Open
Conversation
* handles publish_namespace correctly and register to the coordinator * adds subscribe_namespace request handling * updates moq-pub which correct publish_namespace and handles incoming subscribe request at application level
When a TrackReader in the cache has been closed (due to publisher disconnect, error, etc.), new subscribers would receive the stale cached reader instead of triggering a fresh upstream subscription. This change: - Adds TrackReader::is_closed() to detect closed/dropped tracks - Updates TracksReader::subscribe() to check liveness before returning cached tracks, evicting stale entries when found Includes unit tests for both the bug scenario and normal deduplication.
Replace is_some() + unwrap() patterns with if-let to satisfy clippy::unnecessary_unwrap lint in Rust 1.93+. Affected files: - moq-pub/src/media.rs - moq-relay-ietf/src/relay.rs
A standardized test client for MoQT interoperability testing. Implements test scenarios based on the moq-interop-runner framework. Test cases: - setup-only: Connect, complete SETUP exchange, close gracefully - announce-only: Announce namespace, receive OK - subscribe-error: Subscribe to non-existent track, expect error - announce-subscribe: Publisher announces, subscriber subscribes - subscribe-before-announce: Out-of-order setup handling - publish-namespace-done: Announce then send PUBLISH_NAMESPACE_DONE Also enables TLS_DISABLE_VERIFY environment variable support in moq-native-ietf for containerized testing workflows. Usage: moq-test-client --relay https://localhost:4443 moq-test-client --relay https://localhost:4443 --test setup-only moq-test-client --list
fix: clippy warnings fix: clippy warning fix: cargo fmt
* track is_closed was checking for state as dropped, but we consume track writer when creating stream/subgroup/datagram which drops the track writer and thus mark the state dropped, because of that tracks.subscribe() was not returning the reader even though reader was active. Fixed by checking is_closed on the reader_mode.
…rrors Bug: When multiple tracks (audio + video) are played concurrently in moq-sub, each track's recv_group() writes objects individually to a shared stdout via Arc<Mutex<O>>. Since fMP4 objects alternate between moof (metadata) and mdat (media data) boxes, a concurrent task can acquire the lock between a moof and its corresponding mdat, causing interleaving: [video moof] → [audio moof] → [audio mdat] → [video mdat] ffmpeg then reads video moof + audio mdat as a single segment, parsing MP4 traf box bytes (0x74726166) as H.264 NAL unit sizes, producing: Invalid NAL unit size (1953653094 > 413) Error splitting the input into NAL units This was not observed on the main branch because the relay's announce- based subscribe path (TracksReader::subscribe) creates per-subscriber Track pairs forwarded through the consumer.rs event loop, delivering objects at the natural publish rate with low interleave probability. On feat/publish, the relay's TrackInfo/subscribe_upstream architecture uses shared tracks (OnceLock) with inline forward_upstream(), creating a tighter setup window that increases concurrent delivery probability between audio and video tracks. Fix: Buffer moof objects in recv_group() and write moof+mdat as a single atomic operation under one lock acquisition, preventing any interleaving regardless of relay delivery timing.
Upgrade from web-transport v0.3 / web-transport-quinn v0.3 to web-transport v0.10.4 / web-transport-quinn v0.11.8 (crates.io). The v0.11.x quinn backend has proper support for: - H3 SETTINGS: ENABLE_DATAGRAM, WEBTRANSPORT_MAX_SESSIONS, ENABLE_CONNECT_PROTOCOL, and Chrome-compat deprecated settings - Capsule protocol for graceful session close - Sub-protocol negotiation via wt-available-protocols headers - GREASE filtering in H3 settings - Session::raw() for clean raw-QUIC (moqt:// scheme) fallback API changes adapted: - ALPN constant changed from &[u8] to &str (.as_bytes() added) - accept(conn) -> Request::accept(conn) then .ok().await - connect_with(conn, url) -> Session::connect(conn, request) with ConnectRequest carrying the MoQT version as subprotocol - conn.into() -> Session::raw(conn, url, ConnectResponse::default()) - Three error variants collapsed into one web_transport::Error - Session methods now take &self instead of &mut self - read_buf() returns Option<usize> instead of bool - read_chunk() renamed to read() The subprotocol negotiation is critical for cross-implementation WebTransport interop: per draft-15+, version negotiation uses ALPN for raw QUIC and wt-available-protocols for WebTransport. Without offering the version as a subprotocol, servers like quiche-moq and moxygen that enforce WT-Available-Protocols reject the session.
…riber RequestOk and RequestError are bidirectional messages used as responses to requests from either side: PUBLISH_NAMESPACE (publisher-originated) and SUBSCRIBE_NAMESPACE (subscriber-originated) both receive RequestOk as their acknowledgment. The run_recv dispatch tried Publisher first, so RequestOk always matched the Publisher enum and was routed exclusively to the subscriber handler. When a client sent PUBLISH_NAMESPACE and the relay responded with RequestOk, the client's subscriber handler couldn't find the request ID (it belongs to the publisher) and silently dropped the message. This caused the announce-only and publish-namespace-done interop tests to time out waiting for a response that was received but never delivered. Fix: handle RequestOk and RequestError as a special case before the normal dispatch, trying both subscriber and publisher handlers so the response reaches whichever side owns that request ID.
…aft-16 Upgrade web-transport crates to v0.10 and add WT subprotocol negotiation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.