From c8464e075cc66704c0f9210113a3b852121c5d2a Mon Sep 17 00:00:00 2001
From: cai <cai@nbcai.cc>
Date: Sat, 08 Aug 2026 18:55:07 +0800
Subject: [PATCH] test: cover participant binding rejection
---
src/main.rs | 131 +++++++++++++++++++++++++++++++++++++------
1 files changed, 111 insertions(+), 20 deletions(-)
diff --git a/src/main.rs b/src/main.rs
index c3fa805..9b08934 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -30,7 +30,7 @@
options::TrackPublishOptions,
prelude::{
DataPacket, LocalAudioTrack, LocalTrack, ParticipantIdentity, RemoteAudioTrack,
- RemoteTrack, Room, RoomEvent, RoomOptions,
+ RemoteParticipant, RemoteTrack, Room, RoomEvent, RoomOptions,
},
};
use reqwest::Client;
@@ -737,6 +737,10 @@
})
}
+fn is_bound_user_participant(identity: &str, expected: Option<&str>) -> bool {
+ expected.is_none_or(|value| identity == value)
+}
+
async fn observe_user_audio_events(
mut events: UnboundedReceiver<RoomEvent>,
call_id: String,
@@ -771,25 +775,15 @@
publication: _,
participant,
} => {
- if user_participant_identity
- .as_deref()
- .is_some_and(|expected| participant.identity().to_string() != expected)
- {
+ if !is_bound_user_participant(
+ &participant.identity().to_string(),
+ user_participant_identity.as_deref(),
+ ) {
warn!(call_id = %call_id, trace_id = %trace_id,
metadata_status = "wrong_participant",
"runtime helper ignored non-user audio participant");
continue;
}
- let ingress_metadata =
- match AudioIngressMetadata::from_participant(&participant.attributes()) {
- Ok(value) => value,
- Err(reason) => {
- warn!(call_id = %call_id, trace_id = %trace_id,
- metadata_status = "invalid", reason = reason,
- "runtime helper ignored invalid audio ingress metadata");
- None
- }
- };
let participant_alias = redact(&participant.identity().to_string());
let track_sid_alias = redact(&track.sid().to_string());
let track_name = track.name();
@@ -801,9 +795,6 @@
track_sid_alias = %track_sid_alias,
track_name = %track_name,
track_source = %track_source,
- metadata_status = if ingress_metadata.is_some() { "bound" } else { "absent" },
- metadata_source = ingress_metadata.as_ref().map(|_| "controlled_fixture"),
- metadata_sequence_present = ingress_metadata.is_some(),
"runtime helper user_track_subscribed"
);
spawn_user_audio_frame_observer(
@@ -818,6 +809,7 @@
turn_bridge_config.clone(),
http.clone(),
sink.clone(),
+ participant,
);
}
RoomEvent::TrackSubscribed {
@@ -2929,6 +2921,7 @@
turn_bridge_config: TurnBridgeConfig,
http: Client,
sink: Arc<BotAudioOutputSink>,
+ participant: RemoteParticipant,
) -> JoinHandle<()> {
tokio::spawn(async move {
let mut stream = NativeAudioStream::new(
@@ -3054,14 +3047,14 @@
if !was_in_speech && is_in_speech {
let turn_id = format!("turn-{:04}", vad.turn_index);
- match RealtimeAsrUpload::start(
+ match RealtimeAsrUpload::start_with_participant_attributes(
http.clone(),
turn_bridge_config.realtime_asr_config(),
&call_id,
&trace_id,
&turn_id,
&vad.speech_samples,
- ingress_metadata.as_ref(),
+ || participant.attributes(),
) {
Ok(upload) => {
info!(
@@ -3962,6 +3955,104 @@
use std::collections::HashSet;
#[test]
+ fn production_vad_session_boundary_reads_updated_attributes() {
+ let config = SimpleVadConfig {
+ rms_threshold: 0.001,
+ peak_threshold: 0.01,
+ start_frames: 2,
+ end_silence_ms: 100,
+ min_speech_ms: 1,
+ max_turn_ms: 1_000,
+ initial_ignore_ms: 0,
+ };
+ let mut vad = SimpleVad::new(config);
+ let samples = vec![1_000i16; 160];
+ let frame = AudioFrame {
+ data: samples.as_slice().into(),
+ sample_rate: 16_000,
+ num_channels: 1,
+ samples_per_channel: 160,
+ };
+ let mut attributes = std::collections::HashMap::from([
+ (
+ "inputSourceCategory".to_string(),
+ "controlled_fixture".to_string(),
+ ),
+ (
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ ),
+ ]);
+ let mut starts = Vec::new();
+ for (session_index, sequence) in [(1, "fixture-01"), (2, "fixture-02")] {
+ let was_in_speech = vad.in_speech;
+ vad.observe_frame(
+ "call-001",
+ "trace-001",
+ "participant",
+ "track",
+ session_index * 2 - 1,
+ 1_000 * session_index,
+ &frame,
+ );
+ vad.observe_frame(
+ "call-001",
+ "trace-001",
+ "participant",
+ "track",
+ session_index * 2,
+ 1_000 * session_index + 10,
+ &frame,
+ );
+ let is_in_speech = vad.in_speech;
+ assert!(!was_in_speech && is_in_speech);
+ attributes.insert("clientFixtureSequence".to_string(), sequence.to_string());
+ let metadata = AudioIngressMetadata::from_participant(&attributes)
+ .expect("valid participant attributes")
+ .expect("controlled fixture metadata");
+ let session_line = asr_realtime::session_start_line(
+ "call-001",
+ "trace-001",
+ &format!("turn-{session_index:04}"),
+ "nonce-001",
+ Some(&metadata),
+ )
+ .expect("session start line");
+ let session_json: serde_json::Value =
+ serde_json::from_slice(&session_line).expect("session start json");
+ assert_eq!(sequence, session_json["clientFixtureSequence"]);
+ starts.push(metadata.client_fixture_sequence);
+ vad.reset_current_turn();
+ }
+ assert_eq!(vec!["fixture-01", "fixture-02"], starts);
+ attributes.insert("inputSourceCategory".to_string(), "other".to_string());
+ assert!(AudioIngressMetadata::from_participant(&attributes).is_err());
+ assert!(
+ AudioIngressMetadata::from_participant(&std::collections::HashMap::new())
+ .expect("missing attributes is absent")
+ .is_none()
+ );
+ attributes.insert(
+ "clientFixtureSequence".to_string(),
+ "fixture-01".to_string(),
+ );
+ assert!(AudioIngressMetadata::from_participant(&attributes).is_err());
+ }
+
+ #[test]
+ fn production_observer_rejects_wrong_participant_before_vad_session() {
+ assert!(!is_bound_user_participant(
+ "participant-other",
+ Some("participant-user")
+ ));
+ assert!(is_bound_user_participant(
+ "participant-user",
+ Some("participant-user")
+ ));
+ assert!(is_bound_user_participant("participant-any", None));
+ }
+
+ #[test]
fn reply_chunk_marker_state_emits_turn_first_once_and_later_segment_first_once() {
let mut state = ReplyChunkMarkerState::default();
--
Gitblit v1.9.3