diff --git a/quest/m0/README.md b/quest/m0/README.md index 9141265dc82..929667880e5 100644 --- a/quest/m0/README.md +++ b/quest/m0/README.md @@ -33,6 +33,7 @@ edge, which embeds moq-relay; it moved there on 2026-09-28. - [Publish the release line's drain() wrappers](/quest/m0/ffi-shutdown-backport.md) - Dart `moq` 0.1.1 reaches pub.dev from `release` - [Release's audio encoder rolls groups before the cap](/quest/m0/release-audio-group-cap.md) - a long `groupDuration` with short frames never throws `GroupTooLarge` on `release` - [Local checks pass on unchanged main](/quest/m0/local-checks-main.md) - `sh/go/publish-wrapper.test.sh` passes under a global `tag.gpgsign = true`, so `just check` passes on a clean `main` +- [Bun-spawning test recipes install the JS workspace](/quest/m0/recipe-js-install.md) - `bare-fin`, `max-age`, `stream-limits`, `relay-fill`, and `receive-credit` pass in a fresh checkout - [CI cost and latency](/quest/m0/ci/README.md) - cloud CI runs only in GitHub's merge queue, one tested entry at a time, expensive checks run nightly and gate `release`, caches move to R2, and runners are chosen last - [More tests under load](/quest/m0/test-flakes-2/README.md) - load-only and queue failures, one quest per flake, fixed at the cause, since each one drops a queued PR - [Tests can't wait on the real clock](/quest/m0/clock-guard/README.md) - a test whose real-clock timer fires fails every run, in JS and, if the spike holds, Rust diff --git a/quest/m0/ci/README.md b/quest/m0/ci/README.md index 79289f9f40d..4bbc8732e17 100644 --- a/quest/m0/ci/README.md +++ b/quest/m0/ci/README.md @@ -180,5 +180,6 @@ The historical samples locate candidates; they are not controlled benchmarks. - [Archive the @moq packages on JSR](/quest/m0/ci/jsr-retire.md) - once `release` stops publishing there, the maintainer archives every JSR package, ending its OIDC link - [Verify the R2 Rust cache](/quest/m0/ci/mbx-r2-verify.md) - confirm the hosted Nightly jobs reuse the previous Nightly, then measure restore time and expiry - [Verify the Nix runtime cache](/quest/m0/ci/nix-cache-verify.md) - hosted Cache uploads `main`'s relay and CLI, and Smoke's nix channel substitutes them +- [The Nix CI cache never expires a path the newest cached commit needs](/quest/m0/ci/nix-cache-expiry.md) - reused store paths outlive the bucket's 30-day expiry while a recent commit needs them - [Release tags push to nix.moq.dev, and Cachix stops taking writes](/quest/m0/ci/cachix-retire.md) - backport `release-nix.yml` and its Just module to `release`, drop `cachix.yml`, prove a tag substitutes from nix.moq.dev, then revoke the Cachix token (the cache stays for old tags) - [Measure the Linux cache effect of native media exclusions](/quest/m0/ci/hakari-followups.md) - scheduled Linux jobs measure cache reuse after unrelated tests stop compiling PipeWire and libclang diff --git a/quest/m0/ci/nix-cache-expiry.md b/quest/m0/ci/nix-cache-expiry.md new file mode 100644 index 00000000000..7ede9dfda95 --- /dev/null +++ b/quest/m0/ci/nix-cache-expiry.md @@ -0,0 +1,22 @@ +# [S] The Nix CI cache never expires a path the newest cached commit needs + +## Goal + +The newest `main` commit with a successful Cache run always substitutes fully +from the CI bucket: no store path it references has passed the bucket's +30-day expiry (`infra/nix/lifecycle.json`, an `Age` rule on every object). + +## Plan + +Raised by [moq-dev/moq#8114](https://github.com/moq-dev/moq/pull/8114), which +pins Smoke to that commit. First confirm whether `sh/gh/cache-runtime.sh` (or +the copy it runs) skips uploading paths already in the bucket. If it does, a +path that many commits reuse keeps its first upload's age and expires while +newer commits still need it. Then fix it so the guarantee also holds when no +upload runs for 30 days: refreshing reused paths on upload alone does not +cover such a gap. For example, a scheduled job refreshes the newest cached +commit's closure, or the expiry deletes only paths no recent commit references. + +## Related + +- [Verify the Nix runtime cache](/quest/m0/ci/nix-cache-verify.md) - the scheduled Smoke check this protects diff --git a/quest/m0/recipe-js-install.md b/quest/m0/recipe-js-install.md new file mode 100644 index 00000000000..f84d3c673d6 --- /dev/null +++ b/quest/m0/recipe-js-install.md @@ -0,0 +1,20 @@ +# [XS] Bun-spawning test recipes install the JS workspace + +## Goal + +`just test bare-fin`, `max-age`, `stream-limits`, `relay-fill`, and +`receive-credit` pass in a fresh checkout. Each runs the root installer +before its Rust tests spawn `bun`, instead of relying on an earlier install. +`test/justfile` sets `working-directory := '.'