Repository navigation
fix(quic): sync quinn fixes through 2697de1e - #7287
Conversation
Quest-Claim: 67b46f3f2239c1ed
Record f233510c and 6c69f9f4 as already held or not applying, and port 623a0369's specific_v6_addr test with the driver tests binding through moq_sock::bind. Roll the recurring quest into m2. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed commit: 4ee50ec5a043a38b74f642baf5d1a4409af100c2.
No actionable bugs found in the inspected sync. Reusing moq_sock::bind::udp removes duplicated socket setup, preserves wildcard dual-stack behavior and keeps a specific IPv6 address v6-only; the added connection test exercises the affected driver path. The newly documented dispositions match the inspected fork code: accept staleness uses the non-optional handshake timeout, and reset-stream stop credit uses the remaining reliable-size obligation.
Direction is sound: distinguish already-held/inapplicable fixes from new ports, record exact upstream checkpoints, and carry the recurring audit forward behind the release cut rather than importing unrelated features.
Verification limits: GitHub-only static review of the diff, upstream commit range and relevant fork code, with an independent pass. No tests, Windows runs or builds were executed. This is not an exhaustive security-advisory or cross-platform correctness certification; author-reported validation was not independently reproduced.
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 10 minutes. View limit details
No actionable comments were generated in the recent review.
|
|
Landing via
(Written by Claude Opus 5.5) |
|
quest land removed this pull request from the queue: Conflicts with the base: quest/m2/quinn-sync.md |
|
Codex land fixer: # refs can deploy to.
- environment: publish
+ environment: bot
timeout-minutes: 20
steps:
- uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
merged
result 100644 7c05faf4c9e5f74340793c0a8b56f54578ab3f64 .github/workflows/smoke.yml
our 100644 c3e66ca17a0f37621d88c70418005bd38844cacf .github/workflows/smoke.yml
@@ -492,7 +492,7 @@
contents: read
packages: read
# The published client image and moq-dev relay both follow Git HEAD. Building
- # the Rust relay is the long pole; the 12 protocol cases are bounded and quick.
+ # the Rust relay is the long pole; the protocol cases are bounded and quick.
timeout-minutes: 45
steps:
@@ -523,10 +523,6 @@
dirname "$shim" >> "$GITHUB_PATH"
- name: Moxygen interoperability
- # Current moq-dev HEAD negotiates successfully but fails moxygen's
- # announce-subscribe case. Keep the full diagnostic visible without
- # failing the job; remove this once that protocol mismatch is fixed.
- continue-on-error: true
run: just moxygen
interop-runner:
@@ -559,8 +555,9 @@
path: runner-upstream
persist-credentials: false
- # Any failed pair or setup step fails the job. A complete run still writes
- # its report first, since a failing run is the one worth looking at.
+ # Only moq-dev clients against this checkout's relay gate the job, plus any
+ # setup step; the other pairs are reported only. A complete run still
+ # writes its report first, since a failing run is the one worth looking at.
- name: Run upstream Docker matrices
run: just runner "$GITHUB_WORKSPACE/runner-upstream" "$GITHUB_WORKSPACE/interop-runner-report"
merged
result 100644 76a091624d168b9ab060bcfb766bc7a6605055ac .github/workflows/ts.yml
our 100644 5d97604def1612ec78cb3deb0b0359e69fcc2d49 .github/workflows/ts.yml
@@ -20,8 +20,8 @@
ts:
name: TS validation
runs-on: ubuntu-latest
- # `main` and its merge queue write the Rust cache; see .github/actions/rust-cache.
- environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/gh-readonly-queue/main/')) && 'cache-write' || '' }}
+ # Hosted `main` jobs write the Rust cache; see .github/actions/rust-cache.
+ environment: ${{ github.ref == 'refs/heads/main' && 'cache-write' || '' }}
timeout-minutes: 60
steps:
- name: Checkout
merged
result 100644 052326a850059a2e8a39f78dbf9ece18ea9fa3db .github/workflows/wasm.yml
our 100644 a2d96cbf5c531b44577f072288598800001f8a9f .github/workflows/wasm.yml
@@ -42,11 +42,10 @@
jobs:
wasm:
name: WASM
- # ARM64 like check.yml; the harness builds the relay natively, which the
- # merge queue's Check and Test publish.
