From 6380041448f6ec21f3a7a34d5360ccd82f5822e1 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 28 Oct 2026 17:36:49 -0700 Subject: [PATCH 1/3] chore(quest): claim quest/m1/release-audit/README.md Quest-Claim: 9a18ef5e07104863 From e2e196280806b7bc88a67f52010961f134e25fb0 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Wed, 28 Oct 2026 18:07:24 -0700 Subject: [PATCH 2/3] docs(quest): checkpoint release audit candidate evidence Co-Authored-By: GPT-6.1 Sol --- quest/m1/release-audit/README.md | 68 ++++++++++++++++++++++++++++++++ quest/m2/release/cut.md | 15 +++++++ 2 files changed, 83 insertions(+) diff --git a/quest/m1/release-audit/README.md b/quest/m1/release-audit/README.md index 040306f9d8e..4bd3a886b41 100644 --- a/quest/m1/release-audit/README.md +++ b/quest/m1/release-audit/README.md @@ -88,6 +88,73 @@ The 2026-10-22 binding and OBS audit at main rate recovery child. Its two OBS findings went to m2 next to the other OBS work, because obs-moq ships on its own schedule. +### Candidate recheck (2026-10-28) + +The source candidate is +[main 487759f62a](https://github.com/moq-dev/moq/commit/487759f62a9ef39d916fe5b92df06dcff83478ce), +with +[release 7fad611aae](https://github.com/moq-dev/moq/commit/7fad611aae4ef4e7fa856ed077a0bb467f2de9be) +as the current shipping branch. Main contains the resolved audit fixes and is +usable for their combined source checks. It is not the final cut tree: release +has not been merged into main, and the existing release reviews remain open. +No release, version bump, or merge is authorized by this checkpoint. + +Checks at this candidate: + +- Native protocol, auth, media, relay, archive, FFI/C, and QMux suites passed: + 7,697 tests, 43 skipped. The FFI used its OpenH264 feature. +- JS lint/type/build checks and the unit-test phase passed (4,779 tests). + Python passed 104 tests; C, C++, Kotlin, and Go checks passed. Draft checks + and the OBS ThreadSanitizer suite passed, including reordered timestamps + and EOF draining. These are workspace builds, not registry-install proof. +- Dart analysis and all 45 raw/wrapper tests passed, including idle stream + cancellation. Its publish dry-run failed because the staged package inside + ignored `.tmp/` inherited the checkout's ignore rules (hidden pubspec, + missing LICENSE). The existing + [moq-dev/moq#7117](https://github.com/moq-dev/moq/pull/7117) fixes that exact + staging failure; no duplicate fix is made here. This is not evidence of a + broken published Dart artifact. +- `just test interop --all` passed nine native interop tests and all 176 + configured harness cells: media and denied auth, finite finish, late join, + FETCH/tails, compression, and partial-group drops. Its two indefinite Rust + publisher runs needed forced cleanup after the stop grace period. The + finite-finish and tail checks passed independently; all children were reaped + and ports released. This was the default negotiated-version matrix, not the + optional complete IETF draft matrix. +- The QMux record and CMAF declared-count benchmark sweeps passed. CMAF + refused the hostile 4,294,967,295-sample declarations over both payload sizes. +- The release-mode track benchmark passed all 27 expiry-budget, fetch-park, + and datagram-pressure cases. These are quick, single-candidate measurements + on a shared host, not a new paired main-versus-release performance verdict. + +The full Linux gate is not proven here: the runner-image build was denied +user namespaces, and the local flake probe rejected a scratch Unix socket. +The full native feature-matrix run was stopped in favor of the focused +regression suites above. Swift was skipped on this Linux host. Platform, +installed-artifact, public-API diff, and full OBS runtime proofs retain their +existing owners. Checks used the Nix toolchain with the inherited `BASH_ENV` +hook disabled; that host hook otherwise caused Cargo launches to fail with +`Argument list too long`. + +The