Files
cinny/src/app/plugins/call/CallControl.ts
T
jared 49d9410e3a fix(calls): resolve EC mute hang, robust camera focus, PiP NaN guard (N122/N123/N126)
- N122: setMediaState resolves on EC's transport ACK instead of waiting for a
  DeviceMute state-echo that EC may elide or skip during teardown — which
  previously stranded the promise forever and silently skipped the initial
  deafen state + first StateUpdate on join. Dropped the single-slot
  mediaStatePromiseResolver; onMediaState remains the authoritative sync path.
- N123: focusCameraParticipant now waits for a spotlight videoTile to mount via
  a MutationObserver (with a 600ms hard-timeout fallback) instead of a fixed
  2-frame delay that EC's React commit can exceed on slower devices.
- N126: PiP position restored from localStorage is shape+finiteness validated,
  so corrupt data can't feed NaN into the position math (invalid 'NaNpx' CSS).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-28 09:17:19 -04:00

425 lines
12 KiB
TypeScript

import { ClientWidgetApi } from 'matrix-widget-api';
import { EventEmitter } from 'events';
import { CallControlState } from './CallControlState';
import { ElementMediaStateDetail, ElementMediaStatePayload, ElementWidgetActions } from './types';
export enum CallControlEvent {
StateUpdate = 'state_update',
}
export class CallControl extends EventEmitter implements CallControlState {
private state: CallControlState;
private call: ClientWidgetApi;
private iframe: HTMLIFrameElement;
private bodyMutationObserver: MutationObserver;
private controlMutationObserver: MutationObserver;
private _pipMode = false;
private get document(): Document | undefined {
return this.iframe.contentDocument ?? this.iframe.contentWindow?.document;
}
private get screenshareButton(): HTMLElement | undefined {
const screenshareBtn = this.document?.querySelector(
'[data-testid="incall_screenshare"]',
) as HTMLElement | null;
return screenshareBtn ?? undefined;
}
private get leaveButton(): Element | undefined {
const leaveBtn = this.document?.querySelector('[data-testid="incall_leave"]');
return leaveBtn ?? undefined;
}
private get settingsButton(): HTMLElement | undefined {
// EC 0.20.1: settings button moved to bottom-left; fall back to bottom-center.
const settingsButtonLeft = this.document?.querySelector(
'[data-testid="settings-bottom-left"]',
) as HTMLButtonElement | undefined;
const settingsButtonCenter = this.document?.querySelector(
'[data-testid="settings-bottom-center"]',
) as HTMLButtonElement | undefined;
return settingsButtonLeft ?? settingsButtonCenter ?? undefined;
}
private get reactionsButton(): HTMLElement | undefined {
// EC 0.20.1: reactions/raise-hand button sits just before the leave button.
const reactionsButton = this.leaveButton?.previousElementSibling as HTMLElement | null;
return reactionsButton ?? undefined;
}
private get spotlightButton(): HTMLInputElement | undefined {
const spotlightButton = this.document?.querySelector(
'input[value="spotlight"]',
) as HTMLInputElement | null;
return spotlightButton ?? undefined;
}
private get gridButton(): HTMLInputElement | undefined {
const gridButton = this.document?.querySelector(
'input[value="grid"]',
) as HTMLInputElement | null;
return gridButton ?? undefined;
}
constructor(state: CallControlState, call: ClientWidgetApi, iframe: HTMLIFrameElement) {
super();
this.state = state;
this.call = call;
this.iframe = iframe;
this.bodyMutationObserver = new MutationObserver(this.onBodyMutation.bind(this));
this.controlMutationObserver = new MutationObserver(this.onControlMutation.bind(this));
}
public getState(): CallControlState {
return this.state;
}
public get microphone(): boolean {
return this.state.microphone;
}
public get video(): boolean {
return this.state.video;
}
public get sound(): boolean {
return this.state.sound;
}
public get screenshare(): boolean {
return this.state.screenshare;
}
public get spotlight(): boolean {
return this.state.spotlight;
}
public get screenshareAudioMuted(): boolean {
return this.state.screenshareAudioMuted;
}
public async applyState() {
await this.setMediaState({
audio_enabled: this.microphone,
video_enabled: this.video,
});
this.setSound(this.sound);
this.emitStateUpdate();
}
public async forceState(desired: CallControlState) {
this.state = new CallControlState(
desired.microphone,
desired.video,
desired.sound,
this.screenshare,
this.spotlight,
);
await this.applyState();
}
public startObserving() {
if (!this.document) return;
this.bodyMutationObserver.observe(this.document.body, {
childList: true,
subtree: false, // only direct children of body
