T2.4 — Relay conformance: Tier A (bitrate ceiling)

Status: Pending Review Agent: Kimi Code CLI Started: 2026-05-11T17:20Z Completed: 2026-05-11T17:35Z Commit: 846c98e PRD: ../PRD-relay-conformance.md

What I changed

crates/wzp-relay/src/conformance.rs (new) — Conformance meter + violation enum:
- Violation enum: BitrateExceeded, PacketRateExceeded, TimestampDrift.
- ConformanceMeter with 1-second sliding window tracking bytes_in_window.
- ceiling_bps(codec) — nominal * 3 * 115 / 100 with floor of 2 kbps.
- observe() returns Err(Violation::BitrateExceeded) when window bytes exceed ceiling_bps / 8.
crates/wzp-relay/src/lib.rs — Added pub mod conformance;.
crates/wzp-relay/src/metrics.rs — Added conformance_violations: IntCounterVec (label: violation_type).
crates/wzp-relay/src/room.rs — Wired ConformanceMeter into both forwarding loops:
- run_participant_plain and run_participant_trunked each create a per-participant meter.
- On violation: logs tracing::warn! + bumps Prometheus counter.
- Observe-only — packets are never dropped.
crates/wzp-client/src/featherchat.rs — Added missing TransportFeedback match arm (back-fill from T2.1).

Why these choices

Using a plain struct with &mut self (no atomics/mutex) is correct because each participant runs in exactly one async recv task. The meter is never shared across threads.

The * 3 factor accounts for FEC 2.0 (200% overhead = 3× total bitrate). The * 115 / 100 adds a 15% safety margin. The 2 kbps floor prevents ComfortNoise (0 bps nominal) from having a zero ceiling.

Deviations from the task spec

Task example shows parking_lot::Mutex<Instant>. We don't have parking_lot in the relay crate, and it's unnecessary for a single-threaded async loop. Used plain Instant field instead.

Verification output

$ cargo test -p wzp-relay conformance
running 4 tests
test conformance::tests::bitrate_exceeded_for_opus24k ... ok
test conformance::tests::ceiling_bps_floor ... ok
test conformance::tests::small_packets_stay_within_ceiling ... ok
test conformance::tests::window_resets_after_one_second ... ok

test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 76 filtered out; finished in 0.00s

$ cargo test -p wzp-relay
running 86 tests
...(all 86 pass)...

test result: ok. 86 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s

Test summary

Tests added: 4
- bitrate_exceeded_for_opus24k — 1 MB/s payload declared as Opus24k correctly returns BitrateExceeded
- small_packets_stay_within_ceiling — 100 small packets stay under limit
- window_resets_after_one_second — window rollover works
- ceiling_bps_floor — ComfortNoise gets 2 kbps floor
Tests modified: 0
wzp-relay test count: 86 (was 82 before T2.4)
cargo clippy -p wzp-relay --lib: pass (no new warnings)
cargo fmt --all -- --check: pass

Risks / follow-ups

Tier B (packet-rate) and Tier C (timestamp drift) are reserved for T2.5.
Currently observe-only. Future tasks may add drop/throttle behavior.

Reviewer checklist (filled in by reviewer)

Code matches PRD intent
Verification output is real
No backward-incompat surprises
Tests cover the new behavior
Approved

3.3 KiB Raw Blame History Unescape Escape