip.
- *
- * @returns {{ file: Buffer, clusterAt: number[] }}
- */
-function buildFile() {
-  const info = element(ID_INFO, uintElement(ID_TIMESTAMP_SCALE, 1_000_000));
-  const tracks = element(ID_TRACKS, Buffer.concat([
-    trackEntry({ number: 1, type: 1, codecId: "V_MPEG4/ISO/AVC", language: "und" }),
-    trackEntry({ number: 2, type: 17, codecId: "S_TEXT/UTF8", language: "eng" }),
-    trackEntry({ number: 3, type: 17, codecId: "S_TEXT/ASS", language: "rus" })
-  ]));
-
-  const clusterOne = element(ID_CLUSTER, Buffer.concat([
-    uintElement(ID_TIMESTAMP, 1000),
-    cueBlock({ track: 2, relativeTicks: 0, durationTicks: 2000, text: "First English line" }),
-    cueBlock({
-      track: 3,
-      relativeTicks: 500,
-      durationTicks: 1500,
-      // ReadOrder, Layer, Style, Name, MarginL, MarginR, MarginV, Effect,
-      // then the text — the Dialogue row without its header and its timestamps.
-      text: "1,0,Default,,0000,0000,0000,,Первая русская строка"
-    })
-  ]));
-  const clusterTwo = element(ID_CLUSTER, Buffer.concat([
-    uintElement(ID_TIMESTAMP, 10_000),
-    cueBlock({ track: 2, relativeTicks: 0, durationTicks: 1000, text: "Second English line" }),
-    cueBlock({
-      track: 3,
-      relativeTicks: 250,
-      durationTicks: 1000,
-      text: "2,0,Default,,0000,0000,0000,,Вторая, с запятой"
-    })
-  ]));
-
-  const cuesWith = (firstAt, secondAt) => element(ID_CUES, Buffer.concat([
-    element(ID_CUE_POINT, Buffer.concat([
-      uintElement(ID_CUE_TIME, 1000),
-      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
-        uintElement(ID_CUE_TRACK, 2),
-        uint32Element(ID_CUE_CLUSTER_POSITION, firstAt)
-      ])),
-      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
-        uintElement(ID_CUE_TRACK, 3),
-        uint32Element(ID_CUE_CLUSTER_POSITION, firstAt)
-      ]))
-    ])),
-    element(ID_CUE_POINT, Buffer.concat([
-      uintElement(ID_CUE_TIME, 10_000),
-      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
-        uintElement(ID_CUE_TRACK, 2),
-        uint32Element(ID_CUE_CLUSTER_POSITION, secondAt)
-      ])),
-      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
-        uintElement(ID_CUE_TRACK, 3),
-        uint32Element(ID_CUE_CLUSTER_POSITION, secondAt)
-      ]))
-    ]))
-  ]));
-
-  const seekEntry = (targetId, position) => element(ID_SEEK, Buffer.concat([
-    element(ID_SEEK_ID, idBytes(targetId)),
-    uint32Element(ID_SEEK_POSITION, position)
-  ]));
-  const seekHeadWith = (infoAt, tracksAt, cuesAt) => element(ID_SEEK_HEAD, Buffer.concat([
-    seekEntry(ID_INFO, infoAt),
-    seekEntry(ID_TRACKS, tracksAt),
-    seekEntry(ID_CUES, cuesAt)
-  ]));
-
-  const headLength = seekHeadWith(0, 0, 0).length;
-  const infoAt = headLength;
-  const tracksAt = infoAt + info.length;
-  const cuesAt = tracksAt + tracks.length;
-  const firstClusterAt = cuesAt + cuesWith(0, 0).length;
-  const secondClusterAt = firstClusterAt + clusterOne.length;
-
-  const segmentPayload = Buffer.concat([
-    seekHeadWith(infoAt, tracksAt, cuesAt),
-    info,
-    tracks,
-    cuesWith(firstClusterAt, secondClusterAt),
-    clusterOne,
-    clusterTwo
-  ]);
-  const ebml = element(ID_EBML, Buffer.from([0x42, 0x86, 0x81, 0x01]));
-  const segment = element(ID_SEGMENT, segmentPayload);
-  const segmentDataOffset = ebml.length + segment.length - segmentPayload.length;
-  return {
-    file: Buffer.concat([ebml, segment]),
-    clusterAt: [segmentDataOffset + firstClusterAt, segmentDataOffset + secondClusterAt]
-  };
-}
-
-/**
- * A torrent over one file, holding whichever pieces the caller says.
