fix(voice): reuse the Linux audio capture across routing changes (#2440)

This commit is contained in:
Hampus
2026-09-04 01:00:32 +02:00
committed by GitHub
parent cc3a9c8613
commit 587324fa38
11 changed files with 206 additions and 1 deletions
@@ -556,6 +556,7 @@ export interface NativeAudioApi {
listAudibleApplications: () => Promise<Array<NativeAudioApplication>>;
resolveAudioRootPidForSource: (sourceId: string) => Promise<number | null>;
start: (options: NativeAudioStartOptions) => Promise<NativeAudioStartResult>;
setRule: (captureId: string, linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>) => Promise<boolean>;
stop: (captureId: string) => Promise<void>;
onFrame: (callback: (message: NativeAudioFrameMessage) => void) => () => void;
onEnd: (callback: (message: NativeAudioEndMessage) => void) => () => void;
@@ -75,6 +75,7 @@ import {
captureNativeAudioTrackForLinuxRouting,
commitNativeAudioBridgeReplacement,
disarmNativeAudio,
reconfigureLinuxNativeAudioRouting,
} from '@app/features/voice/utils/NativeAudioCaptureBridge';
import {SCREEN_SHARE_DEGRADATION_PREFERENCE} from '@app/features/voice/utils/ScreenShareOptions';
import {ScreenShareRollbackIncompleteError} from '@app/features/voice/utils/ScreenShareRollbackIncompleteError';
@@ -752,6 +753,14 @@ class VoiceEngineV2AppScreenShareExecutionAdapter extends Store {
const participant = room?.localParticipant;
if (!participant || !participant.isScreenShareEnabled) return false;
if (!linuxRule) return false;
if (options.replaceExisting !== true) {
const reconfigured = await reconfigureLinuxNativeAudioRouting(linuxRule, options);
if (reconfigured !== 'unsupported') {
this.syncLocalScreenShareAudioStateInternal(participant, true);
this.syncPersistedScreenShareAudioPreferenceInternal(participant);
return true;
}
}
const capturedTrack = await captureNativeAudioTrackForLinuxRouting(linuxRule, options);
if (!capturedTrack) return false;
let adopted = false;
@@ -18,6 +18,7 @@ import {
isValidAudioFrameMessage,
type NativeAudioBridgeHandle,
type NativeAudioBridgeStats,
nativeAudioRoutingRulesEqual,
patchTrackStopForCleanup,
replaceStreamAudioTrack,
shouldMixSelfWindowAudioIntoSystemCapture,
@@ -1363,6 +1364,35 @@ export function captureNativeAudioTrackForSelfWindow(): MediaStreamTrack | null
return track;
}
export type LinuxNativeAudioReconfigureResult = 'unchanged' | 'updated' | 'unsupported';
export async function reconfigureLinuxNativeAudioRouting(
linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>,
options: {includeSelfWindowAudio?: boolean} = {},
): Promise<LinuxNativeAudioReconfigureResult> {
const bridge = activeBridge;
if (!bridge) return 'unsupported';
if (bridge.linuxRule === undefined) return 'unsupported';
if (Boolean(bridge.includeSelfWindowAudio) !== (options.includeSelfWindowAudio === true)) return 'unsupported';
if (nativeAudioRoutingRulesEqual(bridge.linuxRule, linuxRule)) return 'unchanged';
const nativeAudioApi = getNativeAudioApi();
if (typeof nativeAudioApi?.setRule !== 'function') return 'unsupported';
let applied = false;
try {
applied = await nativeAudioApi.setRule(bridge.captureId, linuxRule);
} catch (error) {
logger.warn('Failed to reconfigure Linux native audio routing in place', {
captureId: bridge.captureId,
error,
});
return 'unsupported';
}
if (!applied) return 'unsupported';
if (activeBridge !== bridge) return 'unsupported';
activeBridge = {...bridge, linuxRule};
return 'updated';
}
export async function captureNativeAudioTrackForLinuxRouting(
linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>,
options: {includeSelfWindowAudio?: boolean} = {},
@@ -1376,7 +1406,12 @@ export async function captureNativeAudioTrackForLinuxRouting(
? await createNativeAudioBridgeWithSelfWindowAudio(result.captureId)
: await createNativeAudioBridge(result.captureId);
const cleanup = attachNativeAudioCleanup(new MediaStream([handle.track]), result.captureId, handle);
activeBridge = {captureId: result.captureId, cleanup};
activeBridge = {
captureId: result.captureId,
cleanup,
linuxRule,
includeSelfWindowAudio: options.includeSelfWindowAudio === true,
};
if (previousBridge && previousBridge.captureId !== result.captureId) {
supersededBridge = previousBridge;
}
@@ -0,0 +1,54 @@
// SPDX-License-Identifier: AGPL-3.0-or-later
import {
type NativeAudioRoutingRule,
nativeAudioRoutingRulesEqual,
} from '@app/features/voice/utils/native_audio_capture_bridge/shared';
import {describe, expect, it} from 'vitest';
const base: NativeAudioRoutingRule = {
include: [{'application.name': 'Firefox'}],
exclude: [{'node.name': 'fluxer'}],
ignoreDevices: true,
};
describe('nativeAudioRoutingRulesEqual', () => {
it('treats a rule as equal to itself and to a structural copy', () => {
expect(nativeAudioRoutingRulesEqual(base, base)).toBe(true);
expect(nativeAudioRoutingRulesEqual(base, {...base, include: [{'application.name': 'Firefox'}]})).toBe(true);
});
it('ignores the order of include and exclude patterns', () => {
const a: NativeAudioRoutingRule = {include: [{'application.name': 'A'}, {'application.name': 'B'}]};
const b: NativeAudioRoutingRule = {include: [{'application.name': 'B'}, {'application.name': 'A'}]};
expect(nativeAudioRoutingRulesEqual(a, b)).toBe(true);
});
it('separates a missing list from an empty one only when the contents differ', () => {
expect(nativeAudioRoutingRulesEqual({include: []}, {})).toBe(true);
expect(nativeAudioRoutingRulesEqual({include: [{'application.name': 'A'}]}, {})).toBe(false);
});
it('detects a changed selection, which is the case that must restart routing', () => {
expect(nativeAudioRoutingRulesEqual(base, {...base, include: [{'application.name': 'Chromium'}]})).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, {...base, include: []})).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, {...base, exclude: []})).toBe(false);
});
it('detects a changed tuning flag', () => {
expect(nativeAudioRoutingRulesEqual(base, {...base, ignoreDevices: false})).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, {...base, onlySpeakers: true})).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, {...base, onlyDefaultSpeakers: true})).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, {...base, ignoreInputMedia: true})).toBe(false);
});
it('treats an undefined flag as false rather than as a difference', () => {
expect(nativeAudioRoutingRulesEqual({include: []}, {include: [], onlySpeakers: false})).toBe(true);
});
it('never reports equality when a rule is missing', () => {
expect(nativeAudioRoutingRulesEqual(undefined, base)).toBe(false);
expect(nativeAudioRoutingRulesEqual(base, undefined)).toBe(false);
expect(nativeAudioRoutingRulesEqual(undefined, undefined)).toBe(true);
});
});
@@ -40,6 +40,43 @@ export interface NativeAudioBridgeHandle {
export interface ActiveNativeAudioBridge {
captureId: string;
cleanup: (stopRemote?: boolean) => Promise<void>;
linuxRule?: NativeAudioRoutingRule;
includeSelfWindowAudio?: boolean;
}
export type NativeAudioRoutingRule = NonNullable<NativeAudioStartOptions['linuxRule']>;
