olcxjas-softworks/LarpixClient
2
0
Fork 1
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity Actions 3

Compare commits

base: olcxjas-softworks:ffd952b1f0582050be155ec3fc6f6c56985b69d2
Branches Tags
olcxjas-softworks:main
..
compare: olcxjas-softworks:e3b1a6fc8bd87ae700fc53b4ca2c941fe305895f
Branches Tags
olcxjas-softworks:main

No commits in common. "ffd952b1f0582050be155ec3fc6f6c56985b69d2" and "e3b1a6fc8bd87ae700fc53b4ca2c941fe305895f" have entirely different histories.
ffd952b1f0 ... e3b1a6fc8b

8 changed files with 166 additions and 704 deletions
Download patch file Download diff file Expand all files Collapse all files

72
android/app/src/main/assets/public/main.js
View file

@ -4206,55 +4206,6 @@ function handleChatInputKey(event) {
let dmMessagePollInterval = null; let dmMessagePollInterval = null;
let appWebSocket = null; let appWebSocket = null;
let dmMessageLoadTimeout = null; let dmMessageLoadTimeout = null;
let appWsPingTimer = null;
let callStatesResyncTimer = null;
//Keepalive app-WS: tekstowy "ping" co 20 s. Serwer zamyka połączenie po 75 s
//bezczynności (AppWsIdleTimeout) - bez tego half-open TCP po zabiciu aplikacji /
//zmianie sieci wisiał na serwerze godzinami, cleanup ActiveCalls nie biegł i
//wskaźniki rozmów pokazywały uczestników, którzy dawno wyszli. Ping utrzymuje
//też mapping NAT przy danych komórkowych, więc eventy (call_update!) nie giną.
function startAppWsPing() {
if (appWsPingTimer) clearInterval(appWsPingTimer);
appWsPingTimer = setInterval(() => {
if (appWebSocket && appWebSocket.readyState === WebSocket.OPEN) {
try { appWebSocket.send("ping"); } catch (e) {}
}
}, 20000);
}
function stopAppWsPing() {
if (appWsPingTimer) { clearInterval(appWsPingTimer); appWsPingTimer = null; }
}
//Pełna resynchronizacja wskaźników rozmów: zgubiony podczas przerwy w app-WS
//call_update NIGDY nie jest odtwarzany przez serwer, a seedCallStateFromList
//tylko dodaje wpisy - bez tego wskaźniki "znikały dopiero po ręcznym odświeżeniu".
//Dla każdego pokoju znanego jako aktywny (+ otwartego) pyta room/voice/state
//i koryguje mapę oraz UI. Uruchamiane: po (re)otwarciu WS, co 30 s, przy powrocie
//karty/aplikacji na wierzch.
async function resyncAllCallStates() {
if (!window.id || typeof fetchEncrypted !== "function" || typeof ensureActiveCallsMap !== "function") return;
let map = ensureActiveCallsMap();
let keys = Object.keys(map);
let currentKey = window.currentDmId ? sanitizeId(window.currentDmId) : null;
if (currentKey && !keys.includes(currentKey)) keys.push(currentKey);
if (keys.length === 0) return;
for (let key of keys) {
let roomType = window.currentDmId === key && window.currentRoomType ? window.currentRoomType
: (String(key).includes("_") ? "dms" : "groups");
try {
let res = await fetchEncrypted("room/voice/state", JSON.stringify({ string1: roomType, string2: key }));
if (!res || !res.startsWith("[")) continue;
let participants = JSON.parse(res).map(u => String(u).split(";")[0]);
if (participants.length > 0) map[key] = participants;
else delete map[key];
} catch (e) {}
}
//Po serii poprawek pełne przerysowanie wskaźników (listy + top-bary + details)
if (typeof updateVoicePresenceUI === "function") updateVoicePresenceUI();
}
function setupWebSocket() { function setupWebSocket() {
if (appWebSocket && (appWebSocket.readyState === WebSocket.OPEN || appWebSocket.readyState === WebSocket.CONNECTING)) return; if (appWebSocket && (appWebSocket.readyState === WebSocket.OPEN || appWebSocket.readyState === WebSocket.CONNECTING)) return;
@ -4277,17 +4228,6 @@ function setupWebSocket() {
} finally { } finally {
release(); release();
} }
startAppWsPing();
//Po każdym (re)połączeniu odtwarzamy stan wskaźników - eventy call_update
//z czasu rozłączenia są bezpowrotnie stracone
resyncAllCallStates();
if (!callStatesResyncTimer) {
callStatesResyncTimer = setInterval(() => {
if (document.visibilityState !== "visible") return;
resyncAllCallStates();
}, 30000);
}
}; };
appWebSocket.onmessage = (event) => { appWebSocket.onmessage = (event) => {
@ -4333,22 +4273,10 @@ function setupWebSocket() {
}; };
appWebSocket.onclose = () => { appWebSocket.onclose = () => {
stopAppWsPing();
setTimeout(setupWebSocket, 5000); setTimeout(setupWebSocket, 5000);
}; };
} }
//Powrót karty/aplikacji na wierzch: mobilne WebView potrafią trzymać martwy
//socket odkryty bardzo długo (brak ramek = brak detekcji). Wymuszamy próbę
//połączenia i od razu resynchronizujemy wskaźniki rozmów.
document.addEventListener("visibilitychange", () => {
if (document.visibilityState !== "visible") return;
if (appWebSocket && appWebSocket.readyState !== WebSocket.OPEN && appWebSocket.readyState !== WebSocket.CONNECTING) {
setupWebSocket();
}
resyncAllCallStates();
});
let currentContextMenuMsgId = null; let currentContextMenuMsgId = null;
let touchContextMenuFired = false; let touchContextMenuFired = false;

2
android/app/src/main/assets/public/screens.js
View file

@ -141,7 +141,7 @@ var invitesEntry = `
var chatScreen = ` var chatScreen = `
<div style="display: flex; flex-direction: column; height: 100%; width: 100%;"> <div style="display: flex; flex-direction: column; height: 100%; width: 100%;">
<div id="chat-messages" style="flex-grow: 1; overflow-y: auto; overflow-x: hidden; padding: 1rem; display: flex; flex-direction: column; gap: 0.5rem;"> <div id="chat-messages" style="flex-grow: 1; overflow-y: auto; padding: 1rem; display: flex; flex-direction: column; gap: 0.5rem;">
</div> </div>
<div id="replying-bar" style="display: none; padding: 0 1rem; border-top: var(--border-width) solid var(--light-border-color); font-size: 0.85rem; justify-content: space-between; align-items: center; background: var(--main-bg-color);"> <div id="replying-bar" style="display: none; padding: 0 1rem; border-top: var(--border-width) solid var(--light-border-color); font-size: 0.85rem; justify-content: space-between; align-items: center; background: var(--main-bg-color);">
<div style="overflow: hidden; text-overflow: ellipsis; white-space: nowrap; flex-grow: 1;"> <div style="overflow: hidden; text-overflow: ellipsis; white-space: nowrap; flex-grow: 1;">

11
android/app/src/main/assets/public/style.css
View file

@ -686,7 +686,6 @@ space{
flex-shrink: 0; flex-shrink: 0;
cursor: pointer; cursor: pointer;
margin: 0; margin: 0;
position: relative;
} }
.room-pfp { .room-pfp {
width: 2.2rem; width: 2.2rem;
@ -739,7 +738,6 @@ space{
display: flex; display: flex;
flex-direction: column; flex-direction: column;
flex-grow: 1; flex-grow: 1;
min-width: 0;
} }
.chat-message-header { .chat-message-header {
display: flex; display: flex;
@ -1180,15 +1178,7 @@ input:checked + .toggle-slider:before {
margin-top: 0.2rem; margin-top: 0.2rem;
} }
/* Czerwona kropka nieprzeczytanych - pozycjonowana absolutnie. Odstęp liczony tak,
żeby ŚRODEK kropki leżał na tej samej osi co środek 1rem-owego .call-indicator
(0.6rem paddingu + (1rem - 0.65rem)/2) - inaczej kropka "ucieka" w prawo
względem większej ikony słuchawek. */
.unread-badge { .unread-badge {
position: absolute;
right: calc(0.6rem + (1rem - 0.65rem) / 2);
top: 50%;
margin-top: -0.325rem;
width: 0.65rem; width: 0.65rem;
height: 0.65rem; height: 0.65rem;
min-width: 0.65rem; min-width: 0.65rem;
@ -1196,6 +1186,7 @@ input:checked + .toggle-slider:before {
background-color: var(--big-red); background-color: var(--big-red);
border-radius: 50%; border-radius: 50%;
flex-shrink: 0; flex-shrink: 0;
margin-right: 0.5rem;
transition: transform 0.2s cubic-bezier(0.175, 0.885, 0.32, 1.275); transition: transform 0.2s cubic-bezier(0.175, 0.885, 0.32, 1.275);
transform: scale(0); transform: scale(0);
} }

