@torrent-tv/proxy 2.88.2 → 2.88.4

package/services/encode/segment-formats/mp4-boxes.js

@@ -244,12 +244,20 @@ export function readPresentationRanges(raw) {
         sample.start - previous.end <= precision + Number.EPSILON * Math.max(1, sample.end) * 8)) {
         previous.end = Math.max(previous.end, sample.end);
       } else merged.push({ ...sample });
+      // Preserve coded-frame timing separately from the presentation interval.
+      // Chromium joins an adjacent frame within twice the maximum coded-frame
+      // duration of the stream (SourceBufferRange::GetFudgeRoom), measured from
+      // the last frame's timestamp, not from the segment's nominal cut.
+      const range = merged.at(-1);
+      range.lastFrameStart = Math.max(range.lastFrameStart ?? sample.start, sample.start);
+      range.maxFrameDuration = Math.max(range.maxFrameDuration ?? 0, sample.end - sample.start);
     }
     // Movie edits quantize independent pieces to movie ticks. Preserve that
     // declared resolution at joins; it is not a playback buffer threshold.
     for (const range of merged) range.end += precision;
     const decoded = decodeRanges.get(id);
-    return { id, ranges: merged, kind: kinds.get(id), precision,
+    return { id, ranges: merged.map(({ start, end }) => ({ start, end })),
+      frameRanges: merged, kind: kinds.get(id), precision,
       decodeStart: Math.min(...decoded.map(({ start }) => start)),
       decodeEnd: Math.max(...decoded.map(({ end }) => end)) };
   });
@@ -280,8 +288,11 @@ export function continuePresentationRanges(ranges, nextRanges) {
 }
 
 function intersectPresentationTracks(tracks) {
-  return tracks.map(({ ranges }) => ranges).reduce((common, ranges) => common.flatMap((left) => ranges
-    .map((right) => ({ start: Math.max(left.start, right.start), end: Math.min(left.end, right.end) }))
+  return tracks.slice(1).map(({ ranges }) => ranges).reduce((common, ranges) => common.flatMap((left) => ranges
+    .map((right) => ({ start: Math.max(left.start, right.start), end: Math.min(left.end, right.end),
+      ...(left.joinEnd !== undefined || right.joinEnd !== undefined ? {
+        joinEnd: Math.min(left.joinEnd ?? left.end, right.joinEnd ?? right.end)
+      } : {}) }))
     .filter(({ start, end }) => end > start)), tracks[0]?.ranges ?? []);
 }
 
@@ -290,11 +301,15 @@ export function translatePresentationRanges(ranges, initBytes, timestampOffsetSe
   const tracks = presentationTracks.get(ranges);
   if (!tracks || !initBytes?.length || !Number.isFinite(timestampOffsetSeconds)) return ranges;
   const edits = readTrackEditOffsets(initBytes);
-  return intersectPresentationTracks(tracks.map(({ id, ranges: held }) => ({
+  return intersectPresentationTracks(tracks.map(({ id, ranges: held, frameRanges, kind }) => ({
     id,
-    ranges: held.map(({ start, end }) => ({
+    ranges: held.map(({ start, end }, index) => ({
       start: Math.max(0, start + timestampOffsetSeconds - (edits.get(id) ?? 0)),
-      end: end + timestampOffsetSeconds - (edits.get(id) ?? 0)
+      end: end + timestampOffsetSeconds - (edits.get(id) ?? 0),
+      ...(kind === "soun" ? {
+        joinEnd: frameRanges[index].lastFrameStart + 2 * frameRanges[index].maxFrameDuration +
+          timestampOffsetSeconds - (edits.get(id) ?? 0)
+      } : {})
     })).filter(({ start, end }) => end > start)
   })));
 }

