[kixelated]

Every PID of an imported program, other than the null PID, now takes one declared route: decoded, whole-PES pesPacket, verbatim section, captured PSI/SI, or not carried and counted in ts::stats. Finishes the quest quest/m1/ts-reliability/stream-routes.md and deletes it. Builds on moq-dev/moq#6208's route retirement and moq-dev/moq#6371's shared parser.

What changed

• MPEG-1/2 video (0x01/0x02) rides as an opaque pesPacket track instead of being discarded, and keeps its clock role (its PTS still paces the program; stats class video). A Hang player still sees no rendition it cannot decode. A PES carrying a sequence_header_code opens a group and the pictures after it join it, so a joiner or a dropped group never lands on a dependent picture. Pictures before the first sequence header, or after a lost one until the next, are dropped. A header split across two PES opens the group at the PES it begins in: a video PES ending in part of a start code waits for the next PES on its stream before its group is decided; a clock break, a PMT revision, or the end of input ends the wait (one PES of extra latency, only then).
• DSM-CC (0x0A to 0x0D) routes as section-framed verbatim, instead of failing every PES start code as damaged.
• 0x86 without CUEI, and any other listed stream the import cannot carry (no mpegts section, no PCR, bad Opus config), counts one refused with a warning, its row's track empty.
• Unclaimed PIDs (CAT, EMM/ECM, ATSC PSIP, TSDT, DVB RST/DIT/SIT, anything no table lists) get a Class::Data row with an empty track, refused counting packets, and one warning. ts/unclaimed.rs holds the counts; each owner decides what is claimed and when counting starts:
  • A single Import counts a packet on a PID its own routes do not claim (PMT, streams, clock, sections, PAT, captured SI, null), once a PMT is in and the PAT lists only this program. --program N of an MPTS counts none, since it cannot tell another program's PIDs from strays.
  • --program all: Programs counts a packet on a PID no program reads (Programs::readers, from the shared parser), the PAT's PMT PIDs, the SI PIDs and the null PID aside, once for the multiplex. Counting waits until every program the PAT lists has started, or 0.5 s of the first program's clock has passed without its PMT (TR 101 290 PMT_error_2).
• TS-scrambled packets on a carried stream count as refused and lose the unit in progress (a PES or a partial section), rather than parse damage. The route stays, so the stream resumes when clear.
• Export lists program 0 on PID 0x0010 in the PAT whenever the catalog carries a NIT. A NIT that comes or goes later re-versions the PAT beside the PMT at the next version, so the PAT's version never runs ahead of the PMT's.
• Export scheduling: a placed PES with a decode time is due by that decode time (DTS, else PTS) rather than its placement. Its first byte still waits for placement, but a whole MPEG-2 picture no longer holds the multiplex from placement to its last byte. Without this, the MPEG-2 round trip starved MP2 audio past its decoder buffer at 10 Mb/s. The program tables leading a placed PES go out by its placement: they ride one lane, and with the picture's decode time as their deadline they held later tables, leading audio, past theirs (missed a decode deadline on PID 4096 in just test ts --mpeg2 after merging main).
• A PID keeps one row through a PMT that lists it, drops it while its packets continue, and lists it again: the row moves between its route and the unclaimed count with its counters intact, under one Import or Programs.
• Docs: LATM (0x11) already rode pesPacket; the Import docs now say so. doc/bin/cli.md and doc/concept/stats.md describe the routes and the counts.
• Quests: the ts-reliability profile rows this covers are landed; pat-lifecycle notes what a program a later PAT adds does to the count.