`, so its recipes run in +`test/` and invoke it as `uv run --no-project python ../sh/js/install.py +frozen`, as [moq-dev/moq#7978](https://github.com/moq-dev/moq/pull/7978)'s +recipes do (CI's `interop.yml` "Workspace dependencies" step hides the gap today). + +## Plan + +Found by [moq-dev/moq#8104](https://github.com/moq-dev/moq/pull/8104): a fresh +checkout fails `just test bare-fin` with `Cannot find module '@moq/clock'`. +Use the coordinated installer from +[moq-dev/moq#7978](https://github.com/moq-dev/moq/pull/7978) rather than a raw +`bun install`, so the recipes share its lock and cache. Start once #7978 has +merged. diff --git a/quest/m0/test-flakes-2/README.md b/quest/m0/test-flakes-2/README.md index 71bf424e0ad..c35519a6cd2 100644 --- a/quest/m0/test-flakes-2/README.md +++ b/quest/m0/test-flakes-2/README.md @@ -145,3 +145,5 @@ Public API: none. Wire: none. - [Fallback playback](/quest/m0/test-flakes-2/ietf22-fallback.md) - a browser viewer follows a fallback within the media lane's bound without dropping its session - [dispatch-query's terminate test survives load](/quest/m0/test-flakes-2/dispatch-query-terminate.md) - `test_terminating_the_run_stops_the_running_phase` passes reliably under host load, re-enabled once fixed - [The datagram-end close follow-up](/quest/m0/test-flakes-2/datagram-end-close.md) - investigate whether duplicate SETUP can override an accepted lite-06 SUBSCRIBE_END during shutdown +- [moq-transcode's idle-rung cancel test survives load](/quest/m0/test-flakes-2/transcode-idle-rung-cancel.md) - `cancelling_an_idle_rung_reports_cancel` stops seeing `serve` return before its abort lands +- [moq-gst's sink-restart pad test survives load](/quest/m0/test-flakes-2/gst-sink-restart-pads.md) - `a_moqsink_restart_keeps_pads_when_caps_arrive_at_different_times` keeps its pads under load diff --git a/quest/m0/test-flakes-2/gst-sink-restart-pads.md b/quest/m0/test-flakes-2/gst-sink-restart-pads.md new file mode 100644 index 00000000000..4355ab0e32e --- /dev/null +++ b/quest/m0/test-flakes-2/gst-sink-restart-pads.md @@ -0,0 +1,17 @@ +# [S] moq-gst's sink-restart pad test survives load + +## Goal + +`source::imp::session_tests::a_moqsink_restart_keeps_pads_when_caps_arrive_at_different_times` +passes reliably under host load, fixed at the cause rather than retried. + +## Plan + +Seen 2026-11-07 on [moq-dev/moq#8122](https://github.com/moq-dev/moq/pull/8122)'s +Check ([job](https://github.com/moq-dev/moq/actions/runs/42341390803/job/128309636226)), +a PR that doesn't touch moq-gst: `assertion failed: playing.iter().all(|pad| +held.contains(pad) && pad.starts_with("video_"))` at +`rs/moq-gst/src/source/imp.rs:5830`, followed by a GStreamer-CRITICAL that +`moqsink0` was disposed in PAUSED instead of NULL. Reproduce under load first, +then find whether the test reads the pad set before the restart settles or the +source really drops a pad. diff --git a/quest/m0/test-flakes-2/transcode-idle-rung-cancel.md b/quest/m0/test-flakes-2/transcode-idle-rung-cancel.md new file mode 100644 index 00000000000..cc9aed5bcfa --- /dev/null +++ b/quest/m0/test-flakes-2/transcode-idle-rung-cancel.md @@ -0,0 +1,16 @@ +# [S] moq-transcode's idle-rung cancel test survives load + +## Goal + +`rung::tests::cancelling_an_idle_rung_reports_cancel` passes reliably under +host load, fixed at the cause rather than retried. + +## Plan + +Seen 2026-11-07 on [moq-dev/moq#8106](https://github.com/moq-dev/moq/pull/8106)'s +Check ([job](https://github.com/moq-dev/moq/actions/runs/42330347079/job/128275210492)), +a PR that doesn't touch moq-transcode: the test panicked at +`rs/moq-transcode/src/rung.rs:1077` because `serve` returned `Ok(())` before +the abort landed. It passed 5/5 unloaded. Reproduce under load first (pin it +to a few cores beside busy loops), then decide whether the race is in the test +or in how `serve` reports a cancel of an idle rung. diff --git a/quest/m1/README.md b/quest/m1/README.md index 8167de6e377..76390c8578c 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -34,6 +34,8 @@ moves what remains of them once their claims merge. - [Drop set](/quest/m1/drop-set.md) - each cursor judges missing groups against its own budget, so a narrow reader drops promptly and a widening re-offers groups already reported dropped on lite-07 - [Media fuzz findings](/quest/m1/fuzz-findings/README.md) - the nightly media fuzz targets run clean: mp4-atom, MKV block, TS step, SPS overflow, and catalog encoder bugs they found are fixed - [A smaller, safer API than release](/quest/m1/api-audit/README.md) - every surface added since release has a consumer or is gone, overlapping operations are collapsed, names mirror across Rust, JS, and the bindings, and every break says how to migrate +- [moq-relay's tests pass without default features](/quest/m1/relay-no-default-features.md) - QUIC-only tests are gated on `quic`, and the nightly matrix runs the no-default build +- [moq-webrtc drops the ICE wire counters nothing reads](/quest/m1/ice-wire-counters.md) - the per-packet `Wire` atomics go once #8122 makes the ICE module private - [Routing guide](/quest/m1/routing-guide.md) - concepts and practical outbound federation setup - [A reader on a failed copy continues on the held re-offer](/quest/m1/copy-recover.md) - a live reader keeps a group whose copy fails by continuing on the same track's re-offer at the same frame - [Group errors are cached by kind](/quest/m1/group-error-cache.md) - permanent group errors bind every reader, retryable ones are kept only for a bounded time, and a live SUBSCRIBE_DROP or FETCH answered `Old` no longer caches it for good diff --git a/quest/m1/ice-wire-counters.md b/quest/m1/ice-wire-counters.md new file mode 100644 index 00000000000..bff4004ea39 --- /dev/null +++ b/quest/m1/ice-wire-counters.md @@ -0,0 +1,19 @@ +# [XS] moq-webrtc drops the ICE wire counters nothing reads + +## Goal + +moq-webrtc's ICE module no longer counts every packet into `Wire` +(per-packet atomics, public as `moq_ice::Wire` before +[moq-dev/moq#8122](https://github.com/moq-dev/moq/pull/8122)), since only +tests read them. + +## Plan + +Decided 2026-11-07 (maintainer, /quest-plan from #8122): remove the counters +now, and let [Wire stats](/quest/m3/wire-stats/README.md) add an ICE source +with its adapter when it needs one. Not chosen: an ICE child in Wire stats +that wires them into Presence now; leaving them write-only. + +Milestone m1: #8122 makes the module crate-private (`mod ice;` in +`rs/moq-webrtc/src/lib.rs`), so removing `Wire` breaks no public API. Start +after #8122 lands. diff --git a/quest/m1/relay-no-default-features.md b/quest/m1/relay-no-default-features.md new file mode 100644 index 00000000000..720be63714b --- /dev/null +++ b/quest/m1/relay-no-default-features.md @@ -0,0 +1,13 @@ +# [S] moq-relay's tests pass without default features + +## Goal + +`just rs test -p moq-relay --no-default-features` passes: every test that +needs QUIC is gated on the `quic` feature, and the nightly feature matrix runs +that configuration so it stays green. + +## Plan + +Found by [moq-dev/moq#8081](https://github.com/moq-dev/moq/pull/8081): on its +merged head the configuration had 29 failures, all QUIC-only tests that don't +check for the feature. No CI row ran it, so it rotted unseen. diff --git a/quest/m2/README.md b/quest/m2/README.md index ad168b11ccb..d99076ffdb2 100644 --- a/quest/m2/README.md +++ b/quest/m2/README.md @@ -183,6 +183,9 @@ quests that require them. - [Finalize moq-lite-07](/quest/m2/lite07-finalize.md) - when