feat(moq-mux): generic verbatim MPEG-TS carriage (mpegts catalog section)#1815
Conversation
Generalize the SCTE-35-specific path into a generic per-PID verbatim lane: any PMT-listed elementary stream we don't decode is carried byte-for-byte, described in a new `ts` catalog section (PID map + stream configs). SCTE-35 becomes one instance (stream_type 0x86, section framing). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughThe PR removes the dedicated 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches✨ Simplify code
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rs/moq-mux/src/container/ts/catalog.rs`:
- Around line 126-130: The Ext struct needs to be marked with the
#[non_exhaustive] attribute to make the extension schema additive and prevent
breaking changes when new optional fields are added in the future. Add
#[non_exhaustive] as an attribute directly above the pub struct Ext definition
on line 126, before the existing derive attribute. Since Ext already derives
Default, the requirement for a constructor is already satisfied, allowing
consumers to construct the struct while keeping it safe for future extension.
In `@rs/moq-mux/src/container/ts/export.rs`:
- Around line 541-546: The Kind::Verbatim variant with framing set to Pes is
currently passing verbatim PES payloads that may have non-audio stream types
(such as teletext or DVB subtitles) to write_pes, which incorrectly applies an
audio stream ID to all non-video tracks. This violates the verbatim elementary
stream contract by rewriting the stream ID. Thread the original PES stream_id
through the import/catalog/export pipeline, or alternatively derive the correct
non-audio stream ID for PMT private-data stream types before calling write_pes.
Update the handling of Kind::Verbatim with framing: catalog::Framing::Pes to use
the correct stream ID instead of defaulting to audio, ensuring the PES header
preserves the original stream type classification.
- Around line 463-465: The condition at line 465 that detects stream_type 0x86
for the CUEI descriptor addition does not account for the framing type. The
issue is that the same Kind::Verbatim with stream_type 0x86 can be packetized as
either sections or PES depending on the framing mode, but the current code only
checks the stream type. When framing is PES, the write_frame function will
packetize the payload as PES, yet the PMT advertises SCTE-35 section signaling,
causing re-import to route the PID incorrectly. Gate the CUEI descriptor
addition by also checking that the framing is not PES (i.e., verify the framing
is section-based) before adding the descriptor alongside the existing
stream_type check.
- Line 443: The call to StreamType::from_u8(*stream_type) at line 443 will fail
when encountering unknown or proprietary stream types (such as teletext 0x10)
that are not defined in the mpeg2ts enum, causing Verbatim track exports to
fail. Either validate and filter unknown stream types before they reach this
point upstream in the export logic, or modify the handling at the
StreamType::from_u8 call site to gracefully handle unknown types by either using
a fallback stream type, wrapping the conversion in a more lenient parser, or
explicitly handling the error case instead of propagating it as a fatal error.
In `@rs/moq-mux/src/container/ts/import_test.rs`:
- Line 298: The SCTE-35 stream selection in the line with
snap.ts.streams.keys().next().expect("scte35 track") is relying on map iteration
order, which is unreliable when new stream types are added to the fixture.
Instead of taking the first stream from the map, filter through the streams to
find the one with stream_type matching 0x86 (the SCTE-35 identifier), and select
that stream by its type rather than by map order position.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 42f89a16-46fc-4fbb-bbb3-1ec221e302f6
📒 Files selected for processing (8)
rs/moq-mux/src/container/source.rsrs/moq-mux/src/container/ts/catalog.rsrs/moq-mux/src/container/ts/export.rsrs/moq-mux/src/container/ts/export_test.rsrs/moq-mux/src/container/ts/import.rsrs/moq-mux/src/container/ts/import_test.rsrs/moq-mux/src/container/ts/mod.rsrs/moq-mux/src/container/ts/scte35.rs
💤 Files with no reviewable changes (1)
- rs/moq-mux/src/container/ts/scte35.rs
Reshape the ts section into a single per-track map keyed by name: Track{pid,
descriptors, verbatim?}. Captures media-track PIDs and PMT ES descriptors
(audio language, ...) and program-level descriptors, not just verbatim streams.