334
android/app/src/main/assets/public/voice.js
View file

@ -1,6 +1,7 @@
let currentVoiceWs = null; let currentVoiceWs = null;
let currentVoiceCallId = null; let currentVoiceCallId = null;
let currentVoiceCallType = null; let currentVoiceCallType = null;
let voiceWsHeartbeat = null;
let voiceActiveUsers = []; let voiceActiveUsers = [];
let voiceAesKeyBytes = null; // grupowy klucz rozmowy (root) - dystrybuowany hybrydą PQ let voiceAesKeyBytes = null; // grupowy klucz rozmowy (root) - dystrybuowany hybrydą PQ
let voiceUiContainer = null; let voiceUiContainer = null;
@ -106,8 +107,8 @@ function getVoicePeerSenderKey(clientIdStr) {
let micStream = null, audioContext = null, isAudioWorkletLoaded = false; let micStream = null, audioContext = null, isAudioWorkletLoaded = false;
let micSourceNode = null, senderNode = null, receiverNode = null; let micSourceNode = null, senderNode = null, receiverNode = null, audioEncoder = null;
let remoteAudio = {}, audioDecryptChains = {}; let remoteAudio = {}, audioDecoders = {}, audioDecryptChains = {};
let knownVoiceUsers = new Set(); let knownVoiceUsers = new Set();
let isMicMuted = false; let isMicMuted = false;
let remoteVideos = {}, videoDecryptChains = {}; let remoteVideos = {}, videoDecryptChains = {};
@ -125,26 +126,22 @@ let isVoiceSpeakerMode = false;
const SPEAKING_HANGOVER_MS = 400; const SPEAKING_HANGOVER_MS = 400;
//Zawory kolejki socketu: TCP dowiezie zakolejkowane bajty tylko spóźnione, // ===== Voice connection resilience =====
//więc progi są hierarchiczne - wideo ustępuje przy 12 KB, audio dopiero przy // Every client broadcasts a tiny encrypted ping so peers can tell "quiet
//14 KB (~270 ms głosu). Przy korku najpierw znika obraz, głos leci dalej. // user" apart from "crashed/half-open socket". Peers silent for too long are
const VOICE_WS_VIDEO_QUEUE_LIMIT_BYTES = 12 * 1024; // dropped locally; a dead server socket triggers an automatic rejoin.
const VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES = 14 * 1024;
//Resilience: peer bez pakietów >7 s = lokalnie martwy (media wycięte), socket
//bez ruchu od ŻYWYCH peerów >15 s = wymuszony reconnect. Pings co 2 s.
const VOICE_PEER_TIMEOUT_MS = 7000; const VOICE_PEER_TIMEOUT_MS = 7000;
const VOICE_PING_INTERVAL_MS = 2000; const VOICE_PING_INTERVAL_MS = 2000;
const VOICE_SERVER_TIMEOUT_MS = 15000; const VOICE_SERVER_TIMEOUT_MS = 10000;
const VOICE_MAX_RECONNECTS = 4; const VOICE_MAX_RECONNECTS = 2;
const VOICE_ROSTER_POLL_MS = 5000;
let peerLastSeen = {}; let peerLastSeen = {};
let deadPeers = new Set(); let deadPeers = new Set();
let lastVoiceServerPacketAt = 0; let lastVoiceServerPacketAt = 0;
//Czy w bieżącym okresie socketu byli jacyś inni uczestnicy - baza dla
//resetu baseline'u ciszy w watchdogu (patrz komentarz przy użyciu)
let voiceHadRemotePeers = false; let voiceHadRemotePeers = false;
let voicePingTimer = null; let voicePingTimer = null;
let voiceRosterRetryTimer = null; let voiceRosterRetryTimer = null;
let voiceRosterPollTimer = null;
let lastVoiceKeyRequestAt = 0; let lastVoiceKeyRequestAt = 0;
let voiceReconnectAttempts = 0; let voiceReconnectAttempts = 0;
let voiceSocketOpenedAt = 0; let voiceSocketOpenedAt = 0;
@ -170,21 +167,11 @@ function cleanupPeerMedia(userId) {
function sendVoicePing() { function sendVoicePing() {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
const iv = nextVoiceIv(); const iv = nextVoiceIv();
//Typ 5: prośba o echo. Payload = 8B timestamp (ms, WŁASNY zegar). Odbiorca crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, new Uint8Array([Date.now() & 255])).then(cipher => {
//odsyła te same bajty jako typ 6, a RTT liczę dopiero na jego powrocie.
const tsData = BigInt(Date.now());
const tsBuf = new ArrayBuffer(8);
new DataView(tsBuf).setBigUint64(0, tsData);
crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, new Uint8Array(tsBuf)).then(cipher => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
payload[0] = 5; payload[1] = 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14); payload[0] = 5; payload[1] = 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14);
currentVoiceWs.send(payload.buffer); currentVoiceWs.send(payload.buffer);
voiceOutstandingPings.set(tsData, Date.now());
const nowMs = Date.now();
for (const [k, sentAt] of voiceOutstandingPings) {
if (nowMs - sentAt > 15000) voiceOutstandingPings.delete(k);
}
}).catch(() => {}); }).catch(() => {});
} }
@ -230,15 +217,9 @@ const VOICE_FPS_OPTIONS = [15, 24, 30, 60];
let voiceSettings = loadVoiceSettings(); let voiceSettings = loadVoiceSettings();
function loadVoiceSettings() { function loadVoiceSettings() {
//Globalny default 480p@24; stare zapisane defaulty (720p/1080p/30fps) const defaults = { micId: "", camId: "", speakerId: "", camRes: "1280x720", camFps: 24, screenRes: "1920x1080", screenFps: 30, sensitivity: VOICE_SENSITIVITY_DEFAULT };
//migrują na nowe, świadome wybory użytkownika zostają.
const defaults = { micId: "", camId: "", speakerId: "", camRes: "640x480", camFps: 24, screenRes: "640x480", screenFps: 24, sensitivity: VOICE_SENSITIVITY_DEFAULT };
try { try {
let s = Object.assign({}, JSON.parse(localStorage.getItem(VOICE_SETTINGS_KEY)) || {}); return Object.assign(defaults, JSON.parse(localStorage.getItem(VOICE_SETTINGS_KEY)) || {});
if (s.camRes === "1280x720") s.camRes = defaults.camRes;
if (s.screenRes === "1920x1080") s.screenRes = defaults.screenRes;
if (s.screenFps === 30 || s.screenFps === "30") s.screenFps = defaults.screenFps;
return Object.assign(defaults, s);
} catch (e) { } catch (e) {
return defaults; return defaults;
} }
@ -293,35 +274,6 @@ async function applyVoiceSpeakerSink() {
let voiceConnectedAt = null; let voiceConnectedAt = null;
let voiceTimerInterval = null; let voiceTimerInterval = null;
let voicePingMs = null; let voicePingMs = null;
//Kontrola congestionu: bufferedAmount nie widzi backlogu kernela TCP, prawdziwy
//obraz daje RTT z pingów. Baseline = MINIMUM z okna próbek (odporne na skoki
//i na peera sztucznie opóźniającego własne echa - stałe kłamstwo podnosi minimum).
let voiceRttFastMs = 0;
let voiceRttSamples = [];
let voiceRttBaseMs = 0;
let voiceAudioHoldUntil = 0; //histereza zaworu audio
//Echo-ping (typ 5 = prośba, typ 6 = echo): RTT wyłącznie na WŁASNYM zegarze -
//wysyłam timestamp, peer odbija te same bajty, odejmuję czas nadania. Zegar
//peera nigdy nie jest używany (może być przekręcony lub spreparowany).
const voiceOutstandingPings = new Map(); // BigInt(ts) -> czas nadania
let voiceLastEchoAt = 0; // rate-limit odpowiedzi echem
let voiceLastRttSampleAt = 0; // kiedy ostatnio spłynął prawdziwy pomiar
function voiceCongestionDelay() {
if (!voiceRttFastMs || !voiceRttBaseMs) return 0;
return Math.min(5000, Math.max(0, voiceRttFastMs - voiceRttBaseMs));
}
function voiceRecordRttSample(rtt) {
voiceLastRttSampleAt = Date.now();
voicePingMs = voicePingMs === null ? rtt : Math.round(voicePingMs * 0.7 + rtt * 0.3);
voiceRttFastMs = voiceRttFastMs ? Math.round(voiceRttFastMs * 0.5 + rtt * 0.5) : rtt;
voiceRttSamples.push(rtt);
if (voiceRttSamples.length > 12) voiceRttSamples.shift();
voiceRttBaseMs = Math.min.apply(null, voiceRttSamples);
updateVoicePing();
}
function formatVoiceDuration(ms) { function formatVoiceDuration(ms) {
let totalSec = Math.floor(ms / 1000); let totalSec = Math.floor(ms / 1000);
@ -678,11 +630,7 @@ setInterval(() => {
} }
// Half-open socket watchdog + backgrounded-tab audio recovery. // Half-open socket watchdog + backgrounded-tab audio recovery.
if (currentVoiceWs && currentVoiceWs.readyState === WebSocket.OPEN) { if (currentVoiceWs && currentVoiceWs.readyState === WebSocket.OPEN) {
//deadPeers NIE liczą się do "others": ghost na rosterze (martwy TCP const others = voiceActiveUsers.filter(u => u !== String(window.id)).length;
//jeszcze nie zsynchronizowany przez serwer) nigdy nic nie wyśle, więc
//zliczany jak żywy wymuszał reconnecty w pętli - migające "Connecting"
//w zdrowej rozmowie.
const others = voiceActiveUsers.filter(u => u !== String(window.id) && !deadPeers.has(u)).length;
// Baseline ciszy odświeżany przy przejściu 0->N uczestników: solo w // Baseline ciszy odświeżany przy przejściu 0->N uczestników: solo w
// rozmowie NIE MA żadnego ruchu przychodzącego, więc po pojawieniu się // rozmowie NIE MA żadnego ruchu przychodzącego, więc po pojawieniu się
// nowej osoby lastVoiceServerPacketAt bywa nieświeży z okresu solo. // nowej osoby lastVoiceServerPacketAt bywa nieświeży z okresu solo.
@ -745,8 +693,8 @@ class AudioReceiver extends AudioWorkletProcessor {
// PREBUF_START - initial prebuffer before first playback (~43 ms) // PREBUF_START - initial prebuffer before first playback (~43 ms)
// PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms) // PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms)
// HARD_LIMIT - burst ceiling (~128 ms): snap back to target immediately // HARD_LIMIT - burst ceiling (~128 ms): snap back to target immediately
// Backlog above target is drained adaptively (proportional to excess, see // Backlog slightly above target is drained by a gentle trim (1 sample per
// process()) so latency returns to ~43 ms quickly after network bursts. // quantum) instead of letting latency creep up to the hard limit.
constructor(options) { constructor(options) {
super(); super();
this.users = {}; this.users = {};
@ -784,12 +732,7 @@ class AudioReceiver extends AudioWorkletProcessor {
if (available < output.length) { user.buffering = true; } if (available < output.length) { user.buffering = true; }
else { else {
if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536; if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536;
else if (available > 2049) { else if (available > 2049) user.readPtr = (user.readPtr + 1) % 65536;
//Adaptacyjny drain: nadmiar nad target odprowadzany proporcjonalnie
//(1 + excess/256 /kwant) - powrót z ~128 ms do ~43 ms w ułamku sekundy
const excess = available - 2048;
user.readPtr = (user.readPtr + 1 + (excess >> 8)) % 65536;
}
let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0; let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0;
for (let i = 0; i < output.length; i++) { for (let i = 0; i < output.length; i++) {
output[i] += (user.buffer[user.readPtr] / 32768) * vol; output[i] += (user.buffer[user.readPtr] / 32768) * vol;
@ -947,38 +890,6 @@ function bytesToB64u(a) {
return btoa(bin).replace(/\+/g, "-").replace(/\//g, "_").replace(/=+$/, ""); return btoa(bin).replace(/\+/g, "-").replace(/\//g, "_").replace(/=+$/, "");