the maintainer cuts it, lite-07 negotiates as `moq-lite-07` and the next release ships it - [Unbounded groups](/quest/m2/unbounded-groups/README.md) - groups and frames lose their size caps; a group keeps a 32 MB tail, and readers behind it replay through paced flights or refresh - [Session policy lives in moq-net](/quest/m1/session-policy/README.md) - accept and connect deadlines, admission, window widening and reconnect live once in moq-net; moq-tokio and moq-relay's io_uring path call them (held in m1 by an open claim) +- [One typed WebTransport error per QUIC adapter](/quest/m2/web-error.md) - moq-tokio and moq-uring each expose one typed `quic::web::Error`, and uring's `quic::Error` loses its `Web` and `Http3` variants +- [moq-quic drops inherited surfaces with no consumer](/quest/m2/moq-quic-unused-surface.md) - unused accept, connection, and stream-id surfaces leave moq-quic's public API before the cut +- [Publish reports rendition stats as hang's type](/quest/m2/rendition-stats-type.md) - `Rendition.stats` and `Report.Writer` take the rendition stats schema's input type, so encoders write plain numbers - [MoQ on rind](/quest/m1/rind/README.md) - every MoQ crate takes its crypto from rind, a pure-Rust graviola fork, so no build compiles C crypto (held in m1 by an open claim) - [Cut readiness](/quest/m2/cut-gate/README.md) - the checks the cut needs that break nothing: main no slower than release, vendored fixes ported, and moq-c at parity - [Probe](/quest/m2/probe/README.md) - a session probes up to the bitrate it wants to send or receive, so a stopped simulcast rendition resumes and a viewer up-switches; browser padding is parked in m3 (held in m2 by open claims; moves to m3) diff --git a/quest/m2/moq-quic-unused-surface.md b/quest/m2/moq-quic-unused-surface.md new file mode 100644 index 00000000000..2c610dd793a --- /dev/null +++ b/quest/m2/moq-quic-unused-surface.md @@ -0,0 +1,21 @@ +# [S] moq-quic drops inherited surfaces with no consumer + +## Goal + +moq-quic's public API loses the inherited surfaces nothing calls: +`endpoint::{Accepting, Accepted}`, `Endpoint::{start_accept, finish_accept}`, +`Connection::{side, is_handshaking}`, and `stream::Id::{new, dir, initiator}`. +Each is removed, not just made private, unless a caller inside moq-quic needs +it, in which case it becomes private. The PR's Impact section says how to +migrate; the release cut copies that into `MIGRATE.md`. + +## Plan + +Decided 2026-11-07 (maintainer, /quest-plan from +[moq-dev/moq#8088](https://github.com/moq-dev/moq/pull/8088)): remove them +before the cut, since removing public API after it would be another break. +Not chosen: folding the list into the post-cut +[dead-code audit](/quest/m3/audit-dead-code.md). + +Confirm each has no caller across the workspace, bindings, and vendored +crates before removing it. diff --git a/quest/m2/rendition-stats-type.md b/quest/m2/rendition-stats-type.md new file mode 100644 index 00000000000..ca0c15418bf --- /dev/null +++ b/quest/m2/rendition-stats-type.md @@ -0,0 +1,22 @@ +# [XS] Publish reports rendition stats as hang's type + +## Goal + +`@moq/publish`'s `Rendition.stats` takes hang's rendition stats schema's input +type (`z.input` of the schema behind `Stats.Rendition`), so publish no longer +keeps its own `RenditionStats` shape (`js/publish/src/rendition-internal.ts`) +beside the schema it writes. The encoders keep writing plain numbers into it +(`effect.proxy` in `js/publish/src/{video,audio}/encoder.ts`). + +## Plan + +Decided 2026-11-07 (maintainer, /quest-plan from +[moq-dev/moq#8088](https://github.com/moq-dev/moq/pull/8088)): `Report.Writer` +(`js/hang/src/report.ts`) accepts the schema's input type (`z.input`) and +brands on validation, so publishers pass plain numbers while readers keep the +`u53Schema` guarantee. The producer-side `Rendition.stats` signal carries that +same input type, since the encoders write it directly, not through +`Report.Writer`. Not chosen: dropping the `u53` brand from the stats +schemas, which would lose the 53-bit integer check on read. + +It