fix: Hold the 2-way audio backchannel open for the shortest possible time (#2697)

The card claimed a camera's ONVIF audio backchannel in two places that
had
nothing to do with a call: the capability probe at camera init
(`&microphone`),
and a pre-armed `sendonly` audio transceiver on every live WebRTC offer.
Merely
looking at a dashboard occupied the camera's speaker line. Outbound
audio now
travels on its own audio-only WebRTC connection, opened when a call is
answered
and closed when it ends.

- The backchannel is claimed only for the duration of a call. Idle
viewing
  claims nothing.
- Two-way audio now works in `mse`, `mp4` and `mjpeg` modes (note: the
outbound
  audio still traverses WebRTC).
- No renegotiation and no video blink at call start or end.
- A call that cannot carry audio now reports it and ends, instead of
showing a
  live microphone that goes nowhere.
- `live.microphone.always_connected` is now purely about the browser
microphone
  permission prompt.
- Call setup measured at 66ms (LAN) and ~260ms (cellular) for ICE and
DTLS, plus
  ~300ms for `go2rtc` to open an RTSP backchannel.

Verified against a live Frigate + `go2rtc` instance, and by unit tests
at 100%
coverage.

 - Closes #2691
 - Closes #2039
 - Closes #2178

Ref #2299 -- the probe no longer opens a backchannel, but it still runs
per
camera on every load and reconnect, and still dials the camera on the
direct-`go2rtc` path. Caching remains to be done.

Ref AlexxIT/go2rtc#1860 -- once a call has opened a backchannel,
`go2rtc` keeps
that media set up on the camera's RTSP session for the life of the
producer.