} }
//Roster pokoju z voice-servera (źródło prawdy) -> synchronizacja kafelków.
//Steady-poll co VOICE_ROSTER_POLL_MS - bez tego sweep ghosta/leave bez
//call_update nigdy nie docierał do UI.
async function fetchVoiceRoster() {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !currentVoiceCallId) return;
try {
const baseUrl = currentVoiceWs.url.split('/ws')[0];
const httpUrl = baseUrl.replace('ws:', 'http:').replace('wss:', 'https:');
const t0 = performance.now();
const res = await fetch(`${httpUrl}/room/users?room=${encodeURIComponent(currentVoiceCallId)}`);
const usersRaw = await res.json();
//HTTP tylko jako szacunek gdy brak prawdziwych próbek RTT (solo / stary peer)
if (voicePingMs === null || Date.now() - voiceLastRttSampleAt > 25000) {
voicePingMs = Math.round(performance.now() - t0);
updateVoicePing();
}
const rawUsers = usersRaw.map(u => String(u));
//Serwer jest źródłem prawdy o składzie - nie filtrujemy po
//deadPeers (audyt S15: ukryty peer nadal dostaje audio)
for (const gone of [...deadPeers]) {
if (!rawUsers.includes(gone)) deadPeers.delete(gone);
}
const users = rawUsers;
let changed = (users.length !== voiceActiveUsers.length);
for(let u of users) if(!voiceActiveUsers.includes(u)) changed = true;
if(changed) {
voiceActiveUsers = users;
updateVoiceUIUsers();
}
} catch (e) {}
}
async function startVoiceSession(serverUrl, ticket) { async function startVoiceSession(serverUrl, ticket) {
if (!ticket) throw new Error("missing voice ticket"); if (!ticket) throw new Error("missing voice ticket");
// Single-session invariant: a stale socket (e.g. a double-fired join) // Single-session invariant: a stale socket (e.g. a double-fired join)
@ -1021,7 +932,32 @@ async function startVoiceSession(serverUrl, ticket) {
else btn.classList.remove("voice-btn-danger"); else btn.classList.remove("voice-btn-danger");
} }
fetchVoiceRoster(); const fetchUsers = async () => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !currentVoiceCallId) return;
try {
const baseUrl = currentVoiceWs.url.split('/ws')[0];
const httpUrl = baseUrl.replace('ws:', 'http:').replace('wss:', 'https:');
const t0 = performance.now();
const res = await fetch(`${httpUrl}/room/users?room=${currentVoiceCallId}`);
const usersRaw = await res.json();
voicePingMs = Math.round(performance.now() - t0);
updateVoicePing();
const rawUsers = usersRaw.map(u => String(u));
//Serwer jest źródłem prawdy o składzie - nie filtrujemy po
//deadPeers (audyt S15: ukryty peer nadal dostaje audio)
for (const gone of [...deadPeers]) {
if (!rawUsers.includes(gone)) deadPeers.delete(gone);
}
const users = rawUsers;
let changed = (users.length !== voiceActiveUsers.length);
for(let u of users) if(!voiceActiveUsers.includes(u)) changed = true;
if(changed) {
voiceActiveUsers = users;
updateVoiceUIUsers();
}
} catch (e) {}
};
fetchUsers();
//The voice-server roster is joined only AFTER ticket validation (an //The voice-server roster is joined only AFTER ticket validation (an
//HTTP roundtrip to the main server), so the immediate fetch can still //HTTP roundtrip to the main server), so the immediate fetch can still
//miss us - retry quickly until our own tile shows up instead of //miss us - retry quickly until our own tile shows up instead of
@ -1035,17 +971,13 @@ async function startVoiceSession(serverUrl, ticket) {
voiceRosterRetryTimer = null; voiceRosterRetryTimer = null;
return; return;
} }
fetchVoiceRoster(); fetchUsers();
}, 400); }, 400);
voiceReconnectAttempts = 0; voiceReconnectAttempts = 0;
lastVoiceServerPacketAt = Date.now(); lastVoiceServerPacketAt = Date.now();
if (voicePingTimer) clearInterval(voicePingTimer); if (voicePingTimer) clearInterval(voicePingTimer);
voicePingTimer = setInterval(sendVoicePing, VOICE_PING_INTERVAL_MS); voicePingTimer = setInterval(sendVoicePing, VOICE_PING_INTERVAL_MS);
//Steady roster poll - utrzymuje kafelki zgodne z prawdą serwera
//(ghost-sweep, leave bez call_update) przez całą rozmowę.
if (voiceRosterPollTimer) clearInterval(voiceRosterPollTimer);
voiceRosterPollTimer = setInterval(fetchVoiceRoster, VOICE_ROSTER_POLL_MS);
//S16: getUserMedia/worklet/rezump AudioContext potrafią rzucić //S16: getUserMedia/worklet/rezump AudioContext potrafią rzucić
//(brak uprawnień, zniknięte urządzenie) - bez obsługi zostawał //(brak uprawnień, zniknięte urządzenie) - bez obsługi zostawał
@ -1101,18 +1033,12 @@ async function startVoiceSession(serverUrl, ticket) {
} }
let vadHangover = 0; let vadHangover = 0;
const myIdStr = String(window.id);
senderNode.port.onmessage = async (e) => { senderNode.port.onmessage = async (e) => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
//Zawór bajtowy + zawór RTT (kolejka kernela jest niewidoczna dla // Backpressure valve: when the socket is congested, drop mic
//bufferedAmount). Drop >450 ms nadwyżki z histerezą 700 ms - lepiej // frames instead of queueing them - queued audio would arrive
//urwać ułamek sekundy niż mówić z sekundami opóźnienia. // minutes late and drown the connection for every stream.
if (currentVoiceWs.bufferedAmount > VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES) { if (currentVoiceWs.bufferedAmount > 256 * 1024) {
vadHangover = 0;
return;
}
if (voiceCongestionDelay() > 450) voiceAudioHoldUntil = Date.now() + 700;
if (Date.now() < voiceAudioHoldUntil) {
vadHangover = 0; vadHangover = 0;
return; return;
} }
@ -1127,7 +1053,7 @@ async function startVoiceSession(serverUrl, ticket) {
if (db >= sensitivity) vadHangover = 75; if (db >= sensitivity) vadHangover = 75;
let isSpeaking = vadHangover > 0 && !isMicMuted; let isSpeaking = vadHangover > 0 && !isMicMuted;
setTileSpeaking(myIdStr, isSpeaking); setTileSpeaking(String(window.id), isSpeaking);
if (vadHangover > 0 && !isMicMuted) { if (vadHangover > 0 && !isMicMuted) {
vadHangover--; vadHangover--;
@ -1158,44 +1084,7 @@ async function startVoiceSession(serverUrl, ticket) {
if (deadPeers.has(clientId)) deadPeers.delete(clientId); if (deadPeers.has(clientId)) deadPeers.delete(clientId);
const type = view.getUint8(8); const type = view.getUint8(8);
if (type === 5 || type === 6) { if (type === 5) return; // liveness ping - presence already recorded
//Liveness już zapisany wyżej. Dalej tylko przy zainstalowanym kluczu.
if (voiceAesKeyBytes) {
const senderKey = getVoicePeerSenderKey(clientId);
if (senderKey) {
//Offsety ODBIORCZE: serwer dokleja 8B nadawcy, więc iv=[10..22),
//ct=[22..). Format wysyłki (iv=2, ct=14) jest INNY - pomylenie
//psuje każdą deszyfrację cicho (GCM auth fail -> .catch)
const piv = new Uint8Array(arrayBuffer, 10, 12);
const pct = new Uint8Array(arrayBuffer, 22);
crypto.subtle.decrypt({name: "AES-GCM", iv: piv}, senderKey, pct).then(plain => {
if (plain.byteLength !== 8) return; //stary klient: ping bez timestampu
const tsData = new DataView(plain).getBigUint64(0);
if (type === 5) {
//Odbij te same bajty własnym podkluczem i świeżym IV; rate-limit
const nowMs = Date.now();
if (nowMs - voiceLastEchoAt < 150) return;
voiceLastEchoAt = nowMs;
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
const riv = nextVoiceIv();
crypto.subtle.encrypt({name: "AES-GCM", iv: riv}, voiceMySenderCryptoKey, plain).then(rcipher => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
const rpayload = new Uint8Array(14 + rcipher.byteLength);
rpayload[0] = 6; rpayload[1] = 0; rpayload.set(riv, 2); rpayload.set(new Uint8Array(rcipher), 14);
currentVoiceWs.send(rpayload.buffer);
}).catch(() => {});
} else {
//Typ 6: powrót własnego timestampu - RTT na WŁASNYM zegarze.
if (!voiceOutstandingPings.has(tsData)) return;
voiceOutstandingPings.delete(tsData);
const rtt = Date.now() - Number(tsData);
if (rtt >= 1 && rtt < 15000) voiceRecordRttSample(rtt);
}
}).catch(() => {});
}
}
return;
}
if (!voiceActiveUsers.includes(clientId) && clientId !== String(window.id)) { if (!voiceActiveUsers.includes(clientId) && clientId !== String(window.id)) {
const baseUrl = currentVoiceWs.url.split('/ws')[0]; const baseUrl = currentVoiceWs.url.split('/ws')[0];
@ -1276,7 +1165,6 @@ async function startVoiceSession(serverUrl, ticket) {
frame.close(); frame.close();
}, },
error: e => { error: e => {
voiceLogError('videodecoder[' + decoderId + ']', e);
if (remoteVideos[decoderId]) { if (remoteVideos[decoderId]) {
try { remoteVideos[decoderId].decoder.close(); } catch(err) {} try { remoteVideos[decoderId].decoder.close(); } catch(err) {}
remoteVideos[decoderId].decoder = null; remoteVideos[decoderId].decoder = null;
@ -1302,14 +1190,11 @@ async function startVoiceSession(serverUrl, ticket) {
try { try {
streamData.decoder.decode(chunk); streamData.decoder.decode(chunk);
} catch(err) { } catch(err) {
voiceLogError('videodecode[' + decoderId + ']', err);
try { streamData.decoder.close(); } catch(e){} try { streamData.decoder.close(); } catch(e){}
streamData.decoder = null; streamData.decoder = null;
streamData.hasKeyFrame = false; streamData.hasKeyFrame = false;
} }
} catch (err) { } catch (err) {}
voiceLogError('vdecrypt[' + decoderId + ']', err);
}
}); });
} else if (type === 3) { } else if (type === 3) {
(async () => { (async () => {
@ -1447,6 +1332,8 @@ async function handleCallUpdate(data) {
for (let u in remoteAudio) { for (let u in remoteAudio) {
let pureU = u.replace("_screen", ""); let pureU = u.replace("_screen", "");
if (!voiceActiveUsers.includes(pureU) && pureU !== myIdStr) { if (!voiceActiveUsers.includes(pureU) && pureU !== myIdStr) {
if (audioDecoders[u]) try { audioDecoders[u].close(); } catch(e){}
delete audioDecoders[u];
delete remoteAudio[u]; delete remoteAudio[u];
const rn = u.endsWith("_screen") ? screenReceiverNode : receiverNode; const rn = u.endsWith("_screen") ? screenReceiverNode : receiverNode;
if(rn) rn.port.postMessage({ action: 'remove', clientId: u }); if(rn) rn.port.postMessage({ action: 'remove', clientId: u });
@ -2603,6 +2490,7 @@ function toggleVoiceMic() {
// keeping the UI, mic stream and AudioContext alive for an automatic rejoin. // keeping the UI, mic stream and AudioContext alive for an automatic rejoin.
function teardownVoiceTransport() { function teardownVoiceTransport() {
if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; } if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; }
if (voiceWsHeartbeat) { clearInterval(voiceWsHeartbeat); voiceWsHeartbeat = null; }
if (camControl) { try { camControl.stop(); } catch (e) {} camControl = null; } if (camControl) { try { camControl.stop(); } catch (e) {} camControl = null; }