Public API and wire impact

• No Rust signatures change.
• Meaning change in ts::stats::Snapshot::streams: rows now exist for PIDs that publish no track (empty track), and refused also counts unclaimed packets and TS-scrambled ones. MPEG-1/2 video rows are pesPacket with class video and units counted as delivered PES. To migrate, filter on a non-empty track where you read the rows as published tracks.
• Catalog: MPEG-1/2 video and DSM-CC streams now appear in the mpegts section where they were missing before. draft-lcurley-moq-mpegts (-01, in progress) lists them among the pesPacket and section-framed examples, groups MPEG-1/2 video pesPacket frames from one sequence header to the next instead of one PES per group (the PES a header begins in opens the group, even when the next PES completes it), and requires the PAT network entry when a NIT is carried.

Tests

• ts/routes_test.rs: MPEG-2 whole-PES carriage, its groups (a mid-GOP join and a scrambled loss each wait for the next sequence header) and an export round trip, each DSM-CC type, 0x86 without CUEI, LATM, CAT + PSIP + unlisted PID (field by field) + nulls + scrambled ES, 16 scrambled packets wrapping a section PID's counter, media before the PMT, a two-program MPTS under Programs with program 2 media ahead of its PMT, the same MPTS under --program 1, a PMT that never arrives, a PMT on another program's PID, and one PID's lifecycle (unlisted, listed, dropped while flowing, listed again) under both owners.
• routes_test::mpeg2_sequence_header_split_across_pes_opens_a_group: a sequence header split at each of its three inner boundaries opens its group at the PES it begins in, a split picture start code opens none, and a whole PES ending in part of a start code still goes out at a flagged clock break (on a dedicated PCR PID) and at end of input.
• export_test::service_layer_survives_roundtrip checks the PAT's network entry and that the re-import counts no row for the NIT PID.
• schedule: a placed PES yields to units due before it decodes, and the tables leading one go out by its placement (without the fix it fails with the harness's error).
• test/ts --mpeg2: Main Profile at Main Level MPEG-2 video round trip, FFmpeg decode compared picture by picture (aligned by suffix, so a repeated picture is fine), holes.py, and the T-STD model, which gains H.262 parameters (Tables 8-13/8-14, leak method with Rbx = min(1.05 Res, Rmax) at High-1440 and High; the vbv_delay method and a zero VBV buffer are refused, covered in compliance.test.py). Wired into .github/workflows/ts.yml.
• benches/ts_programs.rs (from moq-dev/moq#6371) also sweeps 16 and 256 stray PIDs per frame against the same packets as null stuffing, which skips the accounting. Locally, on a loaded machine: 1 program 1.41 GiB/s stray against 1.70 GiB/s null at 256 strays, and no slope with program count (16 programs: 720 against 626 MiB/s).

Decisions

Recorded from the open questions on this PR, each taking the recommendation:

1. DSM-CC: all four types 0x0A to 0x0D route as sections ✅ (GStreamer's reading; 13818-6 carries 0x0A and 0x0C in DSM-CC sections too), over only 0x0B and 0x0D (TSDuck's StreamTypeIsSection).
2. refused units: one per refused listed stream, one per packet for unclaimed and scrambled PIDs ✅, over one unit for both.
3. Scrambled packets keep the route, so a stream resumes when clear ✅, over refusing the stream outright.
4. Placed PES scheduling by decode time applies to every placed PES ✅ (DVB AC-3 and synchronous KLV arms still pass), over scoping it to MPEG-1/2 video.
5. MPEG-1/2 video grouping, decided by the maintainer on 2026-10-23 (Codex finding): group by sequence header and amend the draft's pesPacket section for 0x01/0x02 ✅, over keeping per-PES groups and documenting resync at the next sequence header.