package/services/encode/segment-formats/mpegts.js

@@ -1,134 +1,134 @@
-/**
- * @file MPEG-TS segment format — self-contained `.ts` segments, no init segment.
- *
- * This is the pre-fMP4 behaviour (recovered from the switch commit `dd1ce09`),
- * kept as a selectable alternative rather than deleted. Each segment carries
- * its own parameter sets and its own timestamps, so it is valid on its own —
- * there is no shared init segment that a seek-restart can invalidate, and
- * therefore none of the timeline problems the fMP4 path has to correct for.
- *
- * Trade-off vs fMP4: higher container overhead, and encoders that do not repeat
- * SPS/PPS (CM4 `h264_v4l2m2m`) emit segments after the first with no parameter
- * sets, which is exactly why fMP4 became the default.
- *
- * See {@link SegmentFormat} in `./index.js` for the interface contract.
- */
-
-// What a piece is called while it is still being written. See
-// `makingFileNameTemplate`.
-const MAKING_PATTERN = /^making-([0-9a-z]+)-(\d{5})\.ts$/;
-const SEGMENT_PATTERN = /^segment-(\d{5})\.ts$/;
-
-/**
- * @type {import("./index.js").SegmentFormat}
- */
-export const mpegtsFormat = {
-  id: "mpegts",
-  // No init segment: every `.ts` segment is self-describing.
-  initFileName: null,
-  initContentType: null,
-  segmentContentType: "video/mp2t",
-  playlistVersion: 3,
-
-  muxerArgs() {
-    return ["-hls_segment_filename", "segment-%05d.ts"];
-  },
-
-  /**
-   * Arguments for cutting at times we choose rather than times ffmpeg picks.
-   *
-   * The `hls` muxer takes only a target duration and finds its own cut points,
-   * which is why the playlist and the real segments drifted apart; the `segment`
-   * muxer takes the list. Self-contained segments make this straightforward
-   * here: no init segment to reconcile, so the only difference is the container
-   * and the file name template.
-   *
-   * @returns {string[]}
-   */
-  explicitTimesMuxerArgs() {
-    return ["-segment_format", "mpegts"];
-  },
-
-  /** The output path template for the `segment` muxer. */
-  /**
-   * The name ffmpeg is told to write a piece under WHILE IT IS MAKING IT.
-   *
-   * Not the name it is served under. A piece being written is not a piece, and
-   * under its final name it is indistinguishable from one — which is how half a
-   * segment came to be served: the proof of completeness was "the next file
-   * exists", true of one writer walking forward and false the moment two runs
-   * share an output, because the next file is then written by another process.
-   * Field 2026-09-08: `segment-00057.mp4` served at 2 268 361 bytes and then at
-   * 4 510 940, and the browser refused the whole one for the rest of the session.
-   *
-   * With a working name, the served name appears only when the encoder has said
-   * the piece is closed, and existence under it IS the proof — one rule, the
-   * same for every branch, and true whether or not our own process is alive.
-   *
-   * @returns {string}
-   */
-  makingFileNameTemplate(tag) {
-    return `making-${String(tag ?? "0").replace(/[^0-9a-z]/g, "")}-%05d.ts`;
-  },
-
-  /**
-   * Which run is writing this working name, or null when the name is not one.
-   *
-   * The tag is what makes clearing up after a dead run possible without
-   * guessing: several runs write into one directory, so "the unfinished pieces"
-   * is only a well-formed question per run. It used to be answered by looking
-   * for the highest SERVED name inside the stretch the ended run was given and
-   * judging its bytes — a guess, and under the naming rule above it would remove
-   * a complete piece somebody else had closed.
-   *
-   * @param {string} name
-   * @returns {string | null}
-   */
-  makingTagOf(name) {
-    const match = MAKING_PATTERN.exec(String(name ?? "").trim());
-    return match ? match[1] : null;
-  },
-
-  /**
-   * The name a piece just closed under a working name is served as, or null when
-   * the name is not one of ours.
-   *
-   * @param {string} makingName
-   * @returns {string | null}
-   */
-  servedNameOf(makingName) {
-    const match = MAKING_PATTERN.exec(String(makingName ?? "").trim());
-    return match ? this.segmentFileName(Number(match[2])) : null;
-  },
-
-  segmentFileNameTemplate() {
-    return "segment-%05d.ts";
-  },
-
-  playlistHeaderLines() {
-    return []; // no `#EXT-X-MAP`
-  },
-
-  segmentFileName(index) {
-    return `segment-${String(index).padStart(5, "0")}.ts`;
-  },
-
-  isSegmentFileName(fileName) {
-    return SEGMENT_PATTERN.test(fileName);
-  },
-
-  segmentIndexFromName(fileName) {
-    const match = SEGMENT_PATTERN.exec(fileName);
-    return match ? Number(match[1]) : -1;
-  },
-
-  /**
-   * MPEG-TS segments carry their own timestamps and need no correction, so they
-   * are streamed straight from disk (no read-into-memory step).
-   */
-  needsSegmentRewrite: false,
-
-  prepareSegmentBytes(bytes) {