- *
- * @param {Buffer} bytes
- * @param {(index: number) => boolean} [holds] - Whether a piece is downloaded.
- * @returns {{ torrent: object, reads: Array<{ start: number, end: number }>, pieceLength: number }}
- */
-function torrentOver(bytes, holds = () => true) {
-  const pieceLength = 64;
-  const reads = [];
-  const file = {
-    name: "film.mkv",
-    length: bytes.length,
-    offset: 0,
-    createReadStream({ start = 0, end = bytes.length - 1 } = {}) {
-      reads.push({ start, end });
-      return Readable.from((async function* chunks() {
-        await new Promise((resolve) => setImmediate(resolve));
-        yield bytes.subarray(start, end + 1);
-      })());
-    }
-  };
-  return {
-    reads,
-    pieceLength,
-    torrent: {
-      name: "film",
-      pieceLength,
-      bitfield: { get: (index) => holds(index) },
-      files: [file]
-    }
-  };
-}
-
-test("a walk of a downloaded file gives up every cue, unframed, with its times", async () => {
-  const { file } = buildFile();
-  const { torrent } = torrentOver(file);
-  const sourceKey = "a".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    const plain = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
-    assert.deepEqual(
-      plain.cues.map((cue) => [cue.startSeconds, cue.endSeconds, cue.text]),
-      [
-        [1, 3, "First English line"],
-        [10, 11, "Second English line"]
-      ]
-    );
-
-    const ass = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 3);
-    assert.deepEqual(
-      ass.cues.map((cue) => [cue.startSeconds, cue.endSeconds, cue.text]),
-      [
-        [1.5, 3, "Первая русская строка"],
-        [10.25, 11.25, "Вторая, с запятой"]
-      ],
-      "the eight fields Matroska puts before the text are the container's framing, not the words"
-    );
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
-
-test("the cursor is the order cues were FOUND, and it never repeats", async () => {
-  const { file } = buildFile();
-  const { torrent } = torrentOver(file);
-  const sourceKey = "b".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    const held = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
-    const cursors = held.cues.map((cue) => cue.seq);
-    assert.deepEqual(cursors, [1, 2], "two cues found, in the order they were read");
-    assert.equal(new Set(cursors).size, cursors.length, "no cue shares a cursor with another");
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
-
-test("a cluster whose bytes are not downloaded is left for next time", async () => {
-  const { file, clusterAt } = buildFile();
-  // Everything except the pieces the SECOND cluster sits in.