Rename the catalog key/types to mpegts. SCTE-35/teletext/AC-3 verbatim carriage
unchanged; media language descriptors and program_info now survive the round-trip.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
kixelated
changed the title
Only derive the program-level SCTE-35 CUEI registration for a section-framed 0x86 verbatim stream; a PES-framed 0x86 must not advertise section signaling. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…stream_type - Mark the mpegts Ext extension #[non_exhaustive] so new sections stay additive. - On export, skip (with a warning) a verbatim stream whose recorded stream_type isn't a known TS stream type rather than aborting the whole export; round-tripped streams always parse, this guards a hand-built catalog. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Addressed the review feedback in
Deferred (wants a maintainer call): preserve the original PES (Written by Claude) |
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
rs/moq-mux/src/container/ts/import.rs (1)
208-219:⚠️ Potential issue | 🟠 Major | ⚡ Quick winClear stale PID routes before accepting a new PMT mapping.
Line 208 checks
sectionsbeforestreams. If a later PMT remaps a section PID to PES or a codec, the oldsectionsentry keeps intercepting it; in the base-catalog path, the oldStream::Ignoredmakesensure_streamreturn before the new route is installed. Finish and remove stale section routes, and allowStream::Ignoredto be replaced before the early return.Proposed fix
- if self.streams.contains_key(&pid) { - return Ok(()); - } + if let Some(mut section) = self.sections.remove(&pid.as_u16()) { + section.finish()?; + self.pending.remove(&pid); + } + if matches!(self.streams.get(&pid), Some(Stream::Ignored)) { + self.streams.remove(&pid); + } else if self.streams.contains_key(&pid) { + return Ok(()); + }Also applies to: 306-307, 400-412
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rs/moq-mux/src/container/ts/import.rs` around lines 208 - 219, The issue is that stale PID routes are not being cleared when a PMT remaps a PID from a section route to a PES or codec route. The current logic at line 208 checks self.sections.get_mut(&pid) first and continues to use the old section route, preventing the new route from being installed. Additionally, in the base-catalog path, the old Stream::Ignored prevents ensure_stream from installing the new route before the early return. To fix this, before processing the packet with the current section route, finish and remove the stale section entry if the PID has been remapped in a new PMT, and ensure that Stream::Ignored entries can be replaced by finishing the stale route before checking for Stream::Ignored and making the early continue. Apply the same logic at the other locations mentioned (lines 306-307 and 400-412).
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@rs/moq-mux/src/container/ts/import.rs`:
- Around line 208-219: The issue is that stale PID routes are not being cleared
when a PMT remaps a PID from a section route to a PES or codec route. The
current logic at line 208 checks self.sections.get_mut(&pid) first and continues
to use the old section route, preventing the new route from being installed.
Additionally, in the base-catalog path, the old Stream::Ignored prevents
ensure_stream from installing the new route before the early return. To fix
this, before processing the packet with the current section route, finish and
remove the stale section entry if the PID has been remapped in a new PMT, and
ensure that Stream::Ignored entries can be replaced by finishing the stale route
before checking for Stream::Ignored and making the early continue. Apply the
same logic at the other locations mentioned (lines 306-307 and 400-412).