if (screenControl) { try { screenControl.stop(); } catch (e) {} screenControl = null; } if (screenControl) { try { screenControl.stop(); } catch (e) {} screenControl = null; }
if (currentVoiceWs) { if (currentVoiceWs) {
@ -2619,7 +2507,6 @@ function teardownVoiceTransport() {
peerLastSeen = {}; peerLastSeen = {};
deadPeers.clear(); deadPeers.clear();
if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; } if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; }
if (voiceRosterPollTimer) { clearInterval(voiceRosterPollTimer); voiceRosterPollTimer = null; }
lastVoiceKeyRequestAt = 0; lastVoiceKeyRequestAt = 0;
//Stale participants from the previous attempt must not linger in the UI //Stale participants from the previous attempt must not linger in the UI
voiceActiveUsers = []; voiceActiveUsers = [];
@ -2629,12 +2516,6 @@ function teardownVoiceTransport() {
setVoiceRootKey(null); setVoiceRootKey(null);
voiceSendEpoch = null; voiceSendEpoch = null;
voiceSendCounter = 0n; voiceSendCounter = 0n;
voiceRttFastMs = 0;
voiceRttSamples = [];
voiceRttBaseMs = 0;
voiceAudioHoldUntil = 0;
voiceOutstandingPings.clear();
voiceLastRttSampleAt = 0;
lastVoiceServerPacketAt = Date.now(); lastVoiceServerPacketAt = Date.now();
} }
@ -2651,10 +2532,6 @@ function scheduleVoiceReconnect() {
voiceReconnectAttempts++; voiceReconnectAttempts++;
setVoiceStatus("connecting"); setVoiceStatus("connecting");
teardownVoiceTransport(); teardownVoiceTransport();
//Zmiana sieci = stary RTT dotyczy martwej ścieżki; bez resetu wskaźnik
//wisiałby zamrożony, a fallback HTTP nie miał prawa go nadpisać
voicePingMs = null;
updateVoicePing();
setTimeout(async () => { setTimeout(async () => {
if (!currentVoiceCallId || !currentVoiceCallType) return; if (!currentVoiceCallId || !currentVoiceCallType) return;
try { try {
@ -2666,7 +2543,7 @@ function scheduleVoiceReconnect() {
console.warn("Voice reconnect failed", err); console.warn("Voice reconnect failed", err);
scheduleVoiceReconnect(); scheduleVoiceReconnect();
} }
}, Math.min(4000, 800 * voiceReconnectAttempts)); }, 800);
} }
function endVoiceSession() { function endVoiceSession() {
@ -2706,27 +2583,24 @@ function endVoiceSession() {
micStream = null; micStream = null;
} }
if (micSourceNode) { micSourceNode.disconnect(); micSourceNode = null; } if (micSourceNode) { micSourceNode.disconnect(); micSourceNode = null; }
if (audioEncoder) { try { audioEncoder.close(); } catch(e){} audioEncoder = null; }
if (voiceWsHeartbeat) { clearInterval(voiceWsHeartbeat); voiceWsHeartbeat = null; }
if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; } if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; }
if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; } if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; }
if (voiceRosterPollTimer) { clearInterval(voiceRosterPollTimer); voiceRosterPollTimer = null; }
peerLastSeen = {}; peerLastSeen = {};
deadPeers.clear(); deadPeers.clear();
lastVoiceKeyRequestAt = 0; lastVoiceKeyRequestAt = 0;
voiceReconnectAttempts = 0;
voiceActiveUsers = []; voiceActiveUsers = [];
voiceHadRemotePeers = false; voiceHadRemotePeers = false;
voiceRttFastMs = 0;
voiceRttSamples = [];
voiceRttBaseMs = 0;
voiceAudioHoldUntil = 0;
voiceOutstandingPings.clear();
voiceLastEchoAt = 0;
voiceLastRttSampleAt = 0;
for (let id in audioDecoders) {
try { audioDecoders[id].close(); } catch(e){}
}
for (let id in audioDecryptChains) { for (let id in audioDecryptChains) {
if(audioDecryptChains[id]) audioDecryptChains[id].clear(); if(audioDecryptChains[id]) audioDecryptChains[id].clear();
} }
audioDecoders = {};
audioDecryptChains = {}; audioDecryptChains = {};
remoteAudio = {}; remoteAudio = {};
@ -2798,11 +2672,8 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
screenSenderNode = new AudioWorkletNode(audioContext, 'audio-sender'); screenSenderNode = new AudioWorkletNode(audioContext, 'audio-sender');
screenSenderNode.port.onmessage = async (e) => { screenSenderNode.port.onmessage = async (e) => {
if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
// Ten sam zawór co mikrofon (limit bajtów + korek RTT) - audio // Same backpressure valve as the mic sender.
// ma pierwszeństwo przed wideo na wspólnym sockecie. if (currentVoiceWs.bufferedAmount > 256 * 1024) return;
if (currentVoiceWs.bufferedAmount > VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES) return;
if (voiceCongestionDelay() > 450) voiceAudioHoldUntil = Date.now() + 700;
if (Date.now() < voiceAudioHoldUntil) return;
const iv = nextVoiceIv(); const iv = nextVoiceIv();
const cipher = await crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, e.data.audioData); const cipher = await crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, e.data.audioData);
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
@ -2832,24 +2703,7 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
return { width, height, fps, bitrate }; return { width, height, fps, bitrate };
}; };
let currentConfig = getCurrentConfig(), isRunning = true, lastKeyframeAt = 0; let currentConfig = getCurrentConfig(), isRunning = true, frameCount = 0;
//AIMD + slow-start: start na 40% celu (pełną prędkością waliłby w nieznany
//uplink, a bufory TCP w kernelu to MB-y = sekundy kolejki zanim cokolwiek
//zmierzymy). Sygnał korka: skipy (zawór 12 KB) lub nadwyżka RTT.
let brScale = 0.4, encTries = 0, encSkips = 0, brWindowAt = performance.now(), lastBrReconfAt = 0;
//Korek trwa mimo podłogi bitrate'u → FPS spada do 1/2, 1/3
let fpsDiv = 1, lowStreak = 0, cleanStreak = 0, tickCount = 0;
let activeBitrate = Math.round(currentConfig.bitrate * brScale);
const applyEncoderBitrate = (force = false) => {
const target = Math.max(120000, Math.round(currentConfig.bitrate * brScale));
if (!force && Math.abs(target - activeBitrate) / activeBitrate < 0.08) return;
if (!force && performance.now() - lastBrReconfAt < 800) return;
activeBitrate = target;
lastBrReconfAt = performance.now();
try {
encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: activeBitrate, latencyMode: 'realtime'});
} catch (e) {}
};
const encoder = new VideoEncoder({ const encoder = new VideoEncoder({
output: async (chunk) => { output: async (chunk) => {
if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
@ -2859,9 +2713,9 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
payload[0] = typeId; payload[1] = chunk.type === 'key' ? 1 : 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14); payload[0] = typeId; payload[1] = chunk.type === 'key' ? 1 : 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14);
currentVoiceWs.send(payload.buffer); currentVoiceWs.send(payload.buffer);
}, error: (e) => { voiceLogError('videoencoder', e); stopFunction(); } }, error: (e) => { stopFunction(); }
}); });
encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: activeBitrate, latencyMode: 'realtime'}); encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: currentConfig.bitrate, latencyMode: 'realtime'});
// Transient capture-mode wobbles must not thrash the encoder - // Transient capture-mode wobbles must not thrash the encoder -
// commit only after ~300 ms stable. Camera size/fps changes arrive // commit only after ~300 ms stable. Camera size/fps changes arrive
@ -2879,62 +2733,20 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
pendingCfg = newConfig; pendingCfg = newConfig;
pendingAt = performance.now(); pendingAt = performance.now();
} else if (performance.now() - pendingAt > 300) { } else if (performance.now() - pendingAt > 300) {
//Zmiana geometrii źródła - uczony scale łącza zostaje, currentConfig = newConfig; forceKeyframe = true; frameCount = 0; pendingCfg = null;
//resetujemy tylko redukcję FPS encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: currentConfig.bitrate, latencyMode: 'realtime'});
currentConfig = newConfig; forceKeyframe = true; lastKeyframeAt = 0; pendingCfg = null;
activeBitrate = Math.round(currentConfig.bitrate * brScale);
fpsDiv = 1; lowStreak = 0; cleanStreak = 0;
applyEncoderBitrate(true);
} }
} else if (pendingCfg) { } else if (pendingCfg) {
pendingCfg = null; // wiggle reverted on its own - ignore pendingCfg = null; // wiggle reverted on its own - ignore
} }
//Zawór kolejki socketu (12 KB) ustępuje DALEJ przed progiem audio if (encoder.state === 'configured' && encoder.encodeQueueSize <= 2 && currentVoiceWs.bufferedAmount < 1024 * 256 && videoEl.readyState >= 2) {
//(14 KB) - przy korku najpierw znika obraz, głos płynie dalej.
tickCount++;
const policySkip = fpsDiv > 1 && (tickCount % fpsDiv !== 0);
let skipped = false;
if (policySkip) {
//celowe tempo (fpsDiv) - nie liczy się do skipRatio
} else if (encoder.state === 'configured' && encoder.encodeQueueSize <= 2 &&
currentVoiceWs.readyState === WebSocket.OPEN &&
currentVoiceWs.bufferedAmount < VOICE_WS_VIDEO_QUEUE_LIMIT_BYTES && videoEl.readyState >= 2) {
let frame = null; let frame = null;
try { try {
const nowMs = performance.now(); frame = new VideoFrame(videoEl, {timestamp: Math.trunc(performance.now() * 1000)});
frame = new VideoFrame(videoEl, {timestamp: Math.trunc(nowMs * 1000)}); frameCount++;
//Keyframe wg zegara co ~1 s (nie licznika klatek - zawór je pomija) encoder.encode(frame, {keyFrame: forceKeyframe || (frameCount % (currentConfig.fps * 2)) === 0});
const wantKeyframe = forceKeyframe || (nowMs - lastKeyframeAt >= 1000);
encoder.encode(frame, {keyFrame: wantKeyframe});
if (wantKeyframe) lastKeyframeAt = nowMs;
} catch (err) { } finally { if (frame) frame.close(); } } catch (err) { } finally { if (frame) frame.close(); }
} else if (encoder.state === 'configured') {
skipped = true;
}
encTries++;
if (skipped) encSkips++;
const winNow = performance.now();
if (winNow - brWindowAt >= 800) {
const skipRatio = encTries > 0 ? encSkips / encTries : 0;
const congestionMs = voiceCongestionDelay();
if (congestionMs > 450 || skipRatio > 0.3 || congestionMs > 250) {
brScale = Math.max(0.15, brScale * (congestionMs > 450 ? 0.6 : 0.75));
applyEncoderBitrate();
if (brScale <= 0.17) lowStreak++; else lowStreak = 0;
if (lowStreak >= 4 && fpsDiv < 3) { fpsDiv++; lowStreak = 0; }
cleanStreak = 0;
} else if (skipRatio === 0 && congestionMs < 80) {
cleanStreak++;
lowStreak = 0;
if (cleanStreak >= 5) {
if (fpsDiv > 1) { fpsDiv--; }
else if (brScale < 1) { brScale = Math.min(1, brScale * 1.2); applyEncoderBitrate(); }
cleanStreak = 0;
}
}
encTries = 0; encSkips = 0; brWindowAt = winNow;
} }
setTimeout(encodeLoop, 1000 / currentConfig.fps); setTimeout(encodeLoop, 1000 / currentConfig.fps);
}; };