-  const { torrent, pieceLength } = torrentOver(file, (index) => index * pieceLength < clusterAt[1]);
-  const sourceKey = "c".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    const held = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
-    assert.deepEqual(
-      held.cues.map((cue) => cue.text),
-      ["First English line"],
-      "the downloaded cluster is read and the other is not — nothing is asked of the swarm"
-    );
-    assert.equal(held.coveredClusters, 1);
-    assert.equal(held.indexedClusters, 2, "the file states two, and one is not here yet");
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
-
-test("one walk fills every track, and a second call reads no cluster again", async () => {
-  const { file, clusterAt } = buildFile();
-  const { torrent, reads } = torrentOver(file);
-  const sourceKey = "d".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
-    const clusterReads = reads.filter((range) => clusterAt.includes(range.start)).length;
-
-    const other = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 3);
-    assert.equal(other.cues.length, 2, "the other track was filled by the same walk");
-    assert.equal(
-      reads.filter((range) => clusterAt.includes(range.start)).length,
-      clusterReads,
-      "asking for the second track reads no cluster a second time"
-    );
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
-
-test("the warm pass reports what is new, by the number the browser knows", async () => {
-  const { file } = buildFile();
-  const { torrent } = torrentOver(file);
-  const sourceKey = "e".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    const first = await warmSubtitleCues(heldFileOver(torrent, 0, sourceKey));
-    assert.deepEqual(
-      first.map((entry) => [entry.trackIndex, entry.cues.length, entry.language]).sort(),
-      [[0, 2, "eng"], [1, 2, "rus"]].sort(),
-      "both text tracks gained two cues, numbered as ffmpeg numbers them"
-    );
-
-    const again = await warmSubtitleCues(heldFileOver(torrent, 0, sourceKey));
-    assert.deepEqual(again, [], "nothing is new the second time round");
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
-
-test("a torrent that cannot say which pieces it holds is refused, not answered emptily", async () => {
-  const { file } = buildFile();
-  const { torrent } = torrentOver(file);
-  const sourceKey = "f".repeat(40);
-  forgetSubtitles(sourceKey);
-  try {
-    const held = await cuesHeldFor({ ...heldFileOver(torrent, 0, sourceKey), heldRanges: null }, 2);
-    assert.deepEqual(held.cues, []);
-    assert.equal(held.track, null, "the answer says nothing was read, not that there is nothing");
-  } finally {
-    forgetSubtitles(sourceKey);
-  }
-});
+/**
+ * @file What comes out of a real cluster walk, end to end.
+ *
+ * This is the check the walk never had. Everything around it was covered — the
+ * numbering a track is asked for by, the framing one cue's bytes carry, the
+ * thread a pull is answered on — and the walk itself, six hundred lines of it,
+ * had none: a file is built here with real clusters carrying real blocks, and
+ * the cues that come back out are stated exactly. Two field defects landed on
+ * this path in one week; a move of this code without a check that would catch a
+ * regression is a change nobody can verify.
+ *
+ * The four things asserted are the four the walk exists for: the text of each
+ * cue with its times, the found-order cursor a browser follows, the rule that
+ * only already-downloaded clusters are read, and the rule that one walk fills
+ * every track of the file.
+ */
+
+import test from "node:test";
+import assert from "node:assert/strict";
+import { Readable } from "node:stream";
+
+import { cuesHeldFor, warmSubtitleCues, forgetSubtitles } from "../services/media/SubtitleCues.js";
+import { heldFileOver } from "./helpers/held-file.js";
+
+const ID_EBML = 0x1a45dfa3;
+const ID_SEGMENT = 0x18538067;
+const ID_SEEK_HEAD = 0x114d9b74;
