From a4474deecf25e4f1fad4d338deef8b9e3a0a0665 Mon Sep 17 00:00:00 2001 From: Luke Curley <432854+kixelated@users.noreply.github.com> Date: Mon, 9 Nov 2026 17:20:53 -0800 Subject: [PATCH] quest: unblock 2026-11-09 waits, round 2 Move the cut gate's candidate recheck and template into the release cut and delete the childless gate, delete the unrecurring IETF22 fallback flake, and record that an agent tags Dart moq 0.1.1 once moq-dev/moq#8378 merges. Co-Authored-By: Claude Opus 5.5 --- quest/README.md | 2 +- quest/m0/ffi-shutdown-backport.md | 18 +++++++++-- quest/m0/test-flakes-2/README.md | 1 - quest/m0/test-flakes-2/ietf22-fallback.md | 39 ----------------------- quest/m2/README.md | 9 +++--- quest/m2/cut-gate/README.md | 28 ---------------- quest/m3/release/cut.md | 25 +++++++++++++-- quest/m3/vendor-sync/README.md | 2 +- 8 files changed, 45 insertions(+), 79 deletions(-) delete mode 100644 quest/m0/test-flakes-2/ietf22-fallback.md delete mode 100644 quest/m2/cut-gate/README.md diff --git a/quest/README.md b/quest/README.md index 93025fa18e8..44d4a25ff1e 100644 --- a/quest/README.md +++ b/quest/README.md @@ -13,7 +13,7 @@ m1 holds this cycle's additive work: improvements and fixes that need no semver-breaking release, so each ships with the next cut or as a `release` patch when backported. m2 is the next breaking release: every change that would need one if it landed right after the cut, plus the -[cut gate](/quest/m2/cut-gate/README.md). The cut waits on m2 alone. m3 and m4 +cut gate the [release cut](/quest/m3/release/cut.md) re-creates there. The cut waits on m2 alone. m3 and m4 are later work split by priority: m3 holds what a named consumer waits on, release execution, and bug fixes in shipped code; m4 holds tests, tooling, benchmarks, optimizations, and features nobody waits on yet. m5 holds work diff --git a/quest/m0/ffi-shutdown-backport.md b/quest/m0/ffi-shutdown-backport.md index 062d391faf1..8a730d5f1db 100644 --- a/quest/m0/ffi-shutdown-backport.md +++ b/quest/m0/ffi-shutdown-backport.md @@ -7,9 +7,7 @@ The `release` line's hand-written Dart wrapper that gained `drain()` in [`moq`](https://pub.dev/packages/moq) 0.1.1, is on pub.dev. Check: pub.dev lists the version. It is already set on `release` -(`dart/moq/pubspec.yaml`), so this needs a tag push, not code. The publish -is outward-facing, so the maintainer runs it; agents do not push tags, -dispatch workflows, or bump versions. +(`dart/moq/pubspec.yaml`), so this needs a tag push, not code. ## Plan @@ -39,8 +37,22 @@ are ready, and merge `main` into `release` for the next minor line separately, later. Not chosen: one combined cut that merges `main` first, or neither this week. +Decided 2026-11-09 (maintainer, /quest-unblock): ✅ an agent pushes +`moq-dart-v0.1.1` at `release`'s head once the combined cut's release-plz +PR, [moq-dev/moq#8378](https://github.com/moq-dev/moq/pull/8378), merges +(rejected: tag before #8378 merges; the maintainer pushes it). Then it checks +that `release-dart.yml` succeeds and pub.dev lists 0.1.1, and deletes this +quest. + Public API: none. Wire: none. +## Waiting + +- moq-dev/moq: the combined patch cut's release-plz PR, + [moq-dev/moq#8378](https://github.com/moq-dev/moq/pull/8378), merges into + `release`. Check with `gh pr view 8378 --repo moq-dev/moq --json state`; + then push the tag as the Plan's 2026-11-09 decision says. + ## Related - [A Dart tag publishes unattended](/quest/m3/release/dart-publish.md) - proves the `moq-dart-v*` tag path this publish uses diff --git a/quest/m0/test-flakes-2/README.md b/quest/m0/test-flakes-2/README.md index fbc232e14b3..1926716bc71 100644 --- a/quest/m0/test-flakes-2/README.md +++ b/quest/m0/test-flakes-2/README.md @@ -247,6 +247,5 @@ Public API: none. Wire: none. ## Required -- [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 - [js/net's bandwidth sampling test holds up under load](/quest/m0/test-flakes-2/bandwidth-sampling.md) - the sampling test counts ticks without racing the mocked clock, and its `test.skip` is removed - [js/net's lite-07 datagram queue test holds up under load](/quest/m0/test-flakes-2/datagram-queues.md) - the test waits for the first track's queue instead of asserting at once, and its `test.skip` is removed diff --git a/quest/m0/test-flakes-2/ietf22-fallback.md b/quest/m0/test-flakes-2/ietf22-fallback.md deleted file mode 100644 index 728c9cd16b7..00000000000 --- a/quest/m0/test-flakes-2/ietf22-fallback.md +++ /dev/null @@ -1,39 +0,0 @@ -# [S] The browser viewer follows a fallback without dropping its session - -## Goal - -When the newest of two live publishers withdraws, a browser `@moq/watch` -viewer on moq-transport-22 or moq-lite keeps its session and plays the -still-live older run within the media lane's derived 1039 ms bound. The -session drop is fixed at its cause once it recurs with enough evidence to -reproduce it deterministically. - -## Waiting - -- moq-dev/moq: the session drop recurs with relay debug logs and the viewer's - close reason. It has not recurred since - [moq-dev/moq#6548](https://github.com/moq-dev/moq/pull/6548)'s IETF22 fixes; - clean runs do not clear this wait. Advance it by running `just test media - --media-cases replacement --media-engines chromium` with - `RUST_LOG=moq_net=debug,moq_relay=debug` (inside the dev shell, under - `env -u MOQ_QUEST`). - -## Plan - -**Session drop:** seen 1 in 9 runs before -[moq-dev/moq#6548](https://github.com/moq-dev/moq/pull/6548)'s IETF22 fixes -(host load average 30 to 70), 0 in 37 since. After the fallback NAMESPACE, the -viewer restarted its subscriptions. Then the relay stopped its own -SUBSCRIBE_NAMESPACE stream to the viewer, and the session closed. The viewer -reconnected and played the older run, but only after the bound. Relay debug -logging was off for that run, so it cannot say whether the relay closed over a -violation or the session closed cleanly. The viewer's reconnect warning now -names the close reason; start there. - -The audio stall that shared this quest was fixed in -[moq-dev/moq#7656](https://github.com/moq-dev/moq/pull/7656): on-demand audio -encoders replayed all PCM retained while unsubscribed, ignoring the demand's -max delay, so a rejoining live viewer drained seconds of stale groups before -audio resumed. - -Public API: none expected. Wire: none expected. diff --git a/quest/m2/README.md b/quest/m2/README.md index f418eca7b58..779caffd76b 100644 --- a/quest/m2/README.md +++ b/quest/m2/README.md @@ -3,8 +3,9 @@ ## Goal Every change that would need a semver-breaking release if it landed right -after the next cut of `main` into `release`, plus the [cut -gate](/quest/m2/cut-gate/README.md). The cut waits on this milestone alone; +after the next cut of `main` into `release`, plus the cut +gate, which the [release cut](/quest/m3/release/cut.md) re-creates here each +cycle. The cut waits on this milestone alone; complete it before cutting main into release. ## Plan @@ -23,7 +24,8 @@ unbounded-groups epic, whose additive compressed-stream child rides with it. The cut gates that break nothing (main no slower than release, vendored fixes, and moq-c parity) live under the recurring -[cut gate](/quest/m2/cut-gate/README.md), decided the same day. +cut gate, decided the same day; the [release cut](/quest/m3/release/cut.md) +holds its template. Held by open claims: the m1 epics and quests marked below stay where they are until their claims merge, and the [probe epic](/quest/m2/probe/README.md) @@ -200,6 +202,5 @@ quests that require them. - [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 - [Small API fixes before the cut](/quest/m2/api-shape/README.md) - one path and one spelling per item, names by role, and no dead or wrong-role public surface across net, transport, media, native, apps, JS, and bindings - [Finishing a media producer returns a Finished handle](/quest/m2/finished-handle.md) - `finish(self)` consumes the media crates' producers, as it does moq-net's, so appending after finish doesn't compile -- [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) - [Finish the milestone re-cut](/quest/m2/recut-held.md) - claim-held epics move to the milestone their kind and priority give them once their claims merge diff --git a/quest/m2/cut-gate/README.md b/quest/m2/cut-gate/README.md deleted file mode 100644 index 261d9744780..00000000000 --- a/quest/m2/cut-gate/README.md +++ /dev/null @@ -1,28 +0,0 @@ -# [S] Cut readiness - -## Goal - -Before each cut of `main` into `release`, the checks the cut needs that break -nothing have passed: `main` is no slower than `release`, vendored code has -the upstream fixes it needs, and the hand-written `moq-c` is at parity with -moq-ffi. This epic's own work, once every entry under `Required` has merged: -confirm the children's results still hold on the release candidate commit and -record that commit in the [release cut](/quest/m3/release/cut.md)'s