72
webroot/main.js
View file

@ -4206,55 +4206,6 @@ function handleChatInputKey(event) {
let dmMessagePollInterval = null; let dmMessagePollInterval = null;
let appWebSocket = null; let appWebSocket = null;
let dmMessageLoadTimeout = null; let dmMessageLoadTimeout = null;
let appWsPingTimer = null;
let callStatesResyncTimer = null;
//Keepalive app-WS: tekstowy "ping" co 20 s. Serwer zamyka połączenie po 75 s
//bezczynności (AppWsIdleTimeout) - bez tego half-open TCP po zabiciu aplikacji /
//zmianie sieci wisiał na serwerze godzinami, cleanup ActiveCalls nie biegł i
//wskaźniki rozmów pokazywały uczestników, którzy dawno wyszli. Ping utrzymuje
//też mapping NAT przy danych komórkowych, więc eventy (call_update!) nie giną.
function startAppWsPing() {
if (appWsPingTimer) clearInterval(appWsPingTimer);
appWsPingTimer = setInterval(() => {
if (appWebSocket && appWebSocket.readyState === WebSocket.OPEN) {
try { appWebSocket.send("ping"); } catch (e) {}
}
}, 20000);
}
function stopAppWsPing() {
if (appWsPingTimer) { clearInterval(appWsPingTimer); appWsPingTimer = null; }
}
//Pełna resynchronizacja wskaźników rozmów: zgubiony podczas przerwy w app-WS
//call_update NIGDY nie jest odtwarzany przez serwer, a seedCallStateFromList
//tylko dodaje wpisy - bez tego wskaźniki "znikały dopiero po ręcznym odświeżeniu".
//Dla każdego pokoju znanego jako aktywny (+ otwartego) pyta room/voice/state
//i koryguje mapę oraz UI. Uruchamiane: po (re)otwarciu WS, co 30 s, przy powrocie
//karty/aplikacji na wierzch.
async function resyncAllCallStates() {
if (!window.id || typeof fetchEncrypted !== "function" || typeof ensureActiveCallsMap !== "function") return;
let map = ensureActiveCallsMap();
let keys = Object.keys(map);
let currentKey = window.currentDmId ? sanitizeId(window.currentDmId) : null;
if (currentKey && !keys.includes(currentKey)) keys.push(currentKey);
if (keys.length === 0) return;
for (let key of keys) {
let roomType = window.currentDmId === key && window.currentRoomType ? window.currentRoomType
: (String(key).includes("_") ? "dms" : "groups");
try {
let res = await fetchEncrypted("room/voice/state", JSON.stringify({ string1: roomType, string2: key }));
if (!res || !res.startsWith("[")) continue;
let participants = JSON.parse(res).map(u => String(u).split(";")[0]);
if (participants.length > 0) map[key] = participants;
else delete map[key];
} catch (e) {}
}
//Po serii poprawek pełne przerysowanie wskaźników (listy + top-bary + details)
if (typeof updateVoicePresenceUI === "function") updateVoicePresenceUI();
}
function setupWebSocket() { function setupWebSocket() {
if (appWebSocket && (appWebSocket.readyState === WebSocket.OPEN || appWebSocket.readyState === WebSocket.CONNECTING)) return; if (appWebSocket && (appWebSocket.readyState === WebSocket.OPEN || appWebSocket.readyState === WebSocket.CONNECTING)) return;
@ -4277,17 +4228,6 @@ function setupWebSocket() {
} finally { } finally {
release(); release();
} }
startAppWsPing();
//Po każdym (re)połączeniu odtwarzamy stan wskaźników - eventy call_update
//z czasu rozłączenia są bezpowrotnie stracone
resyncAllCallStates();
if (!callStatesResyncTimer) {
callStatesResyncTimer = setInterval(() => {
if (document.visibilityState !== "visible") return;
resyncAllCallStates();
}, 30000);
}
}; };
appWebSocket.onmessage = (event) => { appWebSocket.onmessage = (event) => {
@ -4333,22 +4273,10 @@ function setupWebSocket() {
}; };
appWebSocket.onclose = () => { appWebSocket.onclose = () => {
stopAppWsPing();
setTimeout(setupWebSocket, 5000); setTimeout(setupWebSocket, 5000);
}; };
} }
//Powrót karty/aplikacji na wierzch: mobilne WebView potrafią trzymać martwy
//socket odkryty bardzo długo (brak ramek = brak detekcji). Wymuszamy próbę
//połączenia i od razu resynchronizujemy wskaźniki rozmów.
document.addEventListener("visibilitychange", () => {
if (document.visibilityState !== "visible") return;
if (appWebSocket && appWebSocket.readyState !== WebSocket.OPEN && appWebSocket.readyState !== WebSocket.CONNECTING) {
setupWebSocket();
}
resyncAllCallStates();
});
let currentContextMenuMsgId = null; let currentContextMenuMsgId = null;
let touchContextMenuFired = false; let touchContextMenuFired = false;

