feat!: keep the peer's close code over qmux and raw QUIC, on web-transport-trait 0.5 by kixelated · Pull Request #4262 · moq-dev/moq

kixelated · 2026-09-26T17:23:05Z

Problem

A client loses the application close code its server sent (Session::abort(App(4011)), Request::reject(Reject::App(4011))) over WebSocket/qmux (ws://) and raw QUIC (moqt://). Instead it sees Transport("connection closed"), the close as text inside Transport(..), or its own Session(Internal). WebTransport is fine. Both bugs are upstream.

Approach

Fixed at the source:

fix(qmux): keep the first close reason web-transport#399: qmux keeps the first close reason. The reader's EOF / WebSocket Close, a second peer close, and a local close() no longer overwrite it. accept_* / open_* / recv_datagram report it instead of a bare Closed.
fix(web-transport-moq): report a raw QUIC peer's close code noq#11: web-transport-moq decodes a raw QUIC peer's ApplicationClosed code directly, and close() after the connection closed keeps the peer's reason.

Releasing qmux from web-transport main also ships #390's set_priority(i32), so the trait goes to 0.5 (moq-dev/web-transport#401, moq-dev/noq#13). This PR follows it:

rs/moq-tokio/tests/close_code.rs runs the issue's three cases (abort after accept, abort then drop, reject during handshake), 5 fresh connections each over https://, moqt://, and ws://. It fails on the previous pins for every moqt:// and ws:// connection.
Pins the released crates: web-transport-trait 0.5, qmux 0.6, web-transport-iroh 0.8, web-transport-wasm 0.7, and the noq fork's trait-0.5 major (moq-noq-proto, moq-noq-udp, web-transport-moq 2.0). Cargo.lock carries a single trait version.
The set_priority impls take i32. moq-net keeps its u8 send orders and widens them at the trait boundary.

@moq/qmux already keeps the first close, so no JS change.

Impact

Breaking (Rust): moq_net::web_transport_trait (re-exported) is now 0.5. SendStream::set_priority takes i32. moq_tokio::transport::Session<S> requires a trait-0.5 session. Hence dev.
Wire: none.

Alternatives

A qmux 0.5.2 cherry-picked from the tag would have kept this on main. The maintainer chose to release main with the trait bump.
Widening moq-net's own send order to i32 is left for later. Nothing needs it yet.

Follow-ups

Raw QUIC sessions in web-transport-moq still map stream RESET/STOP codes through the HTTP/3 space.

Closes #4249

(Written by Opus 5.5)

🤖 Generated with Claude Code

Runs the #4249 cases (abort after accept, abort then drop, reject during the handshake) over https://, moqt://, and ws://. Fails until the qmux and web-transport-moq fixes are released and pinned. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

SendStream::set_priority takes an i32 send order in trait 0.5. moq-net keeps its u8 send orders and widens them at the trait boundary. Pins qmux 0.6, web-transport-iroh 0.8, and web-transport-wasm 0.7, which implement the new trait. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

kixelated · 2026-09-26T22:05:31Z

Decisions

The noq fork ships one 1.3.2 bundling noq#11, noq#12, and noq#13.
web-transport-moq stops re-exporting generic.

(Written by Opus 5.5)

chatgpt-codex-connector · 2026-09-27T00:45:41Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-28T02:41:59.780204Z	`b998832`	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d43956dfcf

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-09-27T00:48:47Z

 pollster = "1.0"
 pulldown-cmark = { version = "0.13", default-features = false }
-qmux = { version = "0.5.1", default-features = false }
+qmux = { version = "0.6", default-features = false }


Commit the updated dependency lockfile

The manifest now requires qmux 0.6 and the new WebTransport versions, but the unchanged Cargo.lock still pins qmux 0.5.1, web-transport-trait 0.4.0, and the previous adapter versions. The required quest validation invokes cargo run --locked, and cargo check --help defines --locked as asserting that Cargo.lock remains unchanged, so just check and CI cannot resolve this workspace after these manifest changes. Update and commit Cargo.lock with the released dependencies.

AGENTS.md reference: quest/AGENTS.md:L67-L72

Useful? React with 👍 / 👎.