+const ID_SEEK = 0x4dbb;
+const ID_SEEK_ID = 0x53ab;
+const ID_SEEK_POSITION = 0x53ac;
+const ID_INFO = 0x1549a966;
+const ID_TIMESTAMP_SCALE = 0x2ad7b1;
+const ID_TRACKS = 0x1654ae6b;
+const ID_TRACK_ENTRY = 0xae;
+const ID_TRACK_NUMBER = 0xd7;
+const ID_TRACK_TYPE = 0x83;
+const ID_CODEC_ID = 0x86;
+const ID_LANGUAGE = 0x22b59c;
+const ID_CUES = 0x1c53bb6b;
+const ID_CUE_POINT = 0xbb;
+const ID_CUE_TIME = 0xb3;
+const ID_CUE_TRACK_POSITIONS = 0xb7;
+const ID_CUE_TRACK = 0xf7;
+const ID_CUE_CLUSTER_POSITION = 0xf1;
+const ID_CLUSTER = 0x1f43b675;
+const ID_TIMESTAMP = 0xe7;
+const ID_BLOCK_GROUP = 0xa0;
+const ID_BLOCK = 0xa1;
+const ID_BLOCK_DURATION = 0x9b;
+
+function idBytes(id) {
+  const bytes = [];
+  let rest = id;
+  while (rest > 0) {
+    bytes.unshift(rest & 0xff);
+    rest = Math.floor(rest / 256);
+  }
+  return Buffer.from(bytes);
+}
+
+/** A four-byte size: a valid variable-length integer whatever the payload. */
+function sizeBytes(size) {
+  const buffer = Buffer.alloc(4);
+  buffer.writeUInt32BE(size, 0);
+  buffer[0] |= 0x10;
+  return buffer;
+}
+
+function element(id, payload) {
+  return Buffer.concat([idBytes(id), sizeBytes(payload.length), payload]);
+}
+
+function uintElement(id, value) {
+  const bytes = [];
+  let rest = value;
+  do {
+    bytes.unshift(rest & 0xff);
+    rest = Math.floor(rest / 256);
+  } while (rest > 0);
+  return element(id, Buffer.from(bytes));
+}
+
+function stringElement(id, value) {
+  return element(id, Buffer.from(value, "utf8"));
+}
+
+/** A fixed four bytes, so a table measured twice comes out the same length. */
+function uint32Element(id, value) {
+  const payload = Buffer.alloc(4);
+  payload.writeUInt32BE(value, 0);
+  return element(id, payload);
+}
+
+function trackEntry({ number, type, codecId, language }) {
+  return element(ID_TRACK_ENTRY, Buffer.concat([
+    uintElement(ID_TRACK_NUMBER, number),
+    uintElement(ID_TRACK_TYPE, type),
+    stringElement(ID_CODEC_ID, codecId),
+    stringElement(ID_LANGUAGE, language)
+  ]));
+}
+
+/**
+ * One cue, as a BlockGroup so it can carry its own duration.
+ *
+ * @param {{ track: number, relativeTicks: number, durationTicks: number, text: string }} params
+ * @returns {Buffer}
+ */
+function cueBlock({ track, relativeTicks, durationTicks, text }) {
+  const header = Buffer.alloc(4);
+  header[0] = 0x80 | track; // track numbers under 128 are one byte, marker set
+  header.writeInt16BE(relativeTicks, 1);
+  header[3] = 0;
+  const block = element(ID_BLOCK, Buffer.concat([header, Buffer.from(text, "utf8")]));
+  return element(ID_BLOCK_GROUP, Buffer.concat([block, uintElement(ID_BLOCK_DURATION, durationTicks)]));
+}
+
+/**
+ * A file with two text tracks and two clusters, each carrying a line for each
+ * track.
+ *
+ * Track 2 is `S_TEXT/UTF8`, whose block payload IS the text. Track 3 is
+ * `S_TEXT/ASS`, whose payload is the dialogue row without its `Dialogue:`
+ * header — eight comma-separated fields and then the words, which is what
+ * Matroska states and what the framing rule has to strip.
+ *
+ * @returns {{ file: Buffer, clusterAt: number[] }}
+ */
+function buildFile() {
+  const info = element(ID_INFO, uintElement(ID_TIMESTAMP_SCALE, 1_000_000));
+  const tracks = element(ID_TRACKS, Buffer.concat([
+    trackEntry({ number: 1, type: 1, codecId: "V_MPEG4/ISO/AVC", language: "und" }),
+    trackEntry({ number: 2, type: 17, codecId: "S_TEXT/UTF8", language: "eng" }),
+    trackEntry({ number: 3, type: 17, codecId: "S_TEXT/ASS", language: "rus" })
+  ]));
+
+  const clusterOne = element(ID_CLUSTER, Buffer.concat([
+    uintElement(ID_TIMESTAMP, 1000),
+    cueBlock({ track: 2, relativeTicks: 0, durationTicks: 2000, text: "First English line" }),
+    cueBlock({
+      track: 3,
+      relativeTicks: 500,
+      durationTicks: 1500,
+      // ReadOrder, Layer, Style, Name, MarginL, MarginR, MarginV, Effect,
+      // then the text — the Dialogue row without its header and its timestamps.