2
webroot/screens.js
View file

@ -141,7 +141,7 @@ var invitesEntry = `
var chatScreen = ` var chatScreen = `
<div style="display: flex; flex-direction: column; height: 100%; width: 100%;"> <div style="display: flex; flex-direction: column; height: 100%; width: 100%;">
<div id="chat-messages" style="flex-grow: 1; overflow-y: auto; overflow-x: hidden; padding: 1rem; display: flex; flex-direction: column; gap: 0.5rem;"> <div id="chat-messages" style="flex-grow: 1; overflow-y: auto; padding: 1rem; display: flex; flex-direction: column; gap: 0.5rem;">
</div> </div>
<div id="replying-bar" style="display: none; padding: 0 1rem; border-top: var(--border-width) solid var(--light-border-color); font-size: 0.85rem; justify-content: space-between; align-items: center; background: var(--main-bg-color);"> <div id="replying-bar" style="display: none; padding: 0 1rem; border-top: var(--border-width) solid var(--light-border-color); font-size: 0.85rem; justify-content: space-between; align-items: center; background: var(--main-bg-color);">
<div style="overflow: hidden; text-overflow: ellipsis; white-space: nowrap; flex-grow: 1;"> <div style="overflow: hidden; text-overflow: ellipsis; white-space: nowrap; flex-grow: 1;">

11
webroot/style.css
View file

@ -686,7 +686,6 @@ space{
flex-shrink: 0; flex-shrink: 0;
cursor: pointer; cursor: pointer;
margin: 0; margin: 0;
position: relative;
} }
.room-pfp { .room-pfp {
width: 2.2rem; width: 2.2rem;
@ -739,7 +738,6 @@ space{
display: flex; display: flex;
flex-direction: column; flex-direction: column;
flex-grow: 1; flex-grow: 1;
min-width: 0;
} }
.chat-message-header { .chat-message-header {
display: flex; display: flex;
@ -1180,15 +1178,7 @@ input:checked + .toggle-slider:before {
margin-top: 0.2rem; margin-top: 0.2rem;
} }
/* Czerwona kropka nieprzeczytanych - pozycjonowana absolutnie. Odstęp liczony tak,
żeby ŚRODEK kropki leżał na tej samej osi co środek 1rem-owego .call-indicator
(0.6rem paddingu + (1rem - 0.65rem)/2) - inaczej kropka "ucieka" w prawo
względem większej ikony słuchawek. */
.unread-badge { .unread-badge {
position: absolute;
right: calc(0.6rem + (1rem - 0.65rem) / 2);
top: 50%;
margin-top: -0.325rem;
width: 0.65rem; width: 0.65rem;
height: 0.65rem; height: 0.65rem;
min-width: 0.65rem; min-width: 0.65rem;
@ -1196,6 +1186,7 @@ input:checked + .toggle-slider:before {
background-color: var(--big-red); background-color: var(--big-red);
border-radius: 50%; border-radius: 50%;
flex-shrink: 0; flex-shrink: 0;
margin-right: 0.5rem;
transition: transform 0.2s cubic-bezier(0.175, 0.885, 0.32, 1.275); transition: transform 0.2s cubic-bezier(0.175, 0.885, 0.32, 1.275);
transform: scale(0); transform: scale(0);
} }