# Conflicts: # quest/m1/README.md

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b998832473

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-09-28T02:41:57Z

+						let Ok(session) = request.ok().await else { return };
+						session.abort(moq_net::Error::App(CODE));
+						if let Case::Abort = case {
+							tokio::time::sleep(Duration::from_secs(5)).await;


Hold the abort session with synchronization

Case::Abort is supposed to cover aborting while the session handle remains alive, but this wall-clock delay makes that property depend on the client observing the close within five seconds. Under a stalled or heavily loaded CI worker, the timer can expire first, dropping the handle and collapsing this case into AbortThenDrop, so a retained-handle regression can go untested. Hold the session until the client observes closure, or use an explicit latch or pending task instead of sleeping; the Rust testing rules also require time-dependent async tests to pause Tokio time.

AGENTS.md reference: rs/AGENTS.md:L51-L54

Useful? React with 👍 / 👎.

kixelated · 2026-09-28T07:14:13Z

Assessment

The goal is right and fixing close codes upstream (qmux first-close-wins, raw-QUIC ApplicationClosed decoding) is the correct approach rather than papering over losses in moq-net. The trait-0.5 bump, set_priority(i32) widens at the boundary, and the three-case / three-transport close_code regression are well scoped for dev.

But CI currently fails the PR's own regression: moq-tokio::close_code abort, abort_then_drop, and reject all lose the app code on ws:// with exactly MoqNet(Transport("connection closed")) — the #4249 symptom this PR claims to fix. https/moqt may be fine; the qmux path is not green on the pinned crates. Do not land a breaking trait bump whose end-to-end proof is red.

Recommendation: ITERATE — get ws:// close codes through to SessionError::App(4011) under the new pins (or confirm the qmux 0.6 release actually includes the first-close fix), then re-run the close_code suite green before merge.

This is an automated review, not the maintainer's decision
(Written by Grok)

This was referenced Sep 26, 2026

fix(qmux): keep the first close reason moq-dev/web-transport#399

Merged

chore!: bump web-transport-trait to 0.5 and its implementors moq-dev/web-transport#401

Merged

kixelated and others added 3 commits September 26, 2026 11:46

quest: claim close-codes

eadaca2

kixelated force-pushed the quest/m1/close-codes branch from b8fcc96 to c12415d Compare September 26, 2026 18:56

kixelated changed the title ~~fix: keep the peer's close code over qmux and raw QUIC~~ feat!: keep the peer's close code over qmux and raw QUIC, on web-transport-trait 0.5 Sep 26, 2026

kixelated changed the base branch from main to dev September 26, 2026 18:56

chore: pin web-transport-moq 1.3.2 for the raw QUIC close code

d43956d

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

kixelated mentioned this pull request Sep 26, 2026

quest: plan follow-ups from the current PR wave #4304

Merged

kixelated marked this pull request as ready for review September 27, 2026 00:45

chatgpt-codex-connector Bot reviewed Sep 27, 2026

View reviewed changes

Merge remote-tracking branch 'origin/dev' into pr-4262

b998832

# Conflicts: # quest/m1/README.md

chatgpt-codex-connector Bot reviewed Sep 28, 2026

View reviewed changes

Navigation Menu

Uh oh!

feat!: keep the peer's close code over qmux and raw QUIC, on web-transport-trait 0.5 - #4262

feat!: keep the peer's close code over qmux and raw QUIC, on web-transport-trait 0.5#4262
kixelated wants to merge 5 commits into
devfrom
quest/m1/close-codes

kixelated commented Sep 26, 2026 •

edited

Loading

Uh oh!

kixelated commented Sep 26, 2026

Uh oh!

chatgpt-codex-connector Bot commented Sep 27, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Sep 27, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Sep 28, 2026

Uh oh!

kixelated commented Sep 28, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

kixelated commented Sep 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Approach

Impact

Alternatives

Follow-ups

Uh oh!

kixelated commented Sep 26, 2026

Decisions

Uh oh!

chatgpt-codex-connector Bot commented Sep 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Sep 27, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Sep 28, 2026

Choose a reason for hiding this comment

Uh oh!

kixelated commented Sep 28, 2026

Assessment

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

kixelated commented Sep 26, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Sep 27, 2026 •

edited

Loading