Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions quest/m1/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -69,6 +69,8 @@ session outcomes, and embedder refusals inside session outcomes.
- [Dogfood hosted worklets](/quest/m1/dogfood-assets.md) - the moq.pro dashboard hosts the worklets and calls `assets()` after the release
- [TS pair acquisition](/quest/m1/ts-pair-acquisition.md) - fix the clean baseline and prove the grader's controls before #5328 enables its scheduled hard gate
- [Initial catalog routing](/quest/m1/initial-catalog-routing.md) - a usable fallback does not strand initial catalog acquisition with a spurious Unroutable
- [Live TS clock validation](/quest/m1/ts-live-clock.md) - resolve the local PCR-rate failure with reproducible timing evidence and CI controls
- [Same-instance TS resume](/quest/m1/ts-same-instance-resume.md) - prove valid continuation and fix the unexpected final exporter failure without weakening identity
- [More tests under load](/quest/m1/test-flakes-2/README.md) - the second round of load-only failures, one quest per flake, fixed at the cause
- [END_OF_TRACK placement](/quest/m1/ietf-end-of-track-placement.md) - END_OF_TRACK rides the upstream's Location, and Rust's header stops claiming END_OF_GROUP
- [Plan: a public model](/quest/m1/model-public-surface.md) - every crate-private model item is made public or unnecessary, so a custom publisher or subscriber builds from the public model alone
Expand Down
59 changes: 59 additions & 0 deletions quest/m1/ts-live-clock.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,59 @@
# [M] Live TS clock-rate validation identifies and fixes timing defects

## Goal

Resolve the local live PCR-rate failure seen during automated TS validation.
The exporter and the grader agree on the supported release-clock contract,
with reproducible evidence and automated controls that distinguish valid
output from actual timing defects.

## Plan

Decided with the maintainer on 2026-10-11: investigate and fix this finding
independently of the same-instance resume failure. Preserve strict publisher
identity, the existing validation gates, and production negotiation defaults.
Keep validation PR #5328 draft while its local readiness findings remain
unresolved; a passing CI run alone does not explain the local failure.

The epoch-capable 120-second local run completed with publisher and exporter
exit zero, but the grader reported -1382.60 ppm against its +/-30 ppm limit.
Release timing passed with 224.4 ms accumulated lag and -0.003 ms/s over the
last 38.7 seconds; byte-position scheduling passed at 100%. Logs recorded
WebSocket fallback, old-session forced closure, and a later video deadline
miss. These are observations, not an established causal chain. The same-head
full CI run passed with -0.24 ppm and no release-interval violations.

`test/ts/pcr-timing.py` already excludes the first third before fitting PCR
against arrival time. Do not attribute the failure to startup merely because
the tail is nearly flat. The retained local report contains aggregates, not
the raw arrival samples needed to replay that fit. Retain enough clock,
packet-position, discontinuity and arrival evidence to reproduce the verdict
and separate source timing, clock recovery, export release, pipe backpressure
and measurement behavior. Compare a direct connection with controlled
fallback/handover where useful; do not infer transport causality from timing.

Reproduce the cause before changing production code or the oracle. Correct
the layer that is wrong, including the grader if its measurement does not
support its claim. Never obtain a pass by widening the limit, choosing a
convenient measurement window, disabling fallback, or retrying. A necessary
change to the externally promised timing contract needs a maintainer decision.
Check current main and the related open work before duplicating a fix.

Add deterministic positive and negative timing controls with the intended
failure reason asserted. They should separate valid bounded delay and
measurement effects from the clock-rate or release defect being fixed.
Replay evidence where possible and mock time in unit tests. Keep byte-position
PCR accuracy distinct from arrival-clock measurements. Run short controls in
PR CI and real live timing at least nightly, retaining diagnostic artifacts.
Completion requires a causal regression and the corrected live validation,
not a single successful rerun. Update the harness documentation inline;
public API and wire changes are not expected.

