diff --git a/quest/m0/README.md b/quest/m0/README.md index c239f0539e0..73d74b787b4 100644 --- a/quest/m0/README.md +++ b/quest/m0/README.md @@ -35,5 +35,6 @@ edge, which embeds moq-relay; it moved there on 2026-09-28. - [Benchmark sessions can't overlap on a host](/quest/m0/bench-host-lock.md) - a host-level lock keeps a second benchmark or profile session from measuring under the first's load - [A waiter's record limit changes no behavior](/quest/m0/kio-waiter-records.md) - two moq-net tests stop failing when kio's waiters remember 16 lists, and the limit becomes a tuning knob worth 3.9% per delivered group - [Backport the relay and auth hardening to release](/quest/m0/backport-release-hardening.md) - `release`'s relay refuses dot-segment paths, serves its HTTP routes with main's response hardening, and withdraws advertisements a peer starves of send credit; its auth client refuses redirects +- [Agent build output can't fill the desktop's disk](/quest/m0/agent-disk-budget.md) - finished agent checkouts reclaim their mbx targets automatically, so parallel agent builds stop filling the root disk - [HLS lists contributions](/quest/m0/hls-contributions.md) - an HLS master lists each catalog contribution's renditions, and a contribution's media is subscribed only while players fetch it - [@moq/net passes qmux stream limits through](/quest/m0/qmux-stream-limits.md) - Node and WebSocket JS clients stop failing near 100 live tracks diff --git a/quest/m0/agent-disk-budget.md b/quest/m0/agent-disk-budget.md new file mode 100644 index 00000000000..51a4748341d --- /dev/null +++ b/quest/m0/agent-disk-budget.md @@ -0,0 +1,56 @@ +# [S] Agent build output can't fill the desktop's disk + +## Goal + +The desktop host's agent checkouts reclaim their own build output, so parallel +agent builds don't fill the root disk and nobody has to run `mbx clean` by +hand. Done when a finished T3 or Claude Code checkout's managed target is +reclaimed automatically, without touching a checkout whose owner is still +using its outputs, and the PR records the free space over a busy day of +parallel agents. + +## Plan + +Planned 2026-11-04 with the maintainer, after the desktop's root disk filled +three times on 2026-11-04: to 244 MB, then twice to 5-14 GB while ten iterate +agents ran. Each time a human-run cleanup (`mbx gc`, `mbx clean`, worktree +removal, nix gc) freed the space. + +What holds the space (measured 2026-11-04): + +- `~/.cache/mbx/targets` held 430 GiB of managed targets for about 420 + checkouts, with single checkouts at 10-43 GiB. `mbx cache projects` attributes + nearly all of it to `~/worktrees/moq/t3-*` T3 thread worktrees, about 740 of + which existed. Cleaning the targets of those idle over six hours freed about + 150 GiB. +- The dotfiles `mbx/config.toml` sets `target.keep = ["/"]` and + `gc.min_free_size = "none"` on purpose (see dotfiles `docs/mbx.md`): a target + is kept while its checkout exists, because directly launched binaries hold no + mbx lease. So a target is only reclaimed when its checkout is removed. +- Dotfiles' `worktree-clean.service` removes finished checkouts, but it runs + only when started by hand and scans `~/work`, `~/.codex/worktrees`, + `~/.worktrees` and `/tmp`, not `~/worktrees`, where T3 creates its checkouts. + Most T3 worktrees there sit at a base commit or an unmerged branch, so even + the merged-PR rule finds few: on 2026-11-04 only 2 of 742 were at a merged PR + head. +- `docs/mbx.md` points quest repositories at `quest worktree` to reclaim + finished checkouts. It already keeps a per-user disk budget (`quest worktree + capacity`), runs commands supervised so it knows when a checkout's processes + stop, and removes a finished checkout without forcing. It manages the + checkouts it creates under `.worktrees/`: T3's bypass it, while Claude + Code's already go through `.claude/settings.json`'s `WorktreeCreate` and + `WorktreeRemove` hooks. Trace where finished Claude Code checkouts (such + as the 139 under `~/work/moq/.claude/worktrees` on 2026-11-04) leave their + targets before adding another cleaner; T3's need a new path either way. + +The work lands mostly in [kixelated/dotfiles](https://github.com/kixelated/dotfiles) +(cleaner scope, a schedule, mbx retention), and possibly in how T3 retires +thread worktrees. Weigh routing T3's checkouts through `quest worktree` +against extending the dotfiles cleaner. Keep the retention guarantee `docs/mbx.md` gives: never +reclaim a target whose owner is still using it. Find the signal that a T3 or +Claude Code checkout is finished (its thread or task ended, its PR merged or +closed, no process using it) and reclaim on that, rather than on idle time +alone. Also check `/tmp` test scratch, which filled the disk on 2026-11-04 +before the reboot. + +Public API: none. Wire: none. diff --git a/quest/m0/nightly/README.md b/quest/m0/nightly/README.md index 3ca8bfdbebe..bddd0482811 100644 --- a/quest/m0/nightly/README.md +++ b/quest/m0/nightly/README.md @@ -29,8 +29,10 @@ This README's own work, after the children: [audio jitter target](/quest/m1/audio-jitter-target/README.md), the macOS `receive_credit` timeout has no owner since its quest was deleted in [moq-dev/moq#6732](https://github.com/moq-dev/moq/pull/6732), so plan one under - [test flakes](/quest/m0/test-flakes-2/README.md) if it recurs, and the paired bench - threshold to [bench CI](/quest/m1/bench-ci.md). + [test flakes](/quest/m0/test-flakes-2/README.md) if it recurs, the paired bench + threshold to [bench CI](/quest/m1/bench-ci.md), and the moq-net bench + panics in "Rust bench targets" to + [moq-net's benches run to completion](/quest/m1/bench-test-mode.md). Public API: none. Wire: none. diff --git a/quest/m0/test-flakes-2/README.md b/quest/m0/test-flakes-2/README.md index 69a865614f3..bc2c05c9f72 100644 --- a/quest/m0/test-flakes-2/README.md +++ b/quest/m0/test-flakes-2/README.md @@ -94,3 +94,4 @@ Public API: none. Wire: none. - [IETF22 fallback](/quest/m0/test-flakes-2/ietf22-fallback.md) - an IETF22 browser viewer follows a fallback within the media lane's bound, keeping its session and its audio - [Browser buffering and Firefox audio-tone tests pass under load](/quest/m0/test-flakes-2/browser-media-flakes.md) - the two `harness.browser.ts` buffering tests and the Firefox player's audio-tone checks pass reliably in Interop and Nightly, each fixed at its cause under the epic's rules +- [reload.test.ts's close-drain tests pass under load](/quest/m0/test-flakes-2/reload-close-drain.md) - the drain cases in `js/net`'s reload tests pass in loaded multi-file runs diff --git a/quest/m0/test-flakes-2/reload-close-drain.md b/quest/m0/test-flakes-2/reload-close-drain.md new file mode 100644 index 00000000000..5c9ba445584 --- /dev/null +++ b/quest/m0/test-flakes-2/reload-close-drain.md @@ -0,0 +1,30 @@ +# [S] reload.test.ts's close-drain tests pass under load + +## Goal + +`closing the loop keeps the tokens until the session drains` in +`js/net/src/connection/reload.test.ts`, and any drain case sharing its race, +passes under load in multi-file runs, fixed at its cause under the +[epic](/quest/m0/test-flakes-2/README.md)'s rules. Done when the PR explains +the race and shows the repeated multi-file run that reproduced it now passing. + +## Plan + +Planned 2026-11-04 with the maintainer. While iterating +[moq-dev/moq#7090](https://github.com/moq-dev/moq/pull/7090), its agent saw +`closing the loop keeps the tokens until the session drains` fail once in a +four-file batch, then pass 13 runs with and without that PR's change. Its +failure output wasn't kept. The other drain cases in the file also run on a +manual clock and wait for the old session to close: `teardown by … drains the +session` shares its held-FIN fixture (`holdFins`), and `a URL change connects +the replacement while the old session drains until …` waits out a drain +deadline. Check them for the same race. + +Reproduce first: run `reload.test.ts` together with the rest of `js/net` under +CPU contention and `--rerun-each`, as the epic's earlier JS fixes did +([moq-dev/moq#7658](https://github.com/moq-dev/moq/pull/7658) fixed +`lite/connection.test.ts`'s close-drain fixture race that way). If it doesn't +reproduce in a reasonable number of runs, report that and the runs tried +rather than changing the test blind. + +Public API: none. Wire: none. diff --git a/quest/m1/README.md b/quest/m1/README.md index 51c072317b5..4fa043988c1 100644 --- a/quest/m1/README.md +++ b/quest/m1/README.md @@ -170,6 +170,8 @@ their end report, so an embedder's drain alert pages on real cuts. - [Rust releases publish through OIDC](/quest/m1/release-rs-oidc.md) - release-plz publishes with trusted publishing and `CARGO_REGISTRY_TOKEN` is revoked - [Main is no slower than release](/quest/m1/release-perf/README.md) - every benchmark where main is slower than release is fixed or accepted with a reason, checked by a final paired run before the cut - [Calibrate the paired nightly benchmarks](/quest/m1/bench-ci.md) - A/A and deliberately slowed Nightly runs set the paired comparison's threshold, so its regression alerts can be trusted +- [moq-net's benches run to completion](/quest/m1/bench-test-mode.md) - the `origin`, `session` and `memory` benches stop panicking in test mode and Nightly's measured run, and the maintainer picks where CI catches the next break +- [Catalog contributions become includes on the root's timeline](/quest/m1/catalog-include.md) - rename the unreleased `contributions` map before the cut and state that included broadcasts share the root's timeline - [Unbounded groups](/quest/m1/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 - [Poll-first async](/quest/m1/poll-first/README.md) - library crates compile no coroutines by default: async wrappers are generated from `poll_*` and gated behind an `async` feature diff --git a/quest/m1/bench-test-mode.md b/quest/m1/bench-test-mode.md new file mode 100644 index 00000000000..0c0952cfcc9 --- /dev/null +++ b/quest/m1/bench-test-mode.md @@ -0,0 +1,59 @@ +# [S] moq-net's benches run to completion + +## Goal + +Every moq-net bench target runs to completion, both in Criterion's test mode +(`cargo test -p moq-net --features fuzz --benches`) and in Nightly's measured +run, so a bench broken by a behaviour change fails where someone sees it. +Done when that command passes on main, Nightly's "Rust bench targets" job gets +past its `criterion net` step, and the PR records where CI catches the next +such break. + +## Plan + +Planned 2026-11-04 with the maintainer. While iterating +[moq-dev/moq#7729](https://github.com/moq-dev/moq/pull/7729) (authored tracks +default to zero max age, merged 2026-11-04), its agent ran the Rust benches in +test mode: `group` panicked under the new default and was fixed there, while +`origin`, `session` and `track` in `rs/moq-net/benches/` panicked on main +before the change. + +Nightly already runs these targets measured: `sh/gh/nightly.sh`'s `rust-bench` +step runs `criterion net --no-fail-fast -p moq-net --features fuzz --bench '*'`. +Scheduled run +[41656613559](https://github.com/moq-dev/moq/actions/runs/41656613559) +(2026-11-03, before #7729) failed it on four targets: + +- `origin` (`origin.rs:574` and `:576`, `winner changed`) in + `origin/reprice_duplicate/1d_1s_cost2`. +- `session` (`session/mod.rs:371` and `:492`, `Evicted`) in + `session_delivery_burst/moq-transport-22/...