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 1caf070e-4468-4755-9fdf-8a60b92b093c
📒 Files selected for processing (7)
rs/moq-mux/src/codec/legacy.rsrs/moq-mux/src/container/source.rsrs/moq-mux/src/container/ts/catalog.rsrs/moq-mux/src/container/ts/export.rsrs/moq-mux/src/container/ts/export_test.rsrs/moq-mux/src/container/ts/import.rsrs/moq-mux/src/container/ts/import_test.rs
🚧 Files skipped from review as they are similar to previous changes (3)
- rs/moq-mux/src/container/source.rs
- rs/moq-mux/src/container/ts/export_test.rs
- rs/moq-mux/src/container/ts/export.rs
Capture the original PES stream_id on import and re-emit it on export instead of relabeling every non-video stream as audio (0xC0). DVB teletext, DVB subtitles, and DVB AC-3 ride private_stream_1 (0xBD); rewriting them to audio is rejected by strict broadcast demuxers and TR 101 290 analyzers, and mis-frames stream_ids whose PES-header rules differ. Recorded as mpegts Verbatim.stream_id; absent (a non-TS source) falls back to private_stream_1. Validated on a real CNN feed: AC-3 and teletext both preserve 0xBD across the round-trip. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A later PMT can remap a previously section-framed PID to a PES codec/verbatim stream. ensure_stream now finishes and drops the stale sections entry first, so the PID isn't intercepted as a section forever while the new stream starves. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Follow-up in
(Written by Claude) |
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
rs/moq-mux/src/container/ts/export.rs (1)
349-372:⚠️ Potential issue | 🟠 Major | ⚡ Quick winRefresh existing verbatim metadata before PSI is built.
update_catalogcan see multiple snapshots beforebuild_psi(), but an already-created verbatim track is skipped. Since import first registersVerbatim::new(...)withstream_id: Noneand fills the original PES id later, export can latchNoneand then fall back toprivate_stream_1at Lines 622-624 instead of preserving the real stream id.Suggested fix
let Some(verbatim) = &track.verbatim else { continue; }; - if self.tracks.contains_key(name) { - continue; - } // Skip (don't abort) a verbatim stream whose recorded stream_type isn't a // known TS stream type, so it can't be announced in the PMT. Round-tripped // streams always parse (the type came from the import's PMT); this guards a // hand-built catalog. Mirrors the `active` filter above. if StreamType::from_u8(verbatim.stream_type).is_err() { @@ stream_type: verbatim.stream_type, framing: verbatim.framing, stream_id: verbatim.stream_id, }; + if let Some(existing) = self.tracks.get_mut(name) { + existing.kind = kind; + existing.descriptors = track.descriptors.clone(); + continue; + } let source = ExportSource::for_stream(&self.broadcast, name, verbatim, self.latency)?;Also applies to: 619-632
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rs/moq-mux/src/container/ts/export.rs` around lines 349 - 372, The issue is that existing verbatim tracks are being skipped when a track with the same name already exists in self.tracks. This prevents updating the track's metadata (particularly stream_id) when a new snapshot is processed. Instead of unconditionally continuing when self.tracks.contains_key(name) is true, you need to check if the existing track is a verbatim track and update its metadata with the new verbatim information from the current snapshot rather than skipping it. This ensures that when import initially registers a Verbatim with stream_id set to None and later updates it with the real stream_id, the export will preserve the updated stream_id instead of falling back to private_stream_1 at the fallback logic.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@rs/moq-mux/src/container/ts/export.rs`:
- Around line 349-372: The issue is that existing verbatim tracks are being
skipped when a track with the same name already exists in self.tracks. This
prevents updating the track's metadata (particularly stream_id) when a new
snapshot is processed. Instead of unconditionally continuing when
self.tracks.contains_key(name) is true, you need to check if the existing track
is a verbatim track and update its metadata with the new verbatim information
from the current snapshot rather than skipping it. This ensures that when import
initially registers a Verbatim with stream_id set to None and later updates it
with the real stream_id, the export will preserve the updated stream_id instead
of falling back to private_stream_1 at the fallback logic.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: edb96715-8dd3-4575-a9e1-16e9a6fab276
📒 Files selected for processing (3)
rs/moq-mux/src/container/ts/catalog.rsrs/moq-mux/src/container/ts/export.rsrs/moq-mux/src/container/ts/import.rs
🚧 Files skipped from review as they are similar to previous changes (2)
- rs/moq-mux/src/container/ts/catalog.rs
- rs/moq-mux/src/container/ts/import.rs
| /// The underlying MoQ track producer (for its name, etc.). | ||
| pub fn track(&self) -> &moq_net::TrackProducer { |
There was a problem hiding this comment.