Revisited after moq-dev/moq#6371, per moq-dev/moq#6334; agent choices, confirmed by the maintainer on 2026-10-25:

6. Programs classifies through Programs::readers instead of re-reading the PAT and every PMT, and the refcounted claims.rs is gone ✅. The earlier Codex finding on a PMT arriving on another program's PID is now covered by perf(mux): parse an MPTS once for every program #6371's Programs, which starts a program only from a PMT on the PID its PAT entry names; the test stays.
7. A program a later PAT adds, which Programs does not start yet, holds the count back for 0.5 s like any missing PMT, then its elementary PIDs count as unclaimed, since nothing reads them ✅. This replaces the earlier choice that its PIDs stay claimed, which needed a second PMT parse; pat-lifecycle notes it.
8. A single Import counts once its PMT is in, with no 0.5 s lapse: before a PMT it has no clock to time one ✅.
9. The tables leading a placed PES are due by its placement ✅, over reverting decision 4 (which brings back the audio starvation it fixed).
10. MPEG-1/2 sequence header split across two PES, decided by the maintainer on 2026-10-25 (Codex finding): hold back a PES ending in a partial start code until the next PES on that PID, and flush it at end of input ✅. A clock break flushes it too (Codex finding), since no PES on the new clock can complete a header begun on the old.

Checks

just check passes (rs, js, drafts, markdown, gh, quest, shell) after merging main, cargo test -p moq-mux --all-features --lib container::ts passes (653; the whole moq-mux lib, 1754, after the second merge of main), cargo clippy -p moq-mux --all-features --all-targets is clean, python3 test/ts/compliance.test.py passes, and just test ts --mpeg2 passes (473 of 500 pictures decoded as the source's, 0 joined pictures skipped, T-STD clean). Not rerun locally: just test ts --klv --strict (CI runs it).

Follow-ups

• PAT changes already owns starting a program a later PAT adds, which would make its PIDs claimed again.

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[kixelated]

Status: draft until moq-dev/moq#6208 lands.

Remaining: the quest's PID lifecycle test (unlisted, added by a PMT revision, removed while its packets continue, added again, counters checked at each step, under both one Import and Programs). It needs #6208's route retirement and its fixture::Mux::pmt_at. Verified on a local trial merge of #6208 into this branch: the test passes for both owners and the rest of container::ts passes (588 tests). The merge conflicts in import.rs only where #6208 introduced section_framed(): that function takes the DSM-CC types (0x0A to 0x0D), and handle_pat keeps both the claims feed and #6208's identity check. Until then, Programs claims by PMT listing while an import's routes stay sticky, so under --program all a PMT that drops a still-flowing PID shows an empty track on a row the import still carries.

Checks: cargo test -p moq-mux --all-features passes (1594 lib tests). just test ts --mpeg2 and just test ts --klv --strict pass locally, and so does just drafts check. In just check, js, drafts, rs and drill pass. Its flake module fails building the github-runner image (unshare: write failed /proc/self/uid_map), which is the sandbox and unrelated to this change.

Open decisions (recommendation first):

1. DSM-CC routing: all four types 0x0A to 0x0D as sections (GStreamer's reading, and what this PR does), or only 0x0B and 0x0D (TSDuck's StreamTypeIsSection). I recommend all four, since 13818-6 carries 0x0A and 0x0C in DSM-CC sections too.
2. Units of refused: one per refused stream (a listed one), and one per packet (unclaimed and scrambled). I recommend keeping this split: a refused stream is one event, and a dropped PID's packets are the volume an operator moves to --passthrough.
3. Scrambled packets keep the route, so a stream that goes clear resumes. The alternative is refusing the stream outright. I recommend keeping the route.
4. Placed PES scheduling: a placed PES with a PTS is now due by its DTS (else PTS) instead of its placement. This was needed for the MPEG-2 round trip, and it also applies to DVB AC-3 and synchronous KLV, whose arm still passes. The alternative is to scope it to MPEG-1/2 video only. I recommend keeping it general.
5. A program a later PAT adds has no import under Programs, but Programs still claims its PIDs, so they are not counted. That is left to the pat-lifecycle quest.

(Written by Claude Opus 5.5)

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 2 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 9784ec47-c775-43ec-b813-12eeadbc8185

Commits

Reviewing files that changed from the base of the PR and between 28a947e and aa1fc67.