function routingPatternsEqual(a: NativeAudioRoutingRule['include'], b: NativeAudioRoutingRule['include']): boolean {
const left = a ?? [];
const right = b ?? [];
if (left.length !== right.length) return false;
const serialize = (entry: Record<string, unknown>): string =>
JSON.stringify(
Object.entries(entry)
.filter(([, value]) => typeof value === 'string')
.sort(([x], [y]) => x.localeCompare(y)),
);
const serializedLeft = left.map(serialize).sort();
const serializedRight = right.map(serialize).sort();
return serializedLeft.every((entry, index) => entry === serializedRight[index]);
}
export function nativeAudioRoutingRulesEqual(
a: NativeAudioRoutingRule | undefined,
b: NativeAudioRoutingRule | undefined,
): boolean {
if (a === b) return true;
if (!a || !b) return false;
if (!routingPatternsEqual(a.include, b.include)) return false;
if (!routingPatternsEqual(a.exclude, b.exclude)) return false;
return (
Boolean(a.ignoreDevices) === Boolean(b.ignoreDevices) &&
Boolean(a.ignoreInputMedia) === Boolean(b.ignoreInputMedia) &&
Boolean(a.ignoreVirtual) === Boolean(b.ignoreVirtual) &&
Boolean(a.workaround) === Boolean(b.workaround) &&
Boolean(a.onlySpeakers) === Boolean(b.onlySpeakers) &&
Boolean(a.onlyDefaultSpeakers) === Boolean(b.onlyDefaultSpeakers)
);
}
export interface GeneratorAudioTrack extends MediaStreamTrack {
+1
View File
@@ -653,6 +653,7 @@ export interface NativeAudioApi {
listAudibleApplications(): Promise<Array<NativeAudioApplication>>;
resolveAudioRootPidForSource(sourceId: string): Promise<number | null>;
start(options: NativeAudioStartOptions): Promise<NativeAudioStartResult>;
setRule(captureId: string, linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>): Promise<boolean>;
stop(captureId: string): Promise<void>;
getRoutingGraph(captureId?: string): Promise<NativeAudioRoutingGraphResult>;
onFrame(callback: (message: NativeAudioFrameMessage) => void): () => void;
+2
View File
@@ -120,6 +120,8 @@ export declare class ProcessLoopback extends EventEmitter implements ProcessLoop
start(): void;
setRoutingRule(target: {linuxRule: JsRoutingRule} | number, options?: {includeProcessTree?: boolean}): boolean;
routingGraph(): RoutingGraph | null;
stop(): Promise<void>;
@@ -235,6 +235,19 @@ function patternsEqual(a, b) {
return true;
}
function routingRulesEqual(a, b) {
if (a === b) return true;
if (!a || !b) return false;
if (!patternsEqual(a.include, b.include)) return false;
if (!patternsEqual(a.exclude, b.exclude)) return false;
if (!patternsEqual(a.workaround, b.workaround)) return false;
return (
Boolean(a.ignoreDevices) === Boolean(b.ignoreDevices) &&
Boolean(a.onlySpeakers) === Boolean(b.onlySpeakers) &&
Boolean(a.onlyDefaultSpeakers) === Boolean(b.onlyDefaultSpeakers)
);
}
class ProcessLoopback extends EventEmitter {
constructor(target, options = {}) {
super();
@@ -291,6 +304,28 @@ class ProcessLoopback extends EventEmitter {
}
}
setRoutingRule(target, options = {}) {
if (this.closed || !this.started) return false;
if (typeof this.capture?.setRule !== 'function') return false;
let nextRule;
try {
nextRule = routingRuleFromTarget(target, options);
} catch {
return false;
}
if (routingRulesEqual(nextRule, this.rule)) return true;
let applied = false;
try {
applied = this.capture.setRule(nextRule) !== false;
} catch {
return false;
}
if (!applied) return false;
this.rule = nextRule;
this.options = options;
return true;
}
refreshRuleForLateChildren() {
if (this.closed || !this.started || this.targetPid === null) return;
let nextRule;
+1
View File