-    return bytes;
-  }
-};
+/**
+ * @file MPEG-TS segment format — self-contained `.ts` segments, no init segment.
+ *
+ * This is the pre-fMP4 behaviour (recovered from the switch commit `dd1ce09`),
+ * kept as a selectable alternative rather than deleted. Each segment carries
+ * its own parameter sets and its own timestamps, so it is valid on its own —
+ * there is no shared init segment that a seek-restart can invalidate, and
+ * therefore none of the timeline problems the fMP4 path has to correct for.
+ *
+ * Trade-off vs fMP4: higher container overhead, and encoders that do not repeat
+ * SPS/PPS (CM4 `h264_v4l2m2m`) emit segments after the first with no parameter
+ * sets, which is exactly why fMP4 became the default.
+ *
+ * See {@link SegmentFormat} in `./index.js` for the interface contract.
+ */
+
+// What a piece is called while it is still being written. See
+// `makingFileNameTemplate`.
+const MAKING_PATTERN = /^making-([0-9a-z]+)-(\d{5})\.ts$/;
+const SEGMENT_PATTERN = /^segment-(\d{5})\.ts$/;
+
+/**
+ * @type {import("./index.js").SegmentFormat}
+ */
+export const mpegtsFormat = {
+  id: "mpegts",
+  // No init segment: every `.ts` segment is self-describing.
+  initFileName: null,
+  initContentType: null,
+  segmentContentType: "video/mp2t",
+  playlistVersion: 3,
+
+  muxerArgs() {
+    return ["-hls_segment_filename", "segment-%05d.ts"];
+  },
+
+  /**
+   * Arguments for cutting at times we choose rather than times ffmpeg picks.
+   *
+   * The `hls` muxer takes only a target duration and finds its own cut points,
+   * which is why the playlist and the real segments drifted apart; the `segment`
+   * muxer takes the list. Self-contained segments make this straightforward
+   * here: no init segment to reconcile, so the only difference is the container
+   * and the file name template.
+   *
+   * @returns {string[]}
+   */
+  explicitTimesMuxerArgs() {
+    return ["-segment_format", "mpegts"];
+  },
+
+  /** The output path template for the `segment` muxer. */
+  /**
+   * The name ffmpeg is told to write a piece under WHILE IT IS MAKING IT.
+   *
+   * Not the name it is served under. A piece being written is not a piece, and
+   * under its final name it is indistinguishable from one — which is how half a
+   * segment came to be served: the proof of completeness was "the next file
+   * exists", true of one writer walking forward and false the moment two runs
+   * share an output, because the next file is then written by another process.
+   * Field 2026-09-08: `segment-00057.mp4` served at 2 268 361 bytes and then at
+   * 4 510 940, and the browser refused the whole one for the rest of the session.
+   *
+   * With a working name, the served name appears only when the encoder has said
+   * the piece is closed, and existence under it IS the proof — one rule, the
+   * same for every branch, and true whether or not our own process is alive.
+   *
+   * @returns {string}
+   */
+  makingFileNameTemplate(tag) {
+    return `making-${String(tag ?? "0").replace(/[^0-9a-z]/g, "")}-%05d.ts`;
+  },
+
+  /**
+   * Which run is writing this working name, or null when the name is not one.
+   *
+   * The tag is what makes clearing up after a dead run possible without
+   * guessing: several runs write into one directory, so "the unfinished pieces"
+   * is only a well-formed question per run. It used to be answered by looking
+   * for the highest SERVED name inside the stretch the ended run was given and
+   * judging its bytes — a guess, and under the naming rule above it would remove
+   * a complete piece somebody else had closed.
+   *
+   * @param {string} name
+   * @returns {string | null}
+   */
+  makingTagOf(name) {
+    const match = MAKING_PATTERN.exec(String(name ?? "").trim());
+    return match ? match[1] : null;
+  },
+
+  /**
+   * The name a piece just closed under a working name is served as, or null when
+   * the name is not one of ours.
+   *
+   * @param {string} makingName
+   * @returns {string | null}
+   */
+  servedNameOf(makingName) {
+    const match = MAKING_PATTERN.exec(String(makingName ?? "").trim());
+    return match ? this.segmentFileName(Number(match[2])) : null;
+  },
+
+  segmentFileNameTemplate() {
+    return "segment-%05d.ts";
+  },
+
+  playlistHeaderLines() {
+    return []; // no `#EXT-X-MAP`
+  },
+
+  segmentFileName(index) {
+    return `segment-${String(index).padStart(5, "0")}.ts`;
+  },
+
+  isSegmentFileName(fileName) {
+    return SEGMENT_PATTERN.test(fileName);
+  },
+
+  segmentIndexFromName(fileName) {
+    const match = SEGMENT_PATTERN.exec(fileName);
+    return match ? Number(match[1]) : -1;
+  },
+
+  /**
+   * MPEG-TS segments carry their own timestamps and need no correction, so they
+   * are streamed straight from disk (no read-into-memory step).
+   */
+  needsSegmentRewrite: false,
+
+  prepareSegmentBytes(bytes) {
+    return bytes;
+  }
+};