334
webroot/voice.js
View file

@ -1,6 +1,7 @@
let currentVoiceWs = null; let currentVoiceWs = null;
let currentVoiceCallId = null; let currentVoiceCallId = null;
let currentVoiceCallType = null; let currentVoiceCallType = null;
let voiceWsHeartbeat = null;
let voiceActiveUsers = []; let voiceActiveUsers = [];
let voiceAesKeyBytes = null; // grupowy klucz rozmowy (root) - dystrybuowany hybrydą PQ let voiceAesKeyBytes = null; // grupowy klucz rozmowy (root) - dystrybuowany hybrydą PQ
let voiceUiContainer = null; let voiceUiContainer = null;
@ -106,8 +107,8 @@ function getVoicePeerSenderKey(clientIdStr) {
let micStream = null, audioContext = null, isAudioWorkletLoaded = false; let micStream = null, audioContext = null, isAudioWorkletLoaded = false;
let micSourceNode = null, senderNode = null, receiverNode = null; let micSourceNode = null, senderNode = null, receiverNode = null, audioEncoder = null;
let remoteAudio = {}, audioDecryptChains = {}; let remoteAudio = {}, audioDecoders = {}, audioDecryptChains = {};
let knownVoiceUsers = new Set(); let knownVoiceUsers = new Set();
let isMicMuted = false; let isMicMuted = false;
let remoteVideos = {}, videoDecryptChains = {}; let remoteVideos = {}, videoDecryptChains = {};
@ -125,26 +126,22 @@ let isVoiceSpeakerMode = false;
const SPEAKING_HANGOVER_MS = 400; const SPEAKING_HANGOVER_MS = 400;
//Zawory kolejki socketu: TCP dowiezie zakolejkowane bajty tylko spóźnione, // ===== Voice connection resilience =====
//więc progi są hierarchiczne - wideo ustępuje przy 12 KB, audio dopiero przy // Every client broadcasts a tiny encrypted ping so peers can tell "quiet
//14 KB (~270 ms głosu). Przy korku najpierw znika obraz, głos leci dalej. // user" apart from "crashed/half-open socket". Peers silent for too long are
const VOICE_WS_VIDEO_QUEUE_LIMIT_BYTES = 12 * 1024; // dropped locally; a dead server socket triggers an automatic rejoin.
const VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES = 14 * 1024;
//Resilience: peer bez pakietów >7 s = lokalnie martwy (media wycięte), socket
//bez ruchu od ŻYWYCH peerów >15 s = wymuszony reconnect. Pings co 2 s.
const VOICE_PEER_TIMEOUT_MS = 7000; const VOICE_PEER_TIMEOUT_MS = 7000;
const VOICE_PING_INTERVAL_MS = 2000; const VOICE_PING_INTERVAL_MS = 2000;
const VOICE_SERVER_TIMEOUT_MS = 15000; const VOICE_SERVER_TIMEOUT_MS = 10000;
const VOICE_MAX_RECONNECTS = 4; const VOICE_MAX_RECONNECTS = 2;
const VOICE_ROSTER_POLL_MS = 5000;
let peerLastSeen = {}; let peerLastSeen = {};
let deadPeers = new Set(); let deadPeers = new Set();
let lastVoiceServerPacketAt = 0; let lastVoiceServerPacketAt = 0;
//Czy w bieżącym okresie socketu byli jacyś inni uczestnicy - baza dla
//resetu baseline'u ciszy w watchdogu (patrz komentarz przy użyciu)
let voiceHadRemotePeers = false; let voiceHadRemotePeers = false;
let voicePingTimer = null; let voicePingTimer = null;
let voiceRosterRetryTimer = null; let voiceRosterRetryTimer = null;
let voiceRosterPollTimer = null;
let lastVoiceKeyRequestAt = 0; let lastVoiceKeyRequestAt = 0;
let voiceReconnectAttempts = 0; let voiceReconnectAttempts = 0;
let voiceSocketOpenedAt = 0; let voiceSocketOpenedAt = 0;
@ -170,21 +167,11 @@ function cleanupPeerMedia(userId) {
function sendVoicePing() { function sendVoicePing() {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
const iv = nextVoiceIv(); const iv = nextVoiceIv();
//Typ 5: prośba o echo. Payload = 8B timestamp (ms, WŁASNY zegar). Odbiorca crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, new Uint8Array([Date.now() & 255])).then(cipher => {
//odsyła te same bajty jako typ 6, a RTT liczę dopiero na jego powrocie.
const tsData = BigInt(Date.now());
const tsBuf = new ArrayBuffer(8);
new DataView(tsBuf).setBigUint64(0, tsData);
crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, new Uint8Array(tsBuf)).then(cipher => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
payload[0] = 5; payload[1] = 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14); payload[0] = 5; payload[1] = 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14);
currentVoiceWs.send(payload.buffer); currentVoiceWs.send(payload.buffer);
voiceOutstandingPings.set(tsData, Date.now());
const nowMs = Date.now();
for (const [k, sentAt] of voiceOutstandingPings) {
if (nowMs - sentAt > 15000) voiceOutstandingPings.delete(k);
}
}).catch(() => {}); }).catch(() => {});
} }
@ -230,15 +217,9 @@ const VOICE_FPS_OPTIONS = [15, 24, 30, 60];
let voiceSettings = loadVoiceSettings(); let voiceSettings = loadVoiceSettings();
function loadVoiceSettings() { function loadVoiceSettings() {
//Globalny default 480p@24; stare zapisane defaulty (720p/1080p/30fps) const defaults = { micId: "", camId: "", speakerId: "", camRes: "1280x720", camFps: 24, screenRes: "1920x1080", screenFps: 30, sensitivity: VOICE_SENSITIVITY_DEFAULT };
//migrują na nowe, świadome wybory użytkownika zostają.
const defaults = { micId: "", camId: "", speakerId: "", camRes: "640x480", camFps: 24, screenRes: "640x480", screenFps: 24, sensitivity: VOICE_SENSITIVITY_DEFAULT };
try { try {
let s = Object.assign({}, JSON.parse(localStorage.getItem(VOICE_SETTINGS_KEY)) || {}); return Object.assign(defaults, JSON.parse(localStorage.getItem(VOICE_SETTINGS_KEY)) || {});
if (s.camRes === "1280x720") s.camRes = defaults.camRes;
if (s.screenRes === "1920x1080") s.screenRes = defaults.screenRes;
if (s.screenFps === 30 || s.screenFps === "30") s.screenFps = defaults.screenFps;
return Object.assign(defaults, s);
} catch (e) { } catch (e) {
return defaults; return defaults;
} }
@ -293,35 +274,6 @@ async function applyVoiceSpeakerSink() {
let voiceConnectedAt = null; let voiceConnectedAt = null;
let voiceTimerInterval = null; let voiceTimerInterval = null;
let voicePingMs = null; let voicePingMs = null;
//Kontrola congestionu: bufferedAmount nie widzi backlogu kernela TCP, prawdziwy
//obraz daje RTT z pingów. Baseline = MINIMUM z okna próbek (odporne na skoki
//i na peera sztucznie opóźniającego własne echa - stałe kłamstwo podnosi minimum).
let voiceRttFastMs = 0;
let voiceRttSamples = [];
let voiceRttBaseMs = 0;
let voiceAudioHoldUntil = 0; //histereza zaworu audio
//Echo-ping (typ 5 = prośba, typ 6 = echo): RTT wyłącznie na WŁASNYM zegarze -
//wysyłam timestamp, peer odbija te same bajty, odejmuję czas nadania. Zegar
//peera nigdy nie jest używany (może być przekręcony lub spreparowany).
const voiceOutstandingPings = new Map(); // BigInt(ts) -> czas nadania
let voiceLastEchoAt = 0; // rate-limit odpowiedzi echem
let voiceLastRttSampleAt = 0; // kiedy ostatnio spłynął prawdziwy pomiar
function voiceCongestionDelay() {
if (!voiceRttFastMs || !voiceRttBaseMs) return 0;
return Math.min(5000, Math.max(0, voiceRttFastMs - voiceRttBaseMs));
}
function voiceRecordRttSample(rtt) {
voiceLastRttSampleAt = Date.now();
voicePingMs = voicePingMs === null ? rtt : Math.round(voicePingMs * 0.7 + rtt * 0.3);
voiceRttFastMs = voiceRttFastMs ? Math.round(voiceRttFastMs * 0.5 + rtt * 0.5) : rtt;
voiceRttSamples.push(rtt);
if (voiceRttSamples.length > 12) voiceRttSamples.shift();
voiceRttBaseMs = Math.min.apply(null, voiceRttSamples);
updateVoicePing();
}
function formatVoiceDuration(ms) { function formatVoiceDuration(ms) {
let totalSec = Math.floor(ms / 1000); let totalSec = Math.floor(ms / 1000);
@ -678,11 +630,7 @@ setInterval(() => {
} }
// Half-open socket watchdog + backgrounded-tab audio recovery. // Half-open socket watchdog + backgrounded-tab audio recovery.
if (currentVoiceWs && currentVoiceWs.readyState === WebSocket.OPEN) { if (currentVoiceWs && currentVoiceWs.readyState === WebSocket.OPEN) {
//deadPeers NIE liczą się do "others": ghost na rosterze (martwy TCP const others = voiceActiveUsers.filter(u => u !== String(window.id)).length;
//jeszcze nie zsynchronizowany przez serwer) nigdy nic nie wyśle, więc
//zliczany jak żywy wymuszał reconnecty w pętli - migające "Connecting"
//w zdrowej rozmowie.
const others = voiceActiveUsers.filter(u => u !== String(window.id) && !deadPeers.has(u)).length;
// Baseline ciszy odświeżany przy przejściu 0->N uczestników: solo w // Baseline ciszy odświeżany przy przejściu 0->N uczestników: solo w
// rozmowie NIE MA żadnego ruchu przychodzącego, więc po pojawieniu się // rozmowie NIE MA żadnego ruchu przychodzącego, więc po pojawieniu się
// nowej osoby lastVoiceServerPacketAt bywa nieświeży z okresu solo. // nowej osoby lastVoiceServerPacketAt bywa nieświeży z okresu solo.
@ -745,8 +693,8 @@ class AudioReceiver extends AudioWorkletProcessor {
// PREBUF_START - initial prebuffer before first playback (~43 ms) // PREBUF_START - initial prebuffer before first playback (~43 ms)
// PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms) // PREBUF_RESUME - re-prebuffer after a starvation event (~32 ms)
// HARD_LIMIT - burst ceiling (~128 ms): snap back to target immediately // HARD_LIMIT - burst ceiling (~128 ms): snap back to target immediately
// Backlog above target is drained adaptively (proportional to excess, see // Backlog slightly above target is drained by a gentle trim (1 sample per
// process()) so latency returns to ~43 ms quickly after network bursts. // quantum) instead of letting latency creep up to the hard limit.
constructor(options) { constructor(options) {
super(); super();
this.users = {}; this.users = {};
@ -784,12 +732,7 @@ class AudioReceiver extends AudioWorkletProcessor {
if (available < output.length) { user.buffering = true; } if (available < output.length) { user.buffering = true; }
else { else {
if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536; if (available > 6144) user.readPtr = (user.writePtr - 2048 + 65536) % 65536;
else if (available > 2049) { else if (available > 2049) user.readPtr = (user.readPtr + 1) % 65536;
//Adaptacyjny drain: nadmiar nad target odprowadzany proporcjonalnie
//(1 + excess/256 /kwant) - powrót z ~128 ms do ~43 ms w ułamku sekundy
const excess = available - 2048;
user.readPtr = (user.readPtr + 1 + (excess >> 8)) % 65536;
}
let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0; let vol = this.volumes[id] !== undefined ? this.volumes[id] : 1.0;
for (let i = 0; i < output.length; i++) { for (let i = 0; i < output.length; i++) {
output[i] += (user.buffer[user.readPtr] / 32768) * vol; output[i] += (user.buffer[user.readPtr] / 32768) * vol;
@ -947,38 +890,6 @@ function bytesToB64u(a) {
return btoa(bin).replace(/\+/g, "-").replace(/\//g, "_").replace(/=+$/, ""); return btoa(bin).replace(/\+/g, "-").replace(/\//g, "_").replace(/=+$/, "");
} }
//Roster pokoju z voice-servera (źródło prawdy) -> synchronizacja kafelków.
//Steady-poll co VOICE_ROSTER_POLL_MS - bez tego sweep ghosta/leave bez
//call_update nigdy nie docierał do UI.
async function fetchVoiceRoster() {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !currentVoiceCallId) return;
try {
const baseUrl = currentVoiceWs.url.split('/ws')[0];
const httpUrl = baseUrl.replace('ws:', 'http:').replace('wss:', 'https:');
const t0 = performance.now();
const res = await fetch(`${httpUrl}/room/users?room=${encodeURIComponent(currentVoiceCallId)}`);
const usersRaw = await res.json();
//HTTP tylko jako szacunek gdy brak prawdziwych próbek RTT (solo / stary peer)
if (voicePingMs === null || Date.now() - voiceLastRttSampleAt > 25000) {
voicePingMs = Math.round(performance.now() - t0);
updateVoicePing();
}
const rawUsers = usersRaw.map(u => String(u));
//Serwer jest źródłem prawdy o składzie - nie filtrujemy po
//deadPeers (audyt S15: ukryty peer nadal dostaje audio)