+ # ARM64 like check.yml; the harness builds the relay natively.
runs-on: ubuntu-24.04-arm
- # `main` and its merge queue write the Rust cache; see .github/actions/rust-cache.
- environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/gh-readonly-queue/main/')) && 'cache-write' || '' }}
+ # Hosted `main` jobs write the Rust cache; see .github/actions/rust-cache.
+ environment: ${{ github.ref == 'refs/heads/main' && 'cache-write' || '' }}
timeout-minutes: 60
steps:
merged
result 100644 d03b0c48c1d581d732eb57eac1a67ad0679ff580 .github/workflows/x11.yml
our 100644 7d01aaa79e4ecec746271bda00a186bf367859f9 .github/workflows/x11.yml
@@ -16,8 +16,8 @@
capture:
name: SHM and GetImage
runs-on: ubuntu-24.04-arm
- # `main` and its merge queue write the Rust cache; see .github/actions/rust-cache.
- environment: ${{ (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/gh-readonly-queue/main/')) && 'cache-write' || '' }}
+ # Hosted `main` jobs write the Rust cache; see .github/actions/rust-cache.
+ envi
thread 'codex-main' (74) panicked at /rustc/48a229ceaefd4985c50990b14116b6d856af0985/library/std/src/io/stdio.rs:1166:9:
failed printing to stderr: Resource temporarily unavailable (os error 11)
stack backtrace:
0: __rustc::rust_begin_unwind
1: core::panicking::panic_fmt
2: std::io::stdio::_eprint
3: <codex_exec::event_processor_with_human_output::EventProcessorWithHumanOutput as codex_exec::event_processor::EventProcessor>::process_server_notification
4: codex_exec::run_exec_session::{closure#0}
5: <tracing::instrument::Instrumented<codex_exec::run_exec_session::{closure#0}> as core::future::future::Future>::poll
6: codex_exec::run_main::{closure#0}
7: codex::cli_main::{closure#0}
8: <tokio::runtime::park::CachedParkThread>::block_on::<codex_arg0::run_main_with_arg0_guard<codex::main::{closure#0}, codex::main::{closure#0}::{closure#0}>::{closure#0}>
9: tokio::runtime::context::runtime::enter_runtime::<<tokio::runtime::scheduler::multi_thread::MultiThread>::block_on<codex_arg0::run_main_with_arg0_guard<codex::main::{closure#0}, codex::main::{closure#0}::{closure#0}>::{closure#0}>::{closure#0}, core::result::Result<(), anyhow::Error>>
10: <tokio::runtime::runtime::Runtime>::block_on::<codex_arg0::run_main_with_arg0_guard<codex::main::{closure#0}, codex::main::{closure#0}::{closure#0}>::{closure#0}>
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
quest: fixer agent exited 101
(written by Codex) |
Main moved quinn-sync into the vendor-sync epic, which now owns recurrence. Drop the PR's m2/quinn-sync.md and rewrite quest/m1/vendor-sync/quinn.md to the remainder: a2c55b36, the quinn#2481 BBR3 check, and judging read_datagrams. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Merged origin/main to clear the conflict. moq-dev/moq#7422 had moved this quest into The failing (Written by Claude Opus 5.5) |
|
quest land removed this pull request from the queue: the batch check failed (exit 1). * [new tag] docker-v1 -> docker-v1
* [new tag] github-show -> github-show
* [new tag] push -> push
* [new tag] v0.10.0 -> v0.10.0
* [new tag] v0.11.0 -> v0.11.0
* [new tag] v0.12.0 -> v0.12.0
* [new tag] v0.13.0 -> v0.13.0
* [new tag] v0.13.1 -> v0.13.1
* [new tag] v0.14.0 -> v0.14.0
* [new tag] v0.14.1 -> v0.14.1
* [new tag] v0.15.1 -> v0.15.1
* [new tag] v0.16.0 -> v0.16.0
* [new tag] v0.17.0 -> v0.17.0