Diagnoses #2678
This commit is contained in:
Dermot Duffy
2026-08-21 20:13:58 -07:00
committed by GitHub
parent 2d7c86c93c
commit 11cc543406
65 changed files with 2251 additions and 1045 deletions
@@ -568,7 +568,6 @@ describe('MSEStreamSource', () => {
expect(source.getCapabilities()).toEqual({
supportsPause: true,
hasAudio: false,
has2WayAudio: false,
});
});
@@ -579,7 +578,6 @@ describe('MSEStreamSource', () => {
expect(setupResult.source.getCapabilities()).toEqual({
supportsPause: true,
hasAudio: true,
has2WayAudio: false,
});
});
@@ -595,7 +593,6 @@ describe('MSEStreamSource', () => {
expect(setupResult.source.getCapabilities()).toEqual({
supportsPause: true,
hasAudio: false,
has2WayAudio: false,
});
});
@@ -5,17 +5,17 @@ import type {
StreamSourceContext,
VideoStreamTarget,
} from '../../../../../../src/components-lib/live/providers/go2rtc-experimental/types';
import { flushPromises } from '../../../../../test-utils';
import {
FakeMediaStream,
FakeMediaStreamTrack,
FakeRTCPeerConnection,
FakeStreamSourceChannel,
} from '../test-utils';
type FakeMediaStreamTrack,
} from '../../../../../go2rtc/test-utils';
import { flushPromises } from '../../../../../test-utils';
import { FakeStreamSourceChannel } from '../test-utils';
// @vitest-environment jsdom
describe('WebRTCStreamSource', () => {
const setup = (options?: { microphoneStream?: FakeMediaStream | null }) => {
const setup = () => {
const video = document.createElement('video');
const channel = new FakeStreamSourceChannel();
const loadedCallback = vi.fn();
@@ -29,13 +29,10 @@ describe('WebRTCStreamSource', () => {
const pc = new FakeRTCPeerConnection();
const createPeerConnection = vi.fn(() => pc.asPeerConnection());
const microphoneErrorCallback = vi.fn();
const source = new WebRTCStreamSource(context, {
createPeerConnection,
createMediaStream: (tracks) =>
new FakeMediaStream(tracks as unknown as FakeMediaStreamTrack[]).asMediaStream(),
microphoneStream: options?.microphoneStream?.asMediaStream() ?? null,
microphoneErrorCallback,
});
return {
@@ -44,7 +41,6 @@ describe('WebRTCStreamSource', () => {
createPeerConnection,
failedCallback,
loadedCallback,
microphoneErrorCallback,
pc,
source,
video,
@@ -60,27 +56,14 @@ describe('WebRTCStreamSource', () => {
});
describe('transceivers', () => {
it('should pre-arm a sendonly audio transceiver and recvonly video and audio', () => {
it('should offer inbound video and audio only', () => {
const { source, pc } = setup();
source.start();
expect(pc.transceivers).toHaveLength(3);
expect(pc.transceivers[0].direction).toBe('sendonly');
expect(pc.transceivers[1].direction).toBe('recvonly');
expect(pc.transceivers[2].direction).toBe('recvonly');
// Kind-only pre-arm: no track, so no getUserMedia and no permission prompt.
expect(pc.transceivers[0].sender.track).toBeNull();
});
it('should pre-arm with the current microphone track', () => {
const micTrack = new FakeMediaStreamTrack('audio');
const { source, pc } = setup({
microphoneStream: new FakeMediaStream([micTrack]),
});
source.start();
expect(pc.transceivers[0].sender.track).toBe(micTrack);
expect(pc.transceivers.map((transceiver) => transceiver.direction)).toEqual([
'recvonly',
'recvonly',
]);
});
});
@@ -431,6 +414,17 @@ describe('WebRTCStreamSource', () => {
expect(source.getPeerConnection()).toBe(pc.asPeerConnection());
});
it('should report its media capabilities', () => {
const { source, pc } = setup();
source.start();
pc.fireConnectionStateChange('connected');
expect(source.getCapabilities()).toEqual({
supportsPause: true,
hasAudio: expect.any(Boolean),
});
});
it('should report webrtc technology', () => {
const { source } = setup();
@@ -462,139 +456,5 @@ describe('WebRTCStreamSource', () => {
hasAACAudio: false,
});
});
it('should report 2-way audio capability once a mic track is armed', () => {
const micTrack = new FakeMediaStreamTrack('audio');
const { source, pc } = setup({
microphoneStream: new FakeMediaStream([micTrack]),
});
source.start();
pc.fireConnectionStateChange('connected');
expect(source.getCapabilities().has2WayAudio).toBe(true);
expect(source.getCapabilities().supportsPause).toBe(true);
});
});
describe('setMicrophoneStream', () => {
it('should do nothing for an unchanged stream', async () => {
const stream = new FakeMediaStream([new FakeMediaStreamTrack('audio')]);
const { source, pc } = setup({ microphoneStream: stream });
source.start();
await source.setMicrophoneStream(stream.asMediaStream());
expect(pc.getMicrophoneTransceiver().sender.replaceTrack).not.toHaveBeenCalled();
});
it('should replace the outbound track without renegotiating', async () => {
const { source, pc } = setup();
source.start();
const newTrack = new FakeMediaStreamTrack('audio');
await source.setMicrophoneStream(new FakeMediaStream([newTrack]).asMediaStream());
expect(pc.getMicrophoneTransceiver().sender.replaceTrack).toHaveBeenCalledWith(
newTrack,
);
});
it('should clear the outbound track for a null stream', async () => {
const stream = new FakeMediaStream([new FakeMediaStreamTrack('audio')]);
const { source, pc } = setup({ microphoneStream: stream });
source.start();
await source.setMicrophoneStream(null);
expect(pc.getMicrophoneTransceiver().sender.replaceTrack).toHaveBeenCalledWith(
null,
);
});
it('should do nothing before there is a peer connection', async () => {
const { source, microphoneErrorCallback } = setup();
await source.setMicrophoneStream(
new FakeMediaStream([new FakeMediaStreamTrack('audio')]).asMediaStream(),
);
expect(microphoneErrorCallback).not.toHaveBeenCalled();
});
it.each([
[
'what the browser said when the rejection has a message',
new DOMException('The peer connection is closed', 'InvalidStateError'),
'The peer connection is closed',
],
[
'the rejection type when there is no message to quote',
new DOMException('', 'InvalidStateError'),
'InvalidStateError',
],
['nothing when the rejection is not an object', 'nope', undefined],
[
'nothing when the rejection describes itself with neither',
{ message: 5, name: 7 },
undefined,
],
] as const)(
'should report %s when a current replaceTrack rejects',
async (_summary, rejection, expected) => {
const { source, pc, microphoneErrorCallback } = setup();
source.start();
pc.getMicrophoneTransceiver().sender.replaceTrack.mockRejectedValue(rejection);
await source.setMicrophoneStream(
new FakeMediaStream([new FakeMediaStreamTrack('audio')]).asMediaStream(),
);
expect(microphoneErrorCallback).toHaveBeenCalledWith(expected);
},
);
it('should not fail the stream source when the microphone cannot attach', async () => {
const { source, pc, failedCallback } = setup();
source.start();
pc.getMicrophoneTransceiver().sender.replaceTrack.mockRejectedValue(
new DOMException('replace failed', 'InvalidStateError'),
);
await source.setMicrophoneStream(
new FakeMediaStream([new FakeMediaStreamTrack('audio')]).asMediaStream(),
);
// The inbound video is unaffected by an outbound audio failure, so the
// source must keep running rather than failing over to another one.
expect(failedCallback).not.toHaveBeenCalled();
});
it('should not report a rejection when detaching the microphone', async () => {
const stream = new FakeMediaStream([new FakeMediaStreamTrack('audio')]);
const { source, pc, microphoneErrorCallback } = setup({
microphoneStream: stream,
});
source.start();
pc.getMicrophoneTransceiver().sender.replaceTrack.mockRejectedValue(
new DOMException('The peer connection is closed', 'InvalidStateError'),
);
await source.setMicrophoneStream(null);
// Ignore the error, the user is not trying to be heard anyway.
expect(microphoneErrorCallback).not.toHaveBeenCalled();
});
it('should ignore a stale replaceTrack rejection after stop', async () => {
const { source, pc, microphoneErrorCallback } = setup();
source.start();
let rejectReplace: (reason: Error) => void = () => {};
pc.getMicrophoneTransceiver().sender.replaceTrack.mockReturnValue(
new Promise((_resolve, reject) => {
rejectReplace = reject;
}),
);
const promise = source.setMicrophoneStream(
new FakeMediaStream([new FakeMediaStreamTrack('audio')]).asMediaStream(),
);
source.stop();
rejectReplace(new Error('replace failed'));
await promise;
expect(microphoneErrorCallback).not.toHaveBeenCalled();
});
});
});