Files selected for processing: 26

• .github/workflows/ts.yml
• doc/bin/cli.md
• doc/concept/stats.md
• drafts/draft-lcurley-moq-mpegts.md
• quest/m1/ts-reliability/README.md
• quest/m1/ts-reliability/pat-lifecycle.md
• quest/m1/ts-reliability/stream-routes.md
• rs/justfile
• rs/moq-mux/benches/ts_programs.rs
• rs/moq-mux/src/container/ts/catalog.rs
• rs/moq-mux/src/container/ts/export.rs
• rs/moq-mux/src/container/ts/export_test.rs
• rs/moq-mux/src/container/ts/import.rs
• rs/moq-mux/src/container/ts/import_test.rs
• rs/moq-mux/src/container/ts/mod.rs
• rs/moq-mux/src/container/ts/programs.rs
• rs/moq-mux/src/container/ts/routes_test.rs
• rs/moq-mux/src/container/ts/schedule.rs
• rs/moq-mux/src/container/ts/stats.rs
• rs/moq-mux/src/container/ts/unclaimed.rs
• test/justfile
• test/ts/README.md
• test/ts/compliance.py
• test/ts/compliance.test.py
• test/ts/mpeg2.py
• test/ts/run.sh

Walkthrough

The change updates MPEG-TS import routing, packet accounting, export tables, and scheduling. MPEG-1/2 video, LATM AAC, and DSM-CC streams receive updated handling. The importer reports refused packets for unclaimed PIDs and scrambled carried streams. The exporter updates PAT contents for carried NIT data and uses DTS when available. The TS test tools add MPEG-2 generation, compliance modeling, and capture grading.

Priority: Normal

Merge Risk: Moderate · up to 28a94

MPEG-2 validation can give incorrect results for some streams or reject an unchanged capture. Correct the grading model and alignment before merging; update the stale documentation alongside those fixes.

Pre-merge checks: 5

Check name	Status	Explanation
Title check		The title "fix(mux): give every TS PID a declared route" matches the main change. The changeset assigns each imported MPEG-TS PID to a declared route (decoded, whole-PES pesPacket, verbatim section, c…
Docstring Coverage		Docstring coverage is 82.91% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 117 functions across 15 files. (9 skipped: …
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.
Description check		The pull request description directly matches the changeset. It describes MPEG-TS PID routing, MPEG-2 and DSM-CC handling, refusal accounting, export behavior, scheduling, tests, documentation, and qu…

Full details: Title check

Explanation

The title "fix(mux): give every TS PID a declared route" matches the main change. The changeset assigns each imported MPEG-TS PID to a declared route (decoded, whole-PES pesPacket, verbatim section, captured PSI/SI, or counted as refused in ts::stats). The title is short, specific, and uses the conventional commit form. It does not cover every detail, such as export scheduling and PAT network entries, which is expected.

Full details: Docstring Coverage

Explanation

Docstring coverage is 82.91% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 117 functions across 15 files. (9 skipped: 9 unsupported.)

Full details: Description check

Explanation

The pull request description directly matches the changeset. It describes MPEG-TS PID routing, MPEG-2 and DSM-CC handling, refusal accounting, export behavior, scheduling, tests, documentation, and quest updates.

Finishing Touches

✨ Simplify code

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T00:15:55.855177Z	aa1fc67	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: db75614f77

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fc80562b8e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[kixelated]

Maintainer decision: group MPEG-1/2 video by sequence header. A PES carrying 0x000001B3 opens a group and the rest join it; amend the draft's pesPacket section to match. This still waits for moq-dev/moq#6371 to land first.

(Written by Claude Opus 5.5)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8c60315cf9

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: 8c60315

Confirmed the independent P2 clock-discontinuity finding: #6252 (comment) . The new last-PES lookahead filter at rs/moq-mux/src/container/ts/import.rs:2454–2458 remains active when timebase_break settles/releases the old timeline. A complete MPEG-1/2 PES ending in a partial start-code prefix remains held despite a valid timestamp, failing Stream::placed and aborting import. Resolve lookahead before draining the old clock, as EOF/route retirement do; cover a flagged PCR discontinuity, including a dedicated PCR PID.