registry resolver selects moq-cli 0.14.9, moq-relay 0.17.9, hang 0.21.18, +`@moq/net` 0.4.7, `@moq/auth` 0.2.2, `@moq/hang` 0.5.4, `@moq/json` 0.4.3, +and `@moq/web-transport` 0.1.6. The normal compatibility guard refuses the +accepted exception because it still pins moq-cli 0.14.8. The owning +[moq-dev/moq#7061](https://github.com/moq-dev/moq/pull/7061) has already run the +full remeasurement and found failing stable published CLI controls at these +same registry pins; it leaves the exception unchanged. Do not duplicate that +fix or repin from the source-only results here. + +Remaining work: repeat the Dart publish check after its existing staging fix +lands, and complete normal published-wire compatibility after its owner +resolves the stable control failure. Refresh affected source checks if the +candidate changes, then delete this epic and give the completed evidence to +the cut quest. +Recommendation: investigate and publish the corrected CLI through the existing +compatibility quest before closing this audit. Deferring that proof to the +breaking cut is a maintainer decision, not an inferred acceptance of the +published failure. + ## Related - [JS subscription ranges](/quest/m1/subscribe-ranges/js.md) - range lists and frame offsets build on the bounds fixed here @@ -96,5 +163,6 @@ work, because obs-moq ships on its own schedule. - [Measured release performance](/quest/m1/release-perf/README.md) - existing paired benchmark regressions and their confirmation runs - [Transport performance audit](/quest/m2/audit-perf.md) - congestion-control coverage remains outstanding - [Binding timestamps](/quest/m1/binding-timestamps.md) - exact timestamps across language APIs +- [Published-wire comparison](/quest/m0/interop-green/wire-compat-remeasure.md) - stable published controls and the planned-break refresh - [Dependency audit](/quest/m2/audit-deps.md) - existing advisory inventory - [Publish and adopt](/quest/m2/release/README.md) - API/feature review, merge order, candidate checks, publication, and installed-artifact proof diff --git a/quest/m2/release/cut.md b/quest/m2/release/cut.md index df1b39335ff..42d5d658b6b 100644 --- a/quest/m2/release/cut.md +++ b/quest/m2/release/cut.md @@ -45,6 +45,21 @@ From the 2026-10-25 pre-cut run: `MOQ_FFI_RANGE`) to the release that ships the `implementation` field from [moq-dev/moq#6775](https://github.com/moq-dev/moq/pull/6775). +The [2026-10-28 audit checkpoint](https://github.com/moq-dev/moq/pull/7176) +records source candidate +[main 487759f62a](https://github.com/moq-dev/moq/commit/487759f62a9ef39d916fe5b92df06dcff83478ce) +and shipping branch +[release 7fad611aae](https://github.com/moq-dev/moq/commit/7fad611aae4ef4e7fa856ed077a0bb467f2de9be). +The focused native suites (7,697 passes), JS checks, draft checks, OBS sanitizer +suite, and all 176 configured all-language interop cells passed. Dart's 45 tests +passed, but its checkout-local publish dry-run hit the staging ignore bug owned +by [moq-dev/moq#7117](https://github.com/moq-dev/moq/pull/7117). +The checkpoint does not select or approve the final cut tree. The published-wire comparison +is blocked by stale exception pins and the stable released CLI control failures +owned by [moq-dev/moq#7061](https://github.com/moq-dev/moq/pull/7061). Resolve +that existing prerequisite and repeat the candidate compatibility checks after +the documented release-to-main merge; source-only passes do not replace them. + ## Required - [Breaking release readiness](/quest/m1/README.md) - code, protocol and package proofs From cd6e3e68d26e5700a4539517f6f3280d3a350bb2 Mon Sep 17 00:00:00 2001 From: Luke Curley Date: Sat, 31 Oct 2026 00:25:18 -0700 Subject: [PATCH 3/3] docs(quest): close the release audit with its compatibility proof Normal wire-compat passes against moq-cli 0.14.10 and Dart's staged publish dry-runs pass, so the audit's last proofs are in. Delete the epic and hand its evidence to the cut