changes a published TypeScript type, so it lands before the cut. diff --git a/quest/m2/web-error.md b/quest/m2/web-error.md new file mode 100644 index 00000000000..bd17c77ff85 --- /dev/null +++ b/quest/m2/web-error.md @@ -0,0 +1,21 @@ +# [S] One typed WebTransport error per QUIC adapter + +## Goal + +moq-tokio and moq-uring each expose one typed `quic::web::Error`, with +`Connection`, `Closed`, `SessionGone`, `UnknownSession`, `Http3`, `Handshake`, +`Read`, and `Write` variants, before the next cut. moq-uring's `Web` and +`Http3` variants move out of its `quic::Error`, and the `SessionError` alias is +dropped. Callers match on a variant instead of parsing a string. The PR's +Impact section says how to migrate; the release cut copies that into each +crate's `MIGRATE.md`. + +## Plan + +Decided 2026-11-06 (maintainer, API audit decision 3, from +[moq-dev/moq#8088](https://github.com/moq-dev/moq/pull/8088)): one typed +`web::Error` per crate, as a follow-up before the cut rather than inside the +audit PR. + +Today the errors are split across `quic/web/{recv,send}.rs` in both crates, +and moq-uring's `quic::Error` carries `Web(String)` (`rs/moq-uring/src/quic/mod.rs`). diff --git a/quest/m3/ffi-exporter.md b/quest/m3/ffi-exporter.md index bd767d2c069..b14d3cc3d19 100644 --- a/quest/m3/ffi-exporter.md +++ b/quest/m3/ffi-exporter.md @@ -58,10 +58,11 @@ which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and - moq-ffi depends on the `rust2-uniffi*` crates directly, with no `[patch]` and no git dependency, so crates.io moq-ffi still builds. - **Generated code stays identical**: Kotlin package names, library symbol - prefixes, and every wrapper's API do not change. Diff the regenerated - bindings for every language against `main`'s, and run each wrapper's check - plus `just test interop --all`. The C names below are the only expected - differences. + prefixes, and every wrapper's API do not change, except Dart's + `MoqRoute.instanceId()` below. Diff the regenerated bindings for every + language against `main`'s, and run each wrapper's check plus + `just test interop --all`. The C names and that Dart rename below are the + only expected differences. - **C names** (maintainer, 2026-10-24 and 2026-10-25): rust2's C generator names in GLib style ([moq-dev/rust2#66](https://github.com/moq-dev/rust2/pull/66)). @@ -74,6 +75,17 @@ which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and [moq-dev/moq#6052](https://github.com/moq-dev/moq/pull/6052) worked around the old clash with `moq_audio_encoder_finish`; every wrapper's `EncoderFinish` alias follows, so wrapper APIs do not change. +- **Dart `MoqRoute.instanceId()`** (maintainer, 2026-11-07, from + [moq-dev/moq#8088](https://github.com/moq-dev/moq/pull/8088)'s API audit): + the switch renames it to the exporter's name for moq-net's + `Route::instance()`, a Dart API change the PR's Impact section records for + the release cut's `MIGRATE.md`. Open: rust2's C++ generator declares a + private `uint64_t instance` field in every object class + (`uniffi/cpp/bindgen/src/bindings/cpp/templates/obj.hpp`), which is why + moq-ffi names the method `instance_id` (`rs/moq-ffi/src/origin.rs`). Before + the rename, either rename rust2's private field (recommended, since rust2 + is ours and the clash would hit any `instance` method), or keep the + generated `instance_id` and add the Dart name as a wrapper alias. - rust2's C backend still refuses `Duration`, which moq-ffi's audio configs use (moq's current generator accepts it), so the fork must accept it before the switch can generate C. @@ -126,8 +138,8 @@ which publishes `rust2-uniffi`, `rust2-uniffi-core`, `rust2-uniffi-macros`, and Public API: moq-ffi's dependency names change, generated C type names become CamelCase, and the generated record `MoqAudioEncoderFinished` becomes -`MoqAudioEncoderFinish` in every language; wrapper APIs do not change. Wire: -none. +`MoqAudioEncoderFinish` in every language; wrapper APIs do not change except +Dart's `MoqRoute.instanceId()`, which takes the exporter's name. Wire: none. ## Required