Can we not expose the Producer? Just the name() would be fine.
| use crate::catalog::hang::CatalogExt; | ||
|
|
||
| /// Serialize [`Bytes`] as a base64 string in the catalog JSON. | ||
| mod base64_bytes { |
There was a problem hiding this comment.
serde_from or serde_as instead?
| /// PMT program-level descriptors (`program_info`), carried verbatim. Export | ||
| /// re-emits these; the SCTE-35 'CUEI' registration is derived when absent. | ||
| #[serde(default, skip_serializing_if = "Vec::is_empty")] | ||
| pub program_descriptors: Vec<Descriptor>, |
| pub struct Verbatim { | ||
| /// PMT `stream_type` to re-announce (0x86 SCTE-35, 0x06 private PES, 0x05 | ||
| /// private sections, ...). | ||
| pub stream_type: u8, |
There was a problem hiding this comment.
Is it worth having an enum (with unknown(u8) or no?
|
|
||
| /// How the verbatim bytes are timestamp-framed as MoQ frames. | ||
| #[serde(default)] | ||
| pub container: hang::catalog::Container, |
There was a problem hiding this comment.
Just hard-code hang::legacy::Container? Seems dumb to support CMAF.
| /// Keep this stable per catalog: an importer samples support once at | ||
| /// construction, so a result that flips between `Some` and `None` mid-stream | ||
| /// would disable verbatim carriage or fail. | ||
| fn mpegts_mut(&mut self) -> Option<&mut Mpegts>; |
There was a problem hiding this comment.
Why mut? Is this needed?
| /// | ||
| /// Implement this for an application extension to compose MPEG-TS carriage with | ||
| /// additional sections. | ||
| pub trait Catalog: CatalogExt { |
There was a problem hiding this comment.
And why do we need another trait?
| // known TS stream type, so it can't be announced in the PMT. Round-tripped | ||
| // streams always parse (the type came from the import's PMT); this guards a | ||
| // hand-built catalog. Mirrors the `active` filter above. | ||
| if StreamType::from_u8(verbatim.stream_type).is_err() { |
There was a problem hiding this comment.
Do we really need to gate based on StreamType?
| /// (PAT), [`PMT_PID`], `0x1FFF` (null), and any already-assigned PID are avoided | ||
| /// so cross-references survive when possible. | ||
| fn alloc_pid(&self, name: &str, tracks: &std::collections::BTreeMap<String, catalog::Track>) -> u16 { | ||
| let mut used: std::collections::HashSet<u16> = self.tracks.values().map(|t| t.pid).collect(); |
There was a problem hiding this comment.
IDK If we need to construct a HashSet every time, just iterate over the tracks.
- Fix rustdoc broken intra-doc link to `ts` in ts/mod.rs (the CI failure). - Catalog: use serde_with serde_as/Base64 for Descriptor.data instead of a hand-rolled base64 module (matches rs/hang); drop Verbatim.container (always Legacy-framed) and hard-code it. - legacy::Import exposes name() instead of the Producer. - Export: refresh existing tracks' PID/descriptors/kind on each catalog snapshot until the PMT is built, instead of latching the first (partial) snapshot. The importer fills PIDs, ES descriptors, and PES stream_ids across several publishes, so latching dropped them; this fixes media-PID/descriptor and verbatim stream_id preservation (flagged by review + CodeRabbit). PID assignment is a two-pass (prefer original, then fill free) with no per-call HashSet. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Thanks — addressed your comments + a self-review in Changed as suggested:
Correctness fix from the review (you + CodeRabbit both flagged the class): export was latching per-track metadata from the first catalog snapshot. The importer fills a track's PID, ES descriptors, and PES Kept (with reasoning — let me know if you disagree):
Known minor edges I left (rare, happy to fix if you want): a mid-session PMT version change that alters a track's ES descriptors or adds CUEI to an already-non-empty (Written by Claude) |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rs/moq-mux/src/container/ts/export.rs`:
- Around line 269-277: The `verbatim_ok` closure silently skips verbatim tracks
with unsupported stream types without notifying the operator. Add a warning log
statement that triggers when a verbatim track is skipped due to an unsupported
stream_type value (when `StreamType::from_u8(v.stream_type)` returns an error).