quest. Co-Authored-By: Claude Opus 5.5 --- quest/m1/README.md | 1 - quest/m1/cut-audit/README.md | 1 - quest/m1/release-audit/README.md | 167 -------------------------- quest/m2/performance-survey/README.md | 2 +- quest/m2/release/cut.md | 26 ++-- 5 files changed, 13 insertions(+), 184 deletions(-) delete mode 100644 quest/m1/release-audit/README.md diff --git a/quest/m1/README.md b/quest/m1/README.md index aaebf63bfb1..42cd56b8b98 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -123,7 +123,6 @@ removed). ## Required - [Address groups by timestamp](/quest/m1/time-addressing/README.md) - group start timestamps replace sequence numbers in the moq-net API and lite-07 wire, while old subscribers, old publishers, and published JS clients keep working -- [Release audit findings](/quest/m1/release-audit/README.md) - resolve newly identified resource, parser, playback, and interoperability risks before the cut - [Pre-cut audit findings](/quest/m1/cut-audit/README.md) - fix the crashes, hostile-input costs, and correctness bugs the 2026-10-24 audit of main found - [moq-quic has every quinn fix it needs](/quest/m1/quinn-sync.md) - quinn's fixes since the fork point are ported with their tests, once per release cycle - [Security audit findings](/quest/m1/security-audit/README.md) - fix the verified relay, importer, moq-c, browser, and CI findings of the 2026-10-27 external security audit diff --git a/quest/m1/cut-audit/README.md b/quest/m1/cut-audit/README.md index 5fd84d3208c..d7e41667c1c 100644 --- a/quest/m1/cut-audit/README.md +++ b/quest/m1/cut-audit/README.md @@ -63,7 +63,6 @@ leaving JS-only redesigns of `js/net` to [rs2ts](/quest/m2/rs2ts/README.md). ## Related -- [Release audit findings](/quest/m1/release-audit/README.md) - the 2026-10-22 audit this one follows - [Untrusted media parsers have fuzz targets](/quest/m2/media-fuzz.md) - would have caught the media parser findings here - [Announce interest streams share one route table](/quest/m2/announce-interest-memory.md) - the audit's inherited announce memory finding - [A parked read of a resumed group reads its budget without re-syncing](/quest/m2/recover-budget.md) - measure-first resume cost diff --git a/quest/m1/release-audit/README.md b/quest/m1/release-audit/README.md deleted file mode 100644 index dd9f86e5b5b..00000000000 --- a/quest/m1/release-audit/README.md +++ /dev/null @@ -1,167 +0,0 @@ -# [S] Resolve the release audit findings - -## Goal - -Resolve the newly identified correctness, resource, and performance risks from -the 2026-10-22 release-to-main audit before the next cut. Existing quests keep -ownership of previously tracked problems. This epic's remaining work is to -recheck the resolved findings together on the release candidate and hand the -evidence to the existing cut quest. - -## Plan - -The maintainer requested an extensive audit and quests for bugs, potential -issues, and performance concerns. The comparison pins are -[main 6369eca](https://github.com/moq-dev/moq/commit/6369eca26472e120ee20c473027c13d36994edad) -and [release 36f392f](https://github.com/moq-dev/moq/commit/36f392f492f571454e5abc0e441ddc0168f16477). -Their tree diff spans 3,720 files, 575,722 insertions, and 125,659 deletions. - -Review covered protocol/model ownership, compression, auth, relay discovery, -QUIC/WebTransport/QMux, browser ingestion, media containers, FFI lifetimes, -archive migration, E2EE, release-only commits, and CI/publication paths. -This was risk-based source review with targeted reproductions, not a manual -verification of every changed line. Each child distinguishes reproduction, -source inference, and whether the problem is new or inherited. Priorities -describe impact; no unsupported exploit or performance magnitude is assumed. -No production code, public API, wire format, or package version changed in -the audit. Implementation choices remain with each fix; prefer the existing -ownership mechanisms and keep fixes scoped