* [new tag] v0.18.0 -> v0.18.0
* [new tag] v0.19.0 -> v0.19.0
* [new tag] v0.19.1 -> v0.19.1
* [new tag] v0.19.2 -> v0.19.2
* [new tag] v0.19.3 -> v0.19.3
* [new tag] v0.19.4 -> v0.19.4
* [new tag] v0.19.5 -> v0.19.5
* [new tag] v0.19.6 -> v0.19.6
* [new tag] v0.20.0 -> v0.20.0
* [new tag] v0.21.0 -> v0.21.0
* [new tag] v0.21.1 -> v0.21.1
* [new tag] v0.22.0 -> v0.22.0
* [new tag] v0.23.0 -> v0.23.0
* [new tag] v0.24.0 -> v0.24.0
* [new tag] v0.24.1 -> v0.24.1
* [new tag] v0.24.2 -> v0.24.2
* [new tag] v0.24.3 -> v0.24.3
* [new tag] v0.25.0 -> v0.25.0
* [new tag] v0.25.1 -> v0.25.1
* [new tag] v0.25.2 -> v0.25.2
* [new tag] v0.25.3 -> v0.25.3
* [new tag] v0.26.0 -> v0.26.0
* [new tag] v0.26.1 -> v0.26.1
* [new tag] v0.27.0 -> v0.27.0
* [new tag] v0.27.1 -> v0.27.1
* [new tag] v0.27.2 -> v0.27.2
* [new tag] v0.27.3 -> v0.27.3
* [new tag] v0.28.0 -> v0.28.0
* [new tag] v0.28.1 -> v0.28.1
* [new tag] v0.28.2 -> v0.28.2
* [new tag] v0.28.3 -> v0.28.3
* [new tag] v0.29.0 -> v0.29.0
* [new tag] v0.29.1 -> v0.29.1
* [new tag] v0.29.2 -> v0.29.2
* [new tag] v0.29.3 -> v0.29.3
* [new tag] v0.29.4 -> v0.29.4
* [new tag] v0.29.5 -> v0.29.5
* [new tag] v0.30.0 -> v0.30.0
* [new tag] v0.31.0 -> v0.31.0
* [new tag] v0.31.1 -> v0.31.1
* [new tag] v0.31.2 -> v0.31.2
* [new tag] v0.32.0 -> v0.32.0
* [new tag] v0.32.1 -> v0.32.1
* [new tag] v0.32.2 -> v0.32.2
* [new tag] v0.4.0 -> v0.4.0
* [new tag] v0.5.0 -> v0.5.0
* [new tag] v0.6.0 -> v0.6.0
* [new tag] v0.6.1 -> v0.6.1
* [new tag] v0.7.0 -> v0.7.0
* [new tag] v0.7.1 -> v0.7.1
* [new tag] v0.7.2 -> v0.7.2
* [new tag] v0.8.0 -> v0.8.0
* [new tag] v0.9.0 -> v0.9.0
derivation evaluated to /nix/store/nvij27kw5rh5323b8idz8whhaqn49kc4-moq-relay-0.17.2.drv
checking derivation packages.x86_64-linux.default...
derivation evaluated to /nix/store/8njiymqk67rqxixlszn9734gjq417l0f-moq-all.drv
checking derivation packages.x86_64-linux.uniffi-bindgen-dart...
derivation evaluated to /nix/store/1v1hdl7dhiwdis0y2lmh6vjx149is362-uniffi-bindgen-dart-0.3.1-kixelated.6+v0.32.0.drv
checking derivation packages.x86_64-linux.uniffi-bindgen-cpp...
derivation evaluated to /nix/store/im9pz4z9gbnxq8cgg736sqc7d5y07nc3-uniffi-bindgen-cpp-0.11.0-kixelated.7+v0.32.2.drv
checking derivation packages.x86_64-linux.moq...
derivation evaluated to /nix/store/smqzfr9i5yz34lx7plmp7cf7jkha5myb-moq-cli-0.14.2.drv
checking derivation packages.x86_64-linux.moq-cli...
derivation evaluated to /nix/store/hv61q4ncyjdmv18plpyi91gajc3k7n0z-moq.drv
checking derivation packages.x86_64-linux.moq-bench...
derivation evaluated to /nix/store/wgx1wsmlph9lv0if39k3xla0xi826gcx-moq-bench-0.0.1.drv
checking derivation packages.x86_64-linux.moq-boy...
error: interrupted by the user
ci: test gh failed
ci: seconds per module
295 check py
57 test py
526 check rs
195 test rs
230 check rs_shapes (cancelled)
418 check c
364 check cpp
63 check obs_compile
406 check js
67 test js
28 check markdown
4 check gh
27 test gh (failed)
154 check kt
18 check flake (cancelled)
952 wall time
error: recipe `ci` failed on line 107 with exit code 1
. |
|
codex / Codex land fixer: -
-## Goal
-
-moq-mux parses and builds avcC and hvcC records with `mp4_atom::Avcc` and
-`mp4_atom::Hvcc`, already its dependency, instead of about 330 hand-written
-lines, or a comment at the hand-written code names the input mp4-atom
-mishandles.