The warning should provide context about which track or stream type value is
being dropped and why, ensuring operators have visibility into what catalog
entries are being excluded from the PMT announcement.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 0f450662-4415-4e0d-a57e-6b7245ba88d5
⛔ Files ignored due to path filters (1)
Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (7)
rs/moq-mux/Cargo.tomlrs/moq-mux/src/codec/legacy.rsrs/moq-mux/src/container/source.rsrs/moq-mux/src/container/ts/catalog.rsrs/moq-mux/src/container/ts/export.rsrs/moq-mux/src/container/ts/import.rsrs/moq-mux/src/container/ts/mod.rs
🚧 Files skipped from review as they are similar to previous changes (2)
- rs/moq-mux/src/container/ts/mod.rs
- rs/moq-mux/src/container/ts/import.rs
| // The desired track set: media renditions plus carriable verbatim streams. A | ||
| // verbatim stream_type that isn't a known TS stream type can't be announced in | ||
| // the PMT, so it's skipped (not aborted); round-tripped streams always parse, | ||
| // this only guards a hand-built catalog. | ||
| let verbatim_ok = |t: &catalog::Track| { | ||
| t.verbatim | ||
| .as_ref() | ||
| .is_some_and(|v| StreamType::from_u8(v.stream_type).is_ok()) | ||
| }; |
There was a problem hiding this comment.
Warn when skipping unsupported verbatim stream types.
verbatim_ok silently drops verbatim tracks whose stream_type cannot be converted by mpeg2ts. That avoids aborting, but it still loses an explicitly cataloged stream with no operator signal.
Suggested fix
for (name, track) in mpegts.tracks.iter() {
if verbatim_ok(track) {
active.insert(name.clone(), ());
+ } else if let Some(verbatim) = &track.verbatim {
+ tracing::warn!(
+ track = %name,
+ stream_type = verbatim.stream_type,
+ "skipping verbatim TS track with unsupported stream_type"
+ );
}
}Also applies to: 285-287
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@rs/moq-mux/src/container/ts/export.rs` around lines 269 - 277, The
`verbatim_ok` closure silently skips verbatim tracks with unsupported stream
types without notifying the operator. Add a warning log statement that triggers
when a verbatim track is skipped due to an unsupported stream_type value (when
`StreamType::from_u8(v.stream_type)` returns an error). The warning should
provide context about which track or stream type value is being dropped and why,
ensuring operators have visibility into what catalog entries are being excluded
from the PMT announcement.