`. +- `memory` (`memory.rs:194`, `Evicted`) in `memory/streams=16/size=1048576`. +- `track` (`track.rs:1718` and `:1730`) in `track_range_widening/1r/64`, fixed + by [moq-dev/moq#7840](https://github.com/moq-dev/moq/pull/7840) (merged + 2026-11-04). + +Start by re-running every moq-net bench target on main, in test mode and in a +short measured run, since #7729's zero max-age default may have broken more. +Fix each at its cause: a bench whose expectation predates a deliberate +behaviour change is updated; a bench exposing a product bug gets the product +fix. `origin` and `session` need `fuzz`, and Cargo skips a bench whose +`required-features` are off without a word, so a test-mode run without it +passes while they're broken. + +Where CI catches the next break is the maintainer's call; ask in the PR. A +test-mode run in Nightly adds nothing, since the measured run covers the same +targets. Today the PR gate builds benches with their features but leaves them +unrun (`rs/justfile` above the `test` recipe; `.config/nextest.toml`'s +`default-filter = "not kind(bench)"`). Running them there, which would have +caught these before merge, reverses the +[Nightly epic](/quest/m0/nightly/README.md)'s 2026-10-24 decision that the +gate keeps skipping benches and a green Nightly catches them within a day. +Measure the test-mode run's cost before asking. + +Public API: none. Wire: none. + +## Related + +- [Nightly passes on main](/quest/m0/nightly/README.md) - this quest owns the moq-net bench panics in Nightly's "Rust bench targets" job +- [Calibrate the paired nightly benchmarks](/quest/m1/bench-ci.md) - calibrates the measurements; this quest only keeps the bench executables correct diff --git a/quest/m1/catalog-include.md b/quest/m1/catalog-include.md new file mode 100644 index 00000000000..f76e5422d75 --- /dev/null +++ b/quest/m1/catalog-include.md @@ -0,0 +1,41 @@ +# [S] Catalog contributions become includes on the root's timeline + +## Goal + +The hang catalog's root `contributions` map is renamed to say what it is: other +broadcasts the root includes, which share the root's PTS timeline and its +catalog-root settings (clock, join interval). The rename covers the Rust and JS +catalog types and schemas, `@moq/watch`'s `broadcast.contribution(effect, id)`, +the parity fixture, `doc/concept/hang.md`, and the draft's +`{#contributions}` section, with the shared-timeline rule stated in the draft. +Done before the next release cut, while the field is still unreleased. + +## Plan + +Planned 2026-11-04 with the maintainer, while settling +[moq-dev/moq#7836](https://github.com/moq-dev/moq/pull/7836)'s HLS listing. +The maintainer: keep contributions; "consider renaming them. Afaik the +use-case is moq.pro transcodes so they must share PTS and really everything in +the catalog root. Its why I wanted to call the section `include` originally." + +- ✅ **Contributions stay** (maintainer, 2026-11-04). Not chosen: removing them. +- ✅ **They share the root's timeline** (maintainer, 2026-11-04): consumers + such as moq-hls treat a contribution as on the root's PTS timeline and grid, + with no clock or interval comparison; a mismatch is the publisher's bug. + #7836 implements that for HLS. +- **The name is the open decision.** `include` is the maintainer's original + preference; confirm it (or another) in this quest's PR before renaming, then + rename the field, the types (`Contribution`), the watch method and the draft + section consistently in Rust, JS and the bindings that expose them. +- The field arrived in [moq-dev/moq#7312](https://github.com/moq-dev/moq/pull/7312) + (2026-11-01) and isn't on `release`, so no compatibility alias is needed and + the catalog `version` stays. +- Update the [HLS contributions](/quest/m0/hls-contributions.md) quest, or its + code if #7836 has landed, to the new name. + +Public API: breaking but unreleased (Rust `hang`, `@moq/hang`, `@moq/watch`). +Wire: the catalog JSON key changes; unreleased. + +## Related + +- [HLS lists contributions](/quest/m0/hls-contributions.md) - the HLS consumer settled with the shared-timeline rule