for (const gone of [...deadPeers]) {
if (!rawUsers.includes(gone)) deadPeers.delete(gone);
}
const users = rawUsers;
let changed = (users.length !== voiceActiveUsers.length);
for(let u of users) if(!voiceActiveUsers.includes(u)) changed = true;
if(changed) {
voiceActiveUsers = users;
updateVoiceUIUsers();
}
} catch (e) {}
}
async function startVoiceSession(serverUrl, ticket) { async function startVoiceSession(serverUrl, ticket) {
if (!ticket) throw new Error("missing voice ticket"); if (!ticket) throw new Error("missing voice ticket");
// Single-session invariant: a stale socket (e.g. a double-fired join) // Single-session invariant: a stale socket (e.g. a double-fired join)
@ -1021,7 +932,32 @@ async function startVoiceSession(serverUrl, ticket) {
else btn.classList.remove("voice-btn-danger"); else btn.classList.remove("voice-btn-danger");
} }
fetchVoiceRoster(); const fetchUsers = async () => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !currentVoiceCallId) return;
try {
const baseUrl = currentVoiceWs.url.split('/ws')[0];
const httpUrl = baseUrl.replace('ws:', 'http:').replace('wss:', 'https:');
const t0 = performance.now();
const res = await fetch(`${httpUrl}/room/users?room=${currentVoiceCallId}`);
const usersRaw = await res.json();
voicePingMs = Math.round(performance.now() - t0);
updateVoicePing();
const rawUsers = usersRaw.map(u => String(u));
//Serwer jest źródłem prawdy o składzie - nie filtrujemy po
//deadPeers (audyt S15: ukryty peer nadal dostaje audio)
for (const gone of [...deadPeers]) {
if (!rawUsers.includes(gone)) deadPeers.delete(gone);
}
const users = rawUsers;
let changed = (users.length !== voiceActiveUsers.length);
for(let u of users) if(!voiceActiveUsers.includes(u)) changed = true;
if(changed) {
voiceActiveUsers = users;
updateVoiceUIUsers();
}
} catch (e) {}
};
fetchUsers();
//The voice-server roster is joined only AFTER ticket validation (an //The voice-server roster is joined only AFTER ticket validation (an
//HTTP roundtrip to the main server), so the immediate fetch can still //HTTP roundtrip to the main server), so the immediate fetch can still
//miss us - retry quickly until our own tile shows up instead of //miss us - retry quickly until our own tile shows up instead of
@ -1035,17 +971,13 @@ async function startVoiceSession(serverUrl, ticket) {
voiceRosterRetryTimer = null; voiceRosterRetryTimer = null;
return; return;
} }
fetchVoiceRoster(); fetchUsers();
}, 400); }, 400);
voiceReconnectAttempts = 0; voiceReconnectAttempts = 0;
lastVoiceServerPacketAt = Date.now(); lastVoiceServerPacketAt = Date.now();
if (voicePingTimer) clearInterval(voicePingTimer); if (voicePingTimer) clearInterval(voicePingTimer);
voicePingTimer = setInterval(sendVoicePing, VOICE_PING_INTERVAL_MS); voicePingTimer = setInterval(sendVoicePing, VOICE_PING_INTERVAL_MS);
//Steady roster poll - utrzymuje kafelki zgodne z prawdą serwera
//(ghost-sweep, leave bez call_update) przez całą rozmowę.
if (voiceRosterPollTimer) clearInterval(voiceRosterPollTimer);
voiceRosterPollTimer = setInterval(fetchVoiceRoster, VOICE_ROSTER_POLL_MS);
//S16: getUserMedia/worklet/rezump AudioContext potrafią rzucić //S16: getUserMedia/worklet/rezump AudioContext potrafią rzucić
//(brak uprawnień, zniknięte urządzenie) - bez obsługi zostawał //(brak uprawnień, zniknięte urządzenie) - bez obsługi zostawał
@ -1101,18 +1033,12 @@ async function startVoiceSession(serverUrl, ticket) {
} }
let vadHangover = 0; let vadHangover = 0;
const myIdStr = String(window.id);
senderNode.port.onmessage = async (e) => { senderNode.port.onmessage = async (e) => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return; if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
//Zawór bajtowy + zawór RTT (kolejka kernela jest niewidoczna dla // Backpressure valve: when the socket is congested, drop mic
//bufferedAmount). Drop >450 ms nadwyżki z histerezą 700 ms - lepiej // frames instead of queueing them - queued audio would arrive
//urwać ułamek sekundy niż mówić z sekundami opóźnienia. // minutes late and drown the connection for every stream.
if (currentVoiceWs.bufferedAmount > VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES) { if (currentVoiceWs.bufferedAmount > 256 * 1024) {
vadHangover = 0;
return;
}
if (voiceCongestionDelay() > 450) voiceAudioHoldUntil = Date.now() + 700;
if (Date.now() < voiceAudioHoldUntil) {
vadHangover = 0; vadHangover = 0;
return; return;
} }
@ -1127,7 +1053,7 @@ async function startVoiceSession(serverUrl, ticket) {
if (db >= sensitivity) vadHangover = 75; if (db >= sensitivity) vadHangover = 75;
let isSpeaking = vadHangover > 0 && !isMicMuted; let isSpeaking = vadHangover > 0 && !isMicMuted;
setTileSpeaking(myIdStr, isSpeaking); setTileSpeaking(String(window.id), isSpeaking);
if (vadHangover > 0 && !isMicMuted) { if (vadHangover > 0 && !isMicMuted) {
vadHangover--; vadHangover--;
@ -1158,44 +1084,7 @@ async function startVoiceSession(serverUrl, ticket) {
if (deadPeers.has(clientId)) deadPeers.delete(clientId); if (deadPeers.has(clientId)) deadPeers.delete(clientId);
const type = view.getUint8(8); const type = view.getUint8(8);
if (type === 5 || type === 6) { if (type === 5) return; // liveness ping - presence already recorded
//Liveness już zapisany wyżej. Dalej tylko przy zainstalowanym kluczu.
if (voiceAesKeyBytes) {
const senderKey = getVoicePeerSenderKey(clientId);
if (senderKey) {
//Offsety ODBIORCZE: serwer dokleja 8B nadawcy, więc iv=[10..22),
//ct=[22..). Format wysyłki (iv=2, ct=14) jest INNY - pomylenie
//psuje każdą deszyfrację cicho (GCM auth fail -> .catch)
const piv = new Uint8Array(arrayBuffer, 10, 12);
const pct = new Uint8Array(arrayBuffer, 22);
crypto.subtle.decrypt({name: "AES-GCM", iv: piv}, senderKey, pct).then(plain => {
if (plain.byteLength !== 8) return; //stary klient: ping bez timestampu
const tsData = new DataView(plain).getBigUint64(0);
if (type === 5) {
//Odbij te same bajty własnym podkluczem i świeżym IV; rate-limit
const nowMs = Date.now();
if (nowMs - voiceLastEchoAt < 150) return;
voiceLastEchoAt = nowMs;
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
const riv = nextVoiceIv();
crypto.subtle.encrypt({name: "AES-GCM", iv: riv}, voiceMySenderCryptoKey, plain).then(rcipher => {
if (!currentVoiceWs || currentVoiceWs.readyState !== WebSocket.OPEN) return;
const rpayload = new Uint8Array(14 + rcipher.byteLength);
rpayload[0] = 6; rpayload[1] = 0; rpayload.set(riv, 2); rpayload.set(new Uint8Array(rcipher), 14);
currentVoiceWs.send(rpayload.buffer);
}).catch(() => {});
} else {
//Typ 6: powrót własnego timestampu - RTT na WŁASNYM zegarze.
if (!voiceOutstandingPings.has(tsData)) return;
voiceOutstandingPings.delete(tsData);
const rtt = Date.now() - Number(tsData);
if (rtt >= 1 && rtt < 15000) voiceRecordRttSample(rtt);
}
}).catch(() => {});
}
}
return;
}
if (!voiceActiveUsers.includes(clientId) && clientId !== String(window.id)) { if (!voiceActiveUsers.includes(clientId) && clientId !== String(window.id)) {
const baseUrl = currentVoiceWs.url.split('/ws')[0]; const baseUrl = currentVoiceWs.url.split('/ws')[0];
@ -1276,7 +1165,6 @@ async function startVoiceSession(serverUrl, ticket) {
frame.close(); frame.close();
}, },
error: e => { error: e => {
voiceLogError('videodecoder[' + decoderId + ']', e);
if (remoteVideos[decoderId]) { if (remoteVideos[decoderId]) {
try { remoteVideos[decoderId].decoder.close(); } catch(err) {} try { remoteVideos[decoderId].decoder.close(); } catch(err) {}
remoteVideos[decoderId].decoder = null; remoteVideos[decoderId].decoder = null;
@ -1302,14 +1190,11 @@ async function startVoiceSession(serverUrl, ticket) {
try { try {
streamData.decoder.decode(chunk); streamData.decoder.decode(chunk);
} catch(err) { } catch(err) {
voiceLogError('videodecode[' + decoderId + ']', err);
try { streamData.decoder.close(); } catch(e){} try { streamData.decoder.close(); } catch(e){}
streamData.decoder = null; streamData.decoder = null;
streamData.hasKeyFrame = false; streamData.hasKeyFrame = false;
} }
} catch (err) { } catch (err) {}
voiceLogError('vdecrypt[' + decoderId + ']', err);
}
}); });
} else if (type === 3) { } else if (type === 3) {
(async () => { (async () => {
@ -1447,6 +1332,8 @@ async function handleCallUpdate(data) {
for (let u in remoteAudio) { for (let u in remoteAudio) {
let pureU = u.replace("_screen", ""); let pureU = u.replace("_screen", "");
if (!voiceActiveUsers.includes(pureU) && pureU !== myIdStr) { if (!voiceActiveUsers.includes(pureU) && pureU !== myIdStr) {
if (audioDecoders[u]) try { audioDecoders[u].close(); } catch(e){}
delete audioDecoders[u];
delete remoteAudio[u]; delete remoteAudio[u];
const rn = u.endsWith("_screen") ? screenReceiverNode : receiverNode; const rn = u.endsWith("_screen") ? screenReceiverNode : receiverNode;
if(rn) rn.port.postMessage({ action: 'remove', clientId: u }); if(rn) rn.port.postMessage({ action: 'remove', clientId: u });
@ -2603,6 +2490,7 @@ function toggleVoiceMic() {
// keeping the UI, mic stream and AudioContext alive for an automatic rejoin. // keeping the UI, mic stream and AudioContext alive for an automatic rejoin.
function teardownVoiceTransport() { function teardownVoiceTransport() {
if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; } if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; }
if (voiceWsHeartbeat) { clearInterval(voiceWsHeartbeat); voiceWsHeartbeat = null; }
if (camControl) { try { camControl.stop(); } catch (e) {} camControl = null; } if (camControl) { try { camControl.stop(); } catch (e) {} camControl = null; }
if (screenControl) { try { screenControl.stop(); } catch (e) {} screenControl = null; } if (screenControl) { try { screenControl.stop(); } catch (e) {} screenControl = null; }
if (currentVoiceWs) { if (currentVoiceWs) {
@ -2619,7 +2507,6 @@ function teardownVoiceTransport() {
peerLastSeen = {}; peerLastSeen = {};
deadPeers.clear(); deadPeers.clear();
if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; } if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; }
if (voiceRosterPollTimer) { clearInterval(voiceRosterPollTimer); voiceRosterPollTimer = null; }
lastVoiceKeyRequestAt = 0; lastVoiceKeyRequestAt = 0;
//Stale participants from the previous attempt must not linger in the UI //Stale participants from the previous attempt must not linger in the UI
voiceActiveUsers = []; voiceActiveUsers = [];
@ -2629,12 +2516,6 @@ function teardownVoiceTransport() {
setVoiceRootKey(null); setVoiceRootKey(null);
voiceSendEpoch = null; voiceSendEpoch = null;
voiceSendCounter = 0n; voiceSendCounter = 0n;
voiceRttFastMs = 0;
voiceRttSamples = [];
voiceRttBaseMs = 0;
voiceAudioHoldUntil = 0;
voiceOutstandingPings.clear();
voiceLastRttSampleAt = 0;
lastVoiceServerPacketAt = Date.now(); lastVoiceServerPacketAt = Date.now();
} }
@ -2651,10 +2532,6 @@ function scheduleVoiceReconnect() {
voiceReconnectAttempts++; voiceReconnectAttempts++;
setVoiceStatus("connecting"); setVoiceStatus("connecting");
teardownVoiceTransport(); teardownVoiceTransport();
//Zmiana sieci = stary RTT dotyczy martwej ścieżki; bez resetu wskaźnik
//wisiałby zamrożony, a fallback HTTP nie miał prawa go nadpisać
voicePingMs = null;
updateVoicePing();
setTimeout(async () => { setTimeout(async () => {
if (!currentVoiceCallId || !currentVoiceCallType) return; if (!currentVoiceCallId || !currentVoiceCallType) return;
try { try {
@ -2666,7 +2543,7 @@ function scheduleVoiceReconnect() {
console.warn("Voice reconnect failed", err); console.warn