to their consumer. - -Validation at these pins: - -- Native protocol/auth/relay/JSON/stats: 3,830 passed, 18 skipped. -- Native task/clock/QUIC/socket: 923 passed, 2 skipped; Tokio/uring: - 656 passed, 4 skipped. A strict io_uring WebTransport echo also passed. -- Native mux/HLS/archive/E2EE: 1,992 passed, 5 skipped. -- Browser packages under Bun: 4,229 distinct tests passed across the initial - runs and targeted reruns. Initial failures were a denied socket listen and - two missing-libav tests; these passed after enabling loopback access and - building libav. Hang and watch also passed their normal per-package runs. - Combining the two polyfill tests in one custom Bun process polluted shared - WebCodecs globals; the repository isolates packages in separate processes. - This is not a claim that the full JS type/build/browser gate ran. -- Dependency audit reported the four already tracked unmaintained packages - basic-toml, zbus_names, zvariant, and zvariant_derive, not four new advisories. - -Run focused regression checks and relevant benchmark sweeps after the fixes, -then carry the candidate through the existing compatibility, package, draft, -and all-language interop checks. This audit did not run the full platform -matrix, real-browser WebCodecs/capture/GPU coverage, installed registry artifacts, -or a complete public-API diff against published packages. Those proofs retain -their existing owners. Reconcile release-only versions, changelogs, and -backports through the documented merge sequence before the cut. - -The deeper pass advanced main to -[1d43c096e0](https://github.com/moq-dev/moq/commit/1d43c096e04431bfc87d807ce47a546df8017475), -with the same release pin. It reviewed valid-input protocol differences, -subscription bounds, JSON/model expiry ownership, HLS/archive lifecycle, -FFI/C cancellation, and the handwritten Python/Swift/Kotlin/Go/Dart/C++ -wrappers. Native FFI/C checks added 371 passes and two skips. The repeated -309-test archive/HLS run passed with two skips; it is not additional distinct -coverage over the first pass. - -`TMPDIR=/tmp just test interop --all` passed at the deeper pin: seven native -interop tests and the complete configured integration harness. The harness -covered Rust/Python/Go/C++/browser publishers against Rust, Python, Go, C++, -Chromium, Node, Bun, C, and GStreamer consumers, permitted and denied AUTH, -browser session refusal, CLI FETCH close, finite tails, compression, and -partial-group drops. The initial run failed because nested temporary paths -exceeded Unix socket limits in Go tooling and Chromium; the short-path rerun -completed with no skips or failures. This is Linux workspace-artifact proof, -not installed-registry or mobile/macOS/Windows validation. - -Targeted reproductions confirmed a finite HLS startup race, retained HLS -server state after final drop, idle Dart stream cancellation, empty expiry -indexes, JS IETF timestamp changes, JS datagram range violations, and OBS -reordered-frame timestamps and missing EOF output. Each owning quest records -the reproduction's limits. JS reader and publisher datagram paths now enforce -subscription bounds, and the OBS source now stamps each decoded picture with its -own timestamp and drains at clean EOF. -Go reservation disposal was settled by -[moq-dev/moq#6424](https://github.com/moq-dev/moq/pull/6424): `Update(0)` -releases a reservation deterministically. -No platform-wide binding parity, full OBS runtime, fuzzing campaign, or -network-impairment performance result is implied by these checks. - -The 2026-10-22 binding and OBS audit at main -[1f5caa7026](https://github.com/moq-dev/moq/commit/1f5caa7026) added the -rate recovery child. Its two OBS findings went to m2 next to the other OBS -work, because obs-moq ships on its own schedule. - -### Candidate recheck (2026-10-28) - -The source candidate is -[main 487759f62a](https://github.com/moq-dev/moq/commit/487759f62a9ef39d916fe5b92df06dcff83478ce), -with -[release 7fad611aae](https://github.com/moq-dev/moq/commit/7fad611aae4ef4e7fa856ed077a0bb467f2de9be) -as the current shipping branch. Main contains the resolved audit fixes and is -usable for their combined source checks. It is not the final cut tree: release -has not been merged into main, and the existing release reviews remain open. -No release, version bump, or merge is authorized by this checkpoint. - -Checks at this candidate: - -- Native protocol, auth, media, relay, archive, FFI/C, and QMux suites passed: - 7,697 tests, 43 skipped. The FFI used its OpenH264 feature. -- JS lint/type/build checks and the unit-test phase passed (4,779 tests). - Python passed 104 tests; C, C++, Kotlin, and Go checks passed. Draft checks - and the OBS ThreadSanitizer suite passed, including reordered timestamps - and EOF draining. These are workspace builds, not registry-install proof. -- Dart analysis and all 45 raw/wrapper tests passed, including idle stream - cancellation. Its publish dry-run failed because the staged package inside - ignored `.tmp/` inherited the checkout's ignore rules (hidden pubspec, - missing LICENSE). The existing - [moq-dev/moq#7117](https://github.com/moq-dev/moq/pull/7117) fixes that exact - staging failure; no duplicate fix is made here. This is not evidence of a - broken published Dart artifact. -- `just test interop --all` passed nine native interop tests and all 176 - configured harness cells: media and denied auth, finite finish, late join, - FETCH/tails, compression, and partial-group drops. Its two indefinite Rust - publisher runs needed forced cleanup after the stop grace period. The - finite-finish and tail checks passed independently; all children were reaped - and ports released. This was the default negotiated-version matrix, not the - optional complete IETF draft matrix. -- The QMux record and CMAF declared-count benchmark sweeps passed. CMAF - refused the hostile 4,294,967,295-sample declarations over both payload sizes. -- The release-mode track benchmark passed all 27 expiry-budget, fetch-park, - and datagram-pressure cases. These are quick, single-candidate measurements - on a shared host, not a new paired main-versus-release performance verdict. - -The full Linux gate is not proven here: the runner-image build was denied -user namespaces, and the local flake probe rejected a scratch Unix socket. -The full native feature-matrix run was stopped in favor of the focused -regression suites above. Swift was skipped on this Linux host. Platform, -installed-artifact, public-API diff, and full OBS runtime proofs retain their -existing owners. Checks used the Nix toolchain with the inherited `BASH_ENV` -hook disabled; that host hook otherwise caused Cargo launches to fail with -`Argument list too long`. - -The registry resolver selects moq-cli 0.14.9, moq-relay 0.17.9, hang 0.21.18, -`@moq/net` 0.4.7, `@moq/auth` 0.2.2, `@moq/hang` 0.5.4, `@moq/json` 0.4.3, -and `@moq/web-transport` 0.1.6. The normal compatibility guard refuses the -accepted exception because it still pins moq-cli 0.14.8. The owning -[moq-dev/moq#7061](https://github.com/moq-dev/moq/pull/7061) has already run the -full remeasurement and found failing stable published CLI controls at these -same registry pins; it leaves the exception unchanged. Do not duplicate that -fix or repin from the source-only results here. - -Remaining work: repeat the Dart publish check after its existing staging fix -lands, and complete normal published-wire compatibility after its owner -resolves the stable control failure. Refresh affected source checks if the -candidate changes, then delete this epic and give the completed evidence to -the cut quest. -Recommendation: investigate and publish the corrected CLI through the existing -compatibility quest before closing this audit. Deferring that proof to the -breaking cut is a maintainer decision, not an inferred acceptance of the -published