@@ -430,6 +430,7 @@ export interface NativeAudioApi {
listAudibleApplications: () => Promise<Array<NativeAudioApplication>>;
resolveAudioRootPidForSource: (sourceId: string) => Promise<number | null>;
start: (options: NativeAudioStartOptions) => Promise<NativeAudioStartResult>;
setRule: (captureId: string, linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>) => Promise<boolean>;
stop: (captureId: string) => Promise<void>;
getRoutingGraph: (captureId?: string) => Promise<NativeAudioRoutingGraphResult>;
onFrame: (callback: (message: NativeAudioFrameMessage) => void) => () => void;
+28
View File
@@ -40,6 +40,7 @@ interface NativeCaptureInstance {
removeListener(event: 'closed', listener: () => void): this;
start(): Promise<void> | void;
stop(): Promise<void> | void;
setRoutingRule?: (target: {linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>}) => boolean;
routingGraph?: () => VirtmicRoutingGraph | null;
}
@@ -631,6 +632,23 @@ async function stopCaptureById(
await stopActiveSession(session, reason, detail);
}
function reconfigureCaptureById(
senderId: number,
captureId: string,
linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>,
): boolean {
const session = activeSessions.get(captureId);
if (!session || session.sender.id !== senderId) return false;
if (session.finalized || session.stopping) return false;
if (typeof session.capture.setRoutingRule !== 'function') return false;
try {
return session.capture.setRoutingRule({linuxRule});
} catch (error) {
logger.warn('Failed to reconfigure native audio routing in place', {captureId, error});
return false;
}
}
async function makeRoomForSenderSession(senderId: number): Promise<void> {
const senderSessions = activeSessionIdsBySenderId.get(senderId);
if (!senderSessions) return;
@@ -885,6 +903,15 @@ export function registerNativeAudioHandlers(): void {
(event, options: NativeAudioStartOptions): Promise<NativeAudioStartResult> =>
startNativeAudioCapture(event.sender, options),
);
ipcMain.handle(
'native-audio:set-rule',
(event, captureId: unknown, linuxRule: NativeAudioStartOptions['linuxRule']): boolean => {
if (typeof captureId !== 'string' || !isValidLinuxRule(linuxRule)) {
return false;
}
return reconfigureCaptureById(event.sender.id, captureId, linuxRule);
},
);
ipcMain.handle('native-audio:stop', async (event, captureId: string): Promise<void> => {
const session = activeSessions.get(captureId);
if (!session || session.sender.id !== event.sender.id) {
@@ -904,6 +931,7 @@ export function cleanupNativeAudio(): void {
ipcMain.removeHandler('native-audio:list-applications');
ipcMain.removeHandler('native-audio:resolve-root-pid');
ipcMain.removeHandler('native-audio:start');
ipcMain.removeHandler('native-audio:set-rule');
ipcMain.removeHandler('native-audio:stop');
ipcMain.removeHandler('native-audio:get-routing-graph');
handlersRegistered = false;
+2
View File
@@ -684,6 +684,8 @@ const api: ElectronAPI = {
ipcRenderer.invoke('native-audio:resolve-root-pid', sourceId),
start: (options: NativeAudioStartOptions): Promise<NativeAudioStartResult> =>
ipcRenderer.invoke('native-audio:start', options),
setRule: (captureId: string, linuxRule: NonNullable<NativeAudioStartOptions['linuxRule']>): Promise<boolean> =>
ipcRenderer.invoke('native-audio:set-rule', captureId, linuxRule),
stop: (captureId: string): Promise<void> => ipcRenderer.invoke('native-audio:stop', captureId),
getRoutingGraph: (captureId?: string): Promise<NativeAudioRoutingGraphResult> =>
ipcRenderer.invoke('native-audio:get-routing-graph', captureId),