The bounded one-PES lookahead addresses the earlier split-header finding, but needs this lifecycle boundary handled. The scrambled-section reset remains fixed; harness updates address the earlier VBV-model, zero-buffer, rate and repeated-picture findings.

Verification: static update/surrounding-code review, accounting for merged base changes since the last published review. No tests, builds or FFmpeg checks run.

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: 89e9976

The clock-discontinuity finding #6252 (comment) is fixed. timebase_break ends pending start-code lookahead before settling/releasing the old timeline, letting complete held PES drain before discontinuity. Regression covers dedicated PCR PID, partial-prefix tail and new-clock restart.

No new actionable bugs found. Reusing end-of-lookahead is narrow and avoids extra buffering complexity.

Verification: static delta and surrounding code, accounting for base merge. No tests or builds run.

[kixelated]

Landing summary for 89e9976:

• Fixed the Codex finding on split sequence headers the way the maintainer decided on 2026-10-25. An MPEG-1/2 video PES that ends in part of a start code waits for the next PES on its PID, and the PES the header begins in opens the group. The wait also ends at end of input, at a PMT revision, and at a clock break. The clock-break case comes from a later Codex finding. The draft's pesPacket grouping rule now covers a split header.
• Fixed CodeRabbit's T-STD harness findings. The harness refuses the vbv_delay method and a zero VBV buffer. Rbx is min(1.05 Res, Rmax) at High-1440 and High. Decoded pictures are aligned by suffix, so a repeated picture no longer misplaces the run.
• The maintainer confirmed decisions 6-9 on 2026-10-25, and the PR body records that.
• Merged main twice. just check, the whole moq-mux lib (1754), clippy, compliance.test.py, and just test ts --mpeg2 pass. quest check --base origin/main passes.

(Written by Claude Opus 5.5)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fae6c81588

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: aa1fc67

The independent MPEG-video recovery finding #6252 (comment) is fixed: export.rs now uses the existing video classification for Arrival::sync, so whole-PES MPEG-1/2 video waits for a keyframe after a late drop while audio/data retain their behavior. The new regression checks dependent pictures are omitted until the next sequence header. The previously reviewed clock-discontinuity fix remains unchanged.

No new actionable bugs found. Reusing Kind::class is a suitably narrow solution without adding another recovery mechanism.

Verification: static delta, classification and jitter-recovery inspection, accounting for intervening base merges. No tests or builds run.

[kixelated]

Landing summary for aa1fc67, after the merge queue dropped 89e9976 for a conflict with main:

• Merged main twice. The conflicts with moq-dev/moq#6757 (route retirement, SI on the program clock) kept both sides: refuse retires the section or stream's track timeline and still leaves MPEG-1/2 video as a clock-only route; an SI PID's unplaced row names .si, any other row without a track stays empty, as this PR defines. service_layer_survives_roundtrip now accepts the NIT PID's .si unplaced row, since SI ahead of the first PCR is unplaced after fix(mux): bound TS import state by retiring ended routes and admitting SI on the program clock #6757, but still requires that the PID is never counted as unclaimed (refused 0). The second merge only touched the ts-reliability quest README, which keeps main's new validation-green entry.
• Fixed the Codex finding on export jitter recovery: whole MPEG-1/2 video now resumes only at a sequence header after a late drop. Covered by export_resumes_whole_mpeg2_video_at_a_sequence_header, which failed before the fix.
• Reviews on this head: a Codex thumbs-up and an automated OpenAI review, neither with anything actionable.
• Checks: just check passed after the first merge. After the second merge and the fix: cargo test -p moq-mux --all-features (1757 lib tests), clippy with -D warnings, and quest check all pass.

(Written by Claude Opus 5.5)