-
-## Plan
-
-Found by the 2026-10-19 dependency audit. `rs/moq-mux/src/codec/h264/mod.rs`
-has `Avcc::parse`, `build_avcc`, `catalog_avcc`, `avcc_params`, and
-`read_param_sets`; `codec/h265/mod.rs` has `build_hvcc` and `hvcc_params`.
-The tests already check them against mp4-atom. The fMP4 importer decodes an
-`mp4_atom::Avcc`, re-encodes it, and re-parses it with the hand-written parser
-(`container/fmp4/import.rs`, `init_h264`), while its H.265 path reads the
-`mp4_atom::Hvcc` fields directly.
-
-Check first for a leniency reason, such as how mp4-atom handles the
-high-profile extension or a short record from a real encoder; keep the
-hand-written path only for a reproduced case and test it. `moq_mux` exports
-`codec::h264::Avcc`, so replacing or removing it is a breaking change;
-prefer keeping the type and changing only how it is
-filled when that is simpler. No wire change.
-
-## Related
-
-- [H.264 parsing with maintained crates](/quest/m2/h264-reader.md) - the SPS side of the same codec config
diff --git a/quest/m2/README.md b/quest/m2/README.md
index f93399e86..4e19cf61d 100644
--- a/quest/m2/README.md
+++ b/quest/m2/README.md
@@ -218,7 +218,6 @@ lite-07; bitrate enforcement is not part of this release.
- [Task supervision](/quest/m2/task-supervision.md) - the auth decider loop, stats publisher, and moq-tokio's token and WebSocket internals are supervised like #5447's tasks
- [Auth serve bound address](/quest/m2/auth-serve-bound-addr.md) - `moq auth serve` logs the port it actually bound
-- [RTMP client reports a publish or play error](/quest/m2/rtmp-client-error.md) - a refused publish or play surfaces its reason
- [H.264 parsing with maintained crates](/quest/m2/h264-reader.md) - moq-mux parses SPS VUI with h264-reader and shares one bit reader across codec config parsers
- [avcC and hvcC through mp4-atom](/quest/m2/mp4-atom-config.md) - moq-mux stops hand-writing the decoder config records its mp4-atom dependency already parses
- [Relay metrics through prometheus-client](/quest/m2/relay-prometheus-client.md) - `/metrics` renders through the maintained client with the same series
diff --git a/quest/m2/rtmp-client-error.md b/quest/m2/rtmp-client-error.md
deleted file mode 100644
index e50058d8e..000000000
--- a/quest/m2/rtmp-client-error.md
+++ /dev/null
@@ -1,16 +0,0 @@
-# [XS] RTMP client reports a publish or play error
-
-## Goal
-
-`moq_rtmp::Client` returns a server's `_error` on publish or play with its
-code and description, as it already does on connect. Today the `_error` has
-no outstanding transaction, so the session raises
-`UnknownTransactionResultReceived`, `dial.rs` ignores it, and the client ends
-with "rtmp server closed before publish was accepted" once the server closes.
-
-## Plan
-
-Match the publish and play `_error` in `rs/moq-rtmp`'s client and surface it
-through `dial.rs`. Test against a server that answers a publish with `_error`
-and one that answers a play with `_error`, asserting the code and description
-in both.
codex
I reproduced the dispatcher failures on base: the two targeted tests pass with `CHECK_FAIL_FAST` cleared and fail with it set. The test harness inherits that gate setting. No PR-specific fix is indicated, so I’ll requeue all eight unchanged and record the cause.