+      text: "1,0,Default,,0000,0000,0000,,Первая русская строка"
+    })
+  ]));
+  const clusterTwo = element(ID_CLUSTER, Buffer.concat([
+    uintElement(ID_TIMESTAMP, 10_000),
+    cueBlock({ track: 2, relativeTicks: 0, durationTicks: 1000, text: "Second English line" }),
+    cueBlock({
+      track: 3,
+      relativeTicks: 250,
+      durationTicks: 1000,
+      text: "2,0,Default,,0000,0000,0000,,Вторая, с запятой"
+    })
+  ]));
+
+  const cuesWith = (firstAt, secondAt) => element(ID_CUES, Buffer.concat([
+    element(ID_CUE_POINT, Buffer.concat([
+      uintElement(ID_CUE_TIME, 1000),
+      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
+        uintElement(ID_CUE_TRACK, 2),
+        uint32Element(ID_CUE_CLUSTER_POSITION, firstAt)
+      ])),
+      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
+        uintElement(ID_CUE_TRACK, 3),
+        uint32Element(ID_CUE_CLUSTER_POSITION, firstAt)
+      ]))
+    ])),
+    element(ID_CUE_POINT, Buffer.concat([
+      uintElement(ID_CUE_TIME, 10_000),
+      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
+        uintElement(ID_CUE_TRACK, 2),
+        uint32Element(ID_CUE_CLUSTER_POSITION, secondAt)
+      ])),
+      element(ID_CUE_TRACK_POSITIONS, Buffer.concat([
+        uintElement(ID_CUE_TRACK, 3),
+        uint32Element(ID_CUE_CLUSTER_POSITION, secondAt)
+      ]))
+    ]))
+  ]));
+
+  const seekEntry = (targetId, position) => element(ID_SEEK, Buffer.concat([
+    element(ID_SEEK_ID, idBytes(targetId)),
+    uint32Element(ID_SEEK_POSITION, position)
+  ]));
+  const seekHeadWith = (infoAt, tracksAt, cuesAt) => element(ID_SEEK_HEAD, Buffer.concat([
+    seekEntry(ID_INFO, infoAt),
+    seekEntry(ID_TRACKS, tracksAt),
+    seekEntry(ID_CUES, cuesAt)
+  ]));
+
+  const headLength = seekHeadWith(0, 0, 0).length;
+  const infoAt = headLength;
+  const tracksAt = infoAt + info.length;
+  const cuesAt = tracksAt + tracks.length;
+  const firstClusterAt = cuesAt + cuesWith(0, 0).length;
+  const secondClusterAt = firstClusterAt + clusterOne.length;
+
+  const segmentPayload = Buffer.concat([
+    seekHeadWith(infoAt, tracksAt, cuesAt),
+    info,
+    tracks,
+    cuesWith(firstClusterAt, secondClusterAt),
+    clusterOne,
+    clusterTwo
+  ]);
+  const ebml = element(ID_EBML, Buffer.from([0x42, 0x86, 0x81, 0x01]));
+  const segment = element(ID_SEGMENT, segmentPayload);
+  const segmentDataOffset = ebml.length + segment.length - segmentPayload.length;
+  return {
+    file: Buffer.concat([ebml, segment]),
+    clusterAt: [segmentDataOffset + firstClusterAt, segmentDataOffset + secondClusterAt]
+  };
+}
+
+/**
+ * A torrent over one file, holding whichever pieces the caller says.
+ *
+ * @param {Buffer} bytes
+ * @param {(index: number) => boolean} [holds] - Whether a piece is downloaded.