});
this.onBodyMutation();
}
private onBodyMutation() {
if (!this.document) return;
this.document.body.style.setProperty('background', 'none', 'important');
const controls = this.leaveButton?.parentElement?.parentElement;
if (controls) {
controls.style.setProperty('position', 'absolute');
controls.style.setProperty('visibility', 'hidden');
}
this.observeControls();
}
private observeControls() {
this.controlMutationObserver.disconnect();
const screenshareBtn = this.screenshareButton;
if (screenshareBtn) {
this.controlMutationObserver.observe(screenshareBtn, {
attributes: true,
attributeFilter: ['data-kind'],
});
}
const spotlightBtn = this.spotlightButton;
if (spotlightBtn) {
this.controlMutationObserver.observe(spotlightBtn, {
attributes: true,
});
}
this.onControlMutation();
}
public applySound() {
this.setSound(this.sound);
}
private async setMediaState(state: ElementMediaStatePayload) {
// transport.send resolves once EC has ACK'd the command, which is enough to
// consider the mute applied. We deliberately do NOT gate completion on a
// follow-up DeviceMute state-echo: EC may elide it (e.g. when the requested
// state already matches its current state) or skip it during teardown,
// which would strand this promise forever and block applyState(). The echo,
// when it does arrive, is still handled authoritatively by onMediaState().
return this.call.transport.send(ElementWidgetActions.DeviceMute, state);
}
private setSound(sound: boolean): void {
const callDocument = this.iframe.contentDocument ?? this.iframe.contentWindow?.document;
if (callDocument) {
callDocument.querySelectorAll('audio').forEach((el) => {
const isScreenshareAudio = el.getAttribute('data-lk-source') === 'screen_share_audio';
el.muted = !sound || (isScreenshareAudio && this.screenshareAudioMuted);
});
}
}
private applyScreenshareAudioMuted(): void {
if (!this.sound) return;
const callDocument = this.iframe.contentDocument ?? this.iframe.contentWindow?.document;
if (callDocument) {
callDocument
.querySelectorAll<HTMLAudioElement>('audio[data-lk-source="screen_share_audio"]')
.forEach((el) => {
el.muted = this.screenshareAudioMuted;
});
}
}
public onMediaState(evt: CustomEvent<ElementMediaStateDetail>) {
const { data } = evt.detail;
if (!data) return;
const state = new CallControlState(
data.audio_enabled ?? this.microphone,
data.video_enabled ?? this.video,
this.sound,
this.screenshare,
this.spotlight,
this.screenshareAudioMuted,
);
this.state = state;
this.emitStateUpdate();
if (this.microphone && !this.sound) {
this.toggleSound();
}
}
private onControlMutation() {
const screenshare: boolean = this.screenshareButton?.getAttribute('data-kind') === 'primary';
const spotlight: boolean = this.spotlightButton?.checked ?? false;
this.state = new CallControlState(
this.microphone,
this.video,
this.sound,
screenshare,
spotlight,
this.screenshareAudioMuted,
);
this.emitStateUpdate();
}
public setMicrophone(enabled: boolean) {
const payload: ElementMediaStatePayload = {
audio_enabled: enabled,
video_enabled: this.video,
};
return this.setMediaState(payload);
}
public toggleMicrophone() {
const payload: ElementMediaStatePayload = {
audio_enabled: !this.microphone,
video_enabled: this.video,
};
return this.setMediaState(payload);
}
public toggleVideo() {
const payload: ElementMediaStatePayload = {
audio_enabled: this.microphone,
video_enabled: !this.video,
};
return this.setMediaState(payload);
}
public toggleSound() {
const sound = !this.sound;
this.setSound(sound);
// After un-deafening, re-apply screenshare audio mute if active
if (sound) this.applyScreenshareAudioMuted();
const state = new CallControlState(
this.microphone,
this.video,
sound,
this.screenshare,
this.spotlight,
this.screenshareAudioMuted,
);
this.state = state;
this.emitStateUpdate();
if (!this.sound && this.microphone) {
this.toggleMicrophone();
}
}
public toggleScreenshareAudio() {
const screenshareAudioMuted = !this.screenshareAudioMuted;
this.state = new CallControlState(
this.microphone,
this.video,
this.sound,
this.screenshare,
this.spotlight,
screenshareAudioMuted,
);
this.emitStateUpdate();
this.applyScreenshareAudioMuted();
}
public toggleScreenshare() {
this.screenshareButton?.click();
}
public toggleSpotlight() {
if (this.spotlight) {
this.gridButton?.click();
return;
}
this.spotlightButton?.click();
}
public setPipMode(pip: boolean) {
this._pipMode = pip;
}
public toggleReactions() {
this.reactionsButton?.click();
}
public toggleSettings() {
this.settingsButton?.click();
}
/**
* Focus a specific participant's camera tile in Element Call.