## Related

- [Validation PR #5328](https://github.com/moq-dev/moq/pull/5328#issuecomment-6114456593) - retained local failure and process-status evidence
- [Passing full CI](https://github.com/moq-dev/moq/actions/runs/38178905198) - same-head contrast, including real live timing
- [Same-instance resume](/quest/m1/ts-same-instance-resume.md) - independent full-check failure
- [Transport upgrade](/quest/m1/transport-upgrade/README.md) - existing session handover and epoch contract
- [Mux-rate hold](/quest/m1/tstd/mux-rate-hold.md) - #5287 owns initial rate publication and its finite-source behavior
- [TS AC-3 scheduling](/quest/m1/ts-atsc-ac3.md) - #5306 owns decoder deadlines and unknown-rate startup overflow
65 changes: 65 additions & 0 deletions quest/m1/ts-same-instance-resume.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
# [M] Same-instance TS resume has a valid, reliable regression

## Goal

A valid same-instance TS return within the configured linger resumes output
and finishes with the publisher's clean final verdict. Resolve the existing
CLI regression's unexpected exporter failure while preserving explicit
refusal of a different publisher instance.

## Plan

Decided with the maintainer on 2026-10-11: investigate and fix this finding
independently of the live PCR-rate discrepancy. Preserve strict identity:
matching epochs only permit continuation of the same named content. Reusing
an epoch never licenses different groups or catalog content under the same
names. Do not add implicit stitching, automatic replacement acquisition, or
change production protocol defaults.

The integrated validation branch's `just check` stopped after 3,566 passes
at `moq-cli::export::the_same_instance_within_the_linger_continues`. The test
observed resumed output, then expected a clean final exit and received `exit 1`.
Its captured exporter stderr is discarded, so the underlying error is not
established. This is distinct from initial catalog acquisition, whose
same-epoch success and epochless refusal controls pass in #5333.

First retain the exporter's error and reproduce on current main. The fixture
interrupts an importer, then starts a fresh process replaying the same clip
under the same explicit epoch. Establish whether its catalog, tracks, group
numbers and contents satisfy the identity and continuation contract before
calling the exporter's refusal a production bug. Correct an invalid fixture
at its source; do not make the exporter accept invalid identity reuse. If a
valid continuation fails, isolate the responsible ordering and fix that layer.
Do not label it a load flake without evidence.

Check #5305 and the same-epoch importer work before duplicating implementation.
A pending return catalog, a clean final finish, a replacement switch and an
invalid same-epoch replay are different cases until reproduction connects them.
If an existing quest owns the cause, put the regression and fix there and
reconcile this quest's scope. Do not expand this into the full redundant
importer redesign just to obtain a green CLI test.

Make the regression depend on controlled events and mocked time where
possible. Retain real relay coverage for the CLI outcome. Prove resumed
content belongs to the intended instance and preserve controls for a genuine
replacement, a missing return, and a failed final end as applicable to the
cause. Assert useful error evidence on failure. Do not increase sleeps,
linger, or timeouts, add retries, or treat the success fixture's nonzero exit
as acceptable. Every regression runs in CI, at least nightly for long integration
coverage. Completion requires the causal test and the relevant full checks,
not a successful isolated rerun after a loaded failure.

Keep validation PR #5328 and catalog PR #5333 draft until their readiness
findings are resolved. Update affected test documentation and comments
inline. No new public API, wire format, or user guide is expected; a newly
required behavior or API change returns to the maintainer for a decision.

## Related

- [Validation PR #5328](https://github.com/moq-dev/moq/pull/5328#issuecomment-6114456593) - failing full-check evidence
- [Catalog identity controls #5333](https://github.com/moq-dev/moq/pull/5333) - initial acquisition preserves the selected contract
- [Live clock validation](/quest/m1/ts-live-clock.md) - independent local live failure
- [Replacement catalog linger](/quest/m1/ts-follow-catalog-bound.md) - #5305 owns the bound on a mid-stream switch's first catalog
- [Same-epoch importers](/quest/m1/hop-aligned-import.md) - shared catalogs, tracks and source-derived group identity
- [Closed fallback](/quest/m1/transport-upgrade/closed-fallback.md) - #5284 owns early WebSocket closure and its transport-test fix
- [TS duration fidelity](/quest/m0/ts-duration-fidelity.md) - diagnose a shared end-of-stream cause only if reproduced
Loading