+ * @returns {{ torrent: object, reads: Array<{ start: number, end: number }>, pieceLength: number }}
+ */
+function torrentOver(bytes, holds = () => true) {
+  const pieceLength = 64;
+  const reads = [];
+  const file = {
+    name: "film.mkv",
+    length: bytes.length,
+    offset: 0,
+    createReadStream({ start = 0, end = bytes.length - 1 } = {}) {
+      reads.push({ start, end });
+      return Readable.from((async function* chunks() {
+        await new Promise((resolve) => setImmediate(resolve));
+        yield bytes.subarray(start, end + 1);
+      })());
+    }
+  };
+  return {
+    reads,
+    pieceLength,
+    torrent: {
+      name: "film",
+      pieceLength,
+      bitfield: { get: (index) => holds(index) },
+      files: [file]
+    }
+  };
+}
+
+test("a walk of a downloaded file gives up every cue, unframed, with its times", async () => {
+  const { file } = buildFile();
+  const { torrent } = torrentOver(file);
+  const sourceKey = "a".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    const plain = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
+    assert.deepEqual(
+      plain.cues.map((cue) => [cue.startSeconds, cue.endSeconds, cue.text]),
+      [
+        [1, 3, "First English line"],
+        [10, 11, "Second English line"]
+      ]
+    );
+
+    const ass = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 3);
+    assert.deepEqual(
+      ass.cues.map((cue) => [cue.startSeconds, cue.endSeconds, cue.text]),
+      [
+        [1.5, 3, "Первая русская строка"],
+        [10.25, 11.25, "Вторая, с запятой"]
+      ],
+      "the eight fields Matroska puts before the text are the container's framing, not the words"
+    );
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});
+
+test("the cursor is the order cues were FOUND, and it never repeats", async () => {
+  const { file } = buildFile();
+  const { torrent } = torrentOver(file);
+  const sourceKey = "b".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    const held = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
+    const cursors = held.cues.map((cue) => cue.seq);
+    assert.deepEqual(cursors, [1, 2], "two cues found, in the order they were read");
+    assert.equal(new Set(cursors).size, cursors.length, "no cue shares a cursor with another");
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});
+
+test("a cluster whose bytes are not downloaded is left for next time", async () => {
+  const { file, clusterAt } = buildFile();
+  // Everything except the pieces the SECOND cluster sits in.
+  const { torrent, pieceLength } = torrentOver(file, (index) => index * pieceLength < clusterAt[1]);
+  const sourceKey = "c".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    const held = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
+    assert.deepEqual(
+      held.cues.map((cue) => cue.text),
+      ["First English line"],
+      "the downloaded cluster is read and the other is not — nothing is asked of the swarm"
+    );
+    assert.equal(held.coveredClusters, 1);
+    assert.equal(held.indexedClusters, 2, "the file states two, and one is not here yet");
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});
+
+test("one walk fills every track, and a second call reads no cluster again", async () => {
+  const { file, clusterAt } = buildFile();
+  const { torrent, reads } = torrentOver(file);
+  const sourceKey = "d".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 2);
+    const clusterReads = reads.filter((range) => clusterAt.includes(range.start)).length;
+
+    const other = await cuesHeldFor(heldFileOver(torrent, 0, sourceKey), 3);
+    assert.equal(other.cues.length, 2, "the other track was filled by the same walk");
+    assert.equal(
+      reads.filter((range) => clusterAt.includes(range.start)).length,
+      clusterReads,
+      "asking for the second track reads no cluster a second time"
+    );
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});
+
+test("the warm pass reports what is new, by the number the browser knows", async () => {
+  const { file } = buildFile();
+  const { torrent } = torrentOver(file);
+  const sourceKey = "e".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    const first = await warmSubtitleCues(heldFileOver(torrent, 0, sourceKey));
+    assert.deepEqual(
+      first.map((entry) => [entry.trackIndex, entry.cues.length, entry.language]).sort(),
+      [[0, 2, "eng"], [1, 2, "rus"]].sort(),
+      "both text tracks gained two cues, numbered as ffmpeg numbers them"
+    );
+
+    const again = await warmSubtitleCues(heldFileOver(torrent, 0, sourceKey));
+    assert.deepEqual(again, [], "nothing is new the second time round");
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});
+
+test("a torrent that cannot say which pieces it holds is refused, not answered emptily", async () => {
+  const { file } = buildFile();
+  const { torrent } = torrentOver(file);
+  const sourceKey = "f".repeat(40);
+  forgetSubtitles(sourceKey);
+  try {
+    const held = await cuesHeldFor({ ...heldFileOver(torrent, 0, sourceKey), heldRanges: null }, 2);
+    assert.deepEqual(held.cues, []);
+    assert.equal(held.track, null, "the answer says nothing was read, not that there is nothing");
+  } finally {
+    forgetSubtitles(sourceKey);
+  }
+});