*
* EC renders video tiles as `[data-testid="videoTile"]`. Each tile wraps a
* mute-status indicator with `aria-label` set to the participant's Matrix
* user ID. We find the tile containing that user, switch to spotlight mode
* if needed, then click the tile so EC's internal focus handler runs.
*
* Falls back to a plain spotlight toggle if the tile is not found (e.g. the
* participant has their camera off and EC didn't render a video tile for
* them yet).
*/
public focusCameraParticipant(userId: string): void {
const doc = this.document;
if (!doc) return;
// EC labels participant tiles inconsistently across versions — the user's
// matrix id may be the full aria-label, a substring of it, or carried on a
// data attribute (and sometimes the visible label is the display name, not
// the id at all). Try several strategies before giving up, then walk up to
// the enclosing video tile.
const findTile = (): HTMLElement | undefined => {
const escaped = CSS.escape(userId);
const el =
doc.querySelector<HTMLElement>(`[aria-label="${escaped}"]`) ??
doc.querySelector<HTMLElement>(`[data-testid="videoTile"][aria-label*="${escaped}"]`) ??
doc.querySelector<HTMLElement>(`[aria-label*="${escaped}"]`) ??
doc.querySelector<HTMLElement>(`[data-member-id="${escaped}"]`) ??
doc.querySelector<HTMLElement>(`[data-id="${escaped}"]`) ??
undefined;
return (
el?.closest<HTMLElement>('[data-testid="videoTile"]') ??
el?.closest<HTMLElement>('[data-video-fit]') ??
el ??
undefined
);
};
const applyFocus = () => {
const tile = findTile();
if (tile) {
tile.click();
} else if (import.meta.env.DEV) {
console.warn(`[CallControl] focusCameraParticipant: no tile matched ${userId}`);
}
};
if (this.spotlight) {
// Already in spotlight — pin immediately.
applyFocus();
return;
}
// Switching to spotlight re-renders EC's layout asynchronously; clicking the
// tile in the same tick would land in the old (grid) DOM. A fixed frame
// delay is unreliable (EC's React commit can exceed it on slow devices), so
// watch the iframe DOM for a spotlight video tile to mount, then focus —
// with a hard timeout so the click is always attempted at least once.
this.spotlightButton?.click();
const tileSelector = '[data-testid="videoTile"]';
let settled = false;
let observer: MutationObserver | undefined;
let timer: ReturnType<typeof setTimeout> | undefined;
const finish = () => {
if (settled) return;
settled = true;
if (timer) clearTimeout(timer);
observer?.disconnect();
applyFocus();
};
observer = new MutationObserver(() => {
if (doc.querySelector(tileSelector)) finish();
});
observer.observe(doc.body, { childList: true, subtree: true });
timer = setTimeout(finish, 600);
// A tile may already be present immediately after toggling spotlight.
if (doc.querySelector(tileSelector)) finish();
}
public dispose() {
this.bodyMutationObserver.disconnect();
this.controlMutationObserver.disconnect();
}
private emitStateUpdate() {
this.emit(CallControlEvent.StateUpdate);
}
}