A verbatim stream imported from TS always has a known stream_type (mpeg2ts parsed it from the PMT), so the from_u8 eligibility gate was unreachable for real data and only guarded a hand-built catalog. Drop it; build_psi's existing from_u8 conversion is the backstop (a malformed hand-built catalog errors there). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Follow-up (
(Written by Claude) |
|
Re-read at 09a1d50 round-trip is narrower than carriage. Import carries undecoded stream_type verbatim, but export skips anything StreamType::from_u8() doesn't recognize, so an exotic stream type imports but drops on export, for example KLV crosses the bridge but canot be remuxed. |
|
Oh reviews in cascade, too fast for me :) |
|
bradhards' KLV will work transparently with this |
dev advanced 3 commits while this PR was open (#1845 PTS-exposing TS Export + PCR-paced SRT egress, #1847 moq-lite-05 wire sync, #1819 always-on hardware encoders). Re-merged dev and reconciled two conflicts: - nix/overlay.nix: kept this PR's cargo+sccache CI direction (dev's tip still carries the crane `moqChecks`; the cargo switch lives on main, #1821). - rs/moq-mux/src/container/ts/export.rs: combined dev's #1845 Frame-returning `Export::next` (PTS/keyframe-stamped) with this PR's generic `catalog::Catalog` trait (main's #1815 mpegts rename). Updated the moq-cli drain helper to read `frame.payload` and refreshed two stale `scte35::Ext` comments. Verified: moq-mux (279), moq-cli, moq-srt tests pass; workspace clippy -D warnings and fmt clean (excluding the crates that need libva/gstreamer system libs unavailable in this sandbox: moq-video/libmoq/moq-boy/moq-gst); JS type-checks and biome clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQ9o9paZnRLpYRgbyjUwFV
3 participants
Summary
Generalizes the SCTE-35-specific MPEG-TS path into a generic per-PID verbatim lane. Any PMT-listed elementary stream the demuxer doesn't decode (SCTE-35, DVB AC-3, teletext/subtitles, private data, ...) is now carried byte-for-byte instead of being dropped, described in a new
mpegtscatalog section. SCTE-35 becomes one instance of this.This is the "tier 2" lane from the discussion under #1799: keep full demux for browser/ABR, but stop discarding the ancillary streams a contribution feed carries — and faithfully reconstruct them (PIDs, descriptors, PES stream_ids, program signaling) on export.
Catalog: the
mpegtssectionOne unified per-track map keyed by MoQ track name, plus the program-level descriptors:
Decoded media keeps its codec config in the base
video/audiosections; thempegtsentry records only its MPEG-TS identity. Empty section serializes to{}, so a non-TS broadcast is byte-identical.Import/Exportare generic over a smallCatalogtrait so the media-only()catalog and theExt(mpegts) catalog share one code path.Behaviour
stream_id.mpegts.tracks: preserves PIDs, re-emits captured ES + program descriptors verbatim (deriving the SCTE-35 'CUEI' registration only when none were captured), re-emits each verbatim PES under its originalstream_id(falling back toprivate_stream_1), and repacketizes perframing. It refreshes track metadata across catalog snapshots until the PMT is built, so PID/descriptors/stream_id filled by the importer across several publishes aren't dropped.Public API (moq-mux library)
container::ts::catalog::{Mpegts, Track, Verbatim, Framing, Descriptor, Ext, Catalog (trait)}.container::ts::scte35module;Export::with_scte35→Export::with_ts.Per [Branch Targeting], a breaking library-API change would normally target
dev; onmainat the maintainer's request (the SCTE-35 surface is recent and unreleased in this form).Validated on a real CNN 1080i contribution feed
CNNiEMEA2.ts, 80 MB prefix, import → export → re-import is identical by PID:(Video decode for this feed's two-PPS H.264 requires #1812; the two PRs are independent and compose cleanly.)
This is faithful at the elementary-stream + PMT level (content + signaling), not byte-for-byte at the transport level — PCR cadence, continuity counters, stuffing, and SI tables outside the PMT are regenerated. Literal byte fidelity (and CAS, exact PCR for TR 101 290) is the opaque whole-mux lane (Option B in #1799), future work.
Out of scope / follow-ups
moq pub/sub.Test plan
cargo test -p moq-mux(237), incl.private_pes_carried_verbatim,carries_all_sections_verbatim,mpegtscatalog (de)serialization, SCTE-35 fixture round-tripscargo clippy -p moq-mux --all-targets -- -D warnings,cargo fmt --check, andcargo doc -D warningscleanmoq-cli/moq-relay/hang/libmoqbuildCross-package sync
No
js/hang/js/netmirror: thempegtssection is a moq-mux-side extension; browser consumers ignore the unknown section. Basehangcatalog and themoq-netwire are unchanged.(Written by Claude)