failure. - -## Related - -- [JS subscription ranges](/quest/m1/subscribe-ranges/js.md) - range lists and frame offsets build on the bounds fixed here -- [OBS video source](/quest/m2/obs-moq-video/source.md) - the decoder replacement keeps the timing and EOF behavior fixed here -- [Measured release performance](/quest/m1/release-perf/README.md) - existing paired benchmark regressions and their confirmation runs -- [Transport performance audit](/quest/m2/audit-perf.md) - congestion-control coverage remains outstanding -- [Binding timestamps](/quest/m1/binding-timestamps.md) - exact timestamps across language APIs -- [Published-wire comparison](/quest/m0/interop-green/wire-compat-remeasure.md) - stable published controls and the planned-break refresh -- [Dependency audit](/quest/m2/audit-deps.md) - existing advisory inventory -- [Publish and adopt](/quest/m2/release/README.md) - API/feature review, merge order, candidate checks, publication, and installed-artifact proof diff --git a/quest/m2/performance-survey/README.md b/quest/m2/performance-survey/README.md index 708bad5c8f5..7d8e9fe11da 100644 --- a/quest/m2/performance-survey/README.md +++ b/quest/m2/performance-survey/README.md @@ -150,7 +150,7 @@ No duplicate quest or workflow edit belongs here. [performance audit](/quest/m2/audit-perf.md) repeats it before later cuts. - [Perf](/quest/m2/perf/README.md) owns measured channel contention, relay allocations, ingest batching, worker quiescence and runtime selection. -- The [release audit](/quest/m1/release-audit/README.md) bounded the media +- The 2026-10-22 release audit bounded the media reorder queues in [moq-dev/moq#6453](https://github.com/moq-dev/moq/pull/6453) and fixed QMux copying in [moq-dev/moq#6381](https://github.com/moq-dev/moq/pull/6381). MKV diff --git a/quest/m2/release/cut.md b/quest/m2/release/cut.md index 42d5d658b6b..cd93641cc2d 100644 --- a/quest/m2/release/cut.md +++ b/quest/m2/release/cut.md @@ -45,20 +45,18 @@ From the 2026-10-25 pre-cut run: `MOQ_FFI_RANGE`) to the release that ships the `implementation` field from [moq-dev/moq#6775](https://github.com/moq-dev/moq/pull/6775). -The [2026-10-28 audit checkpoint](https://github.com/moq-dev/moq/pull/7176) -records source candidate -[main 487759f62a](https://github.com/moq-dev/moq/commit/487759f62a9ef39d916fe5b92df06dcff83478ce) -and shipping branch -[release 7fad611aae](https://github.com/moq-dev/moq/commit/7fad611aae4ef4e7fa856ed077a0bb467f2de9be). -The focused native suites (7,697 passes), JS checks, draft checks, OBS sanitizer -suite, and all 176 configured all-language interop cells passed. Dart's 45 tests -passed, but its checkout-local publish dry-run hit the staging ignore bug owned -by [moq-dev/moq#7117](https://github.com/moq-dev/moq/pull/7117). -The checkpoint does not select or approve the final cut tree. The published-wire comparison -is blocked by stale exception pins and the stable released CLI control failures -owned by [moq-dev/moq#7061](https://github.com/moq-dev/moq/pull/7061). Resolve -that existing prerequisite and repeat the candidate compatibility checks after -the documented release-to-main merge; source-only passes do not replace them. +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), +the focused native suites (7,697 passes), JS checks, draft checks, OBS sanitizer +suite, and all 176 configured all-language interop cells passed. At +[main 1ea4b1a2c3](https://github.com/moq-dev/moq/commit/1ea4b1a2c333527e25c8c29ee827290ed6e7d410), +`just test wire-compat` against moq-cli 0.14.10, moq-relay 0.17.10, and +`@moq/net` 0.4.7 passed every stable lite-01 through lite-06 session cell, +token, and catalog/container lane, with lite-07-wip skipped as its accepted +planned break. Dart's tests and both staged publish dry-runs passed with no +warnings. Neither main is the cut tree, and the full Linux gate was not run: +repeat these checks after the documented release-to-main merge. ## Required