exec
/bin/bash -lc "python3 - <<'PY'
import json, pathlib, subprocess
numbers=[7397,7401,7418,7441,7287,7256,7336,7352]
reason='healthy; batch failure is base test isolation: sh/dispatch-query.test.py inherits CHECK_FAIL_FAST from the landing gate; targeted base tests pass with it cleared'
result=[dict(number=n,action='requeue',reason=reason) for n in numbers]
p=pathlib.Path('quest-land-result.json')
p.write_text(json.dumps(result,indent=2)+'\\n')
assert len(json.loads(p.read_text()))==8
expected={7397:'1bac738fa',7401:'5f4f6db97',7418:'08201ee65',7441:'353bbd9d0',7287:'976bf7422',7256:'ec9560ca5',7336:'28b51f575',7352:'6fc4f8c90'}
for n,prefix in expected.items():
head=subprocess.check_output(['git','rev-parse',f'pr-{n}'],text=True).strip()
assert head.startswith(prefix),(n,head)
print('Validated 8 requeue entries and unchanged pinned branch tips.')
PY
git status --short" in /work
succeeded in 25ms:
Validated 8 requeue entries and unchanged pinned branch tips.
?? quest-land-result.json
codex
Wrote [quest-land-result.json](/work/quest-land-result.json): all eight PRs requeued unchanged.
The failure comes from base’s dispatcher tests inheriting `CHECK_FAIL_FAST`. Targeted tests pass with it cleared and reproduce the failure when set.
All PR branches remain intact; the result file is uncommitted.
tokens used
64,429
Wrote [quest-land-result.json](/work/quest-land-result.json): all eight PRs requeued unchanged.
The failure comes from base’s dispatcher tests inheriting `CHECK_FAIL_FAST`. Targeted tests pass with it cleared and reproduce the failure when set.
All PR branches remain intact; the result file is uncommitted.
fixer result written
(written by codex / Codex) |
Problem
moq-quic is a hard fork of quinn, so cargo-deny cannot see quinn fixes. This is the first recurring sync since the fork point
7616e6b2. The Plan named three upstream fixes missing from the ledger.Approach
Checked quinn
mainthrough2697de1e,0.11.xthrough4b84837d, and quinn's security advisories. Every advisory was published by 2026-09-30, which the fork point already covers.f233510c(a disabled idle timeout abandons accepts): does not apply. The fork's stale-accept check (rs/moq-quic/src/endpoint.rs:577) useshandshake_timeout, a plainDurationthat cannot be disabled.6c69f9f4(stop credits a reset stream twice): already held.Recv::unread_creditcredits only what the reset still owes, andstop_after_reset_credits_oncetests it.623a0369(specific IPv6 endpoints made dual-stack): production binds already had the fix inmoq_sock::bind(moq-dev/moq#5339). The driver tests'bindhelper still had the old logic. It now callsmoq_sock::bind::udp, and upstream'sspecific_v6_addrtest is ported. The bug only shows on Windows, and the Windows CI job runs this test.2697de1e(read_datagrams) is a new feature, not a fix, so it is skipped. The others in the range were dependency bumps or are already in the ledger.The README ledger lists all three fixes and records where the next sync starts.
Since this PR opened, moq-dev/moq#7422 moved the quest to
quest/m1/vendor-sync/quinn.mdunder the vendor-sync epic. That epic now handles recurrence and addsa2c55b36(Apple IPv4 source address, macOS-only) and a check of the quinn#2481 BBR3 PR. The merge from main follows the newer layout. It drops this PR'squest/m2/quinn-sync.mdand its m2 link, and rewritesquest/m1/vendor-sync/quinn.mddown to what's left:a2c55b36, the quinn#2481 check, and judgingread_datagrams. It also records the new last-checked commits and what this run learned.Choices made without asking:
read_datagramsis judged for/quest-planin the remaining run.a2c55b36is left to the remaining run, not ported here. It touches only Apple code paths inmoq_sock::udp, which can't be tested on this Linux host.Impact
None: no public API or wire change. The only code change is in tests.
Alternatives
Fixing the old dual-stack logic inside the driver's test helper in place, instead of reusing
moq_sock::bind::udp. That would keep a second copy of socket setup that could drift again.Follow-ups
None.
Review
Codex, gpt-6.1-sol, medium effort (
quest review --base origin/main, new session on the merge). Reviewed head976bf742245f2dadfb6e50001d23485ee81037bf, round 1: no findings. An earlier round on4ee50ec5(gpt-6-astra,--deep) also had no findings.(Written by Claude Opus 5.5)
🤖 Generated with Claude Code