Plan. - -## Plan - -Decided 2026-11-05 (maintainer, milestone re-cut): the cut waits on m2 alone, -so the checks that gate it without breaking anything live under one recurring -gate in m2 instead of in m1, where the cut no longer looks. - -Recurs once per release cycle, always in m2. It does not re-create itself: -m2 holds it and the cut requires m2, so a copy made before the cut would -block that cut forever. The release cut's PR re-creates it at this path while -it shifts the milestones, after removing this path from every `Required`. The -new gate keeps this Goal and this paragraph, gets a fresh "main is no slower -than release" child with its Goal and no measurements yet, and takes the -[vendor sync](/quest/m3/vendor-sync/README.md) questline that this -run re-created in m3. Only `/quest-delete` ends the recurrence. - -The completed [release performance comparison](/.github/release-perf.md) -records the accepted costs and final relay pairing for this cut. diff --git a/quest/m3/release/cut.md b/quest/m3/release/cut.md index c5a98fc7d03..de2c5460b4c 100644 --- a/quest/m3/release/cut.md +++ b/quest/m3/release/cut.md @@ -60,8 +60,8 @@ moved path is rewritten. A recurring cut-gate child that its last run re-created in m3, such as the [vendor sync](/quest/m3/vendor-sync/README.md), moves under the re-created gate instead. That PR also removes this path from every `Required`, which -releases the audits waiting on it. It then re-creates the [cut -gate](/quest/m2/cut-gate/README.md) in m2 as that gate's Plan describes, and +releases the audits waiting on it. It then re-creates the cut gate at `quest/m2/cut-gate/README.md` from +the template below, and this quest, the [API shape review](/quest/m3/release/api-review.md), the [feature review](/quest/m3/release/feature-review.md), and the [Cargo feature review](/quest/m3/release/feature-flags.md) at the same paths under a fresh @@ -84,6 +84,27 @@ merged moq-e2ee and moq-sign into moq-crypto (decided 2026-11-02, no yank): already points at moq-crypto, and `rs/moq-e2ee` is gone. Don't merge a `release` branch release-plz proposal that bumps it. +Cut gate template (moved here 2026-11-09 from the gate's own Plan; see the +decision below). The gate is an epic in m2 for the checks the cut needs that +break nothing: `main` is no slower than `release`, vendored code has the +upstream fixes it needs, and the hand-written `moq-c` is at parity with +moq-ffi. It has no work of its own and completes when every child merges. It +recurs once per release cycle, always in m2, and never re-creates itself: m2 +holds it and the cut requires m2, so a copy made before the cut would block +that cut forever. Re-create its README with that Goal and this recurrence, a +fresh "main is no slower than release" child with its Goal and no +measurements yet, and the [vendor sync](/quest/m3/vendor-sync/README.md) +questline that the last run re-created in m3. Only `/quest-delete` ends the +recurrence. + +Decided 2026-11-09 (maintainer, /quest-unblock): the cut, not the gate, +confirms on its candidate commit that the gate's children's results still +hold, and records that commit here. ✅ Move the recheck into the cut, and +delete the gate while it has no children (rejected: the gate requiring every +other m2 entry by hand; a /quest-plan interview). The gate had no children +left and nothing made it wait for the rest of m2, so it showed as ready, and +a candidate commit recorded then would go stale before the cut. + The [2026-10-22 release audit](https://github.com/moq-dev/moq/pull/7176) closed with its fixes rechecked together. At source candidate [main 487759f62a](https://github.com/moq-dev/moq/commit/487759f62a9ef39d916fe5b92df06dcff83478ce), diff --git a/quest/m3/vendor-sync/README.md b/quest/m3/vendor-sync/README.md index df81b08ee45..2c958955bb1 100644 --- a/quest/m3/vendor-sync/README.md +++ b/quest/m3/vendor-sync/README.md @@ -45,7 +45,7 @@ This README's own work, once every child has merged: first, what was ruled out). Each child requires [the release cut](/quest/m3/release/cut.md), which removes itself from every `Required` and moves the questline under the re-created - [cut gate](/quest/m2/cut-gate/README.md) in m2, so the next run becomes + cut gate (`quest/m2/cut-gate/`) in m2, so the next run becomes ready there, once per release cycle (rule replaced 2026-11-05). 3. Drop a child whose code was deleted or whose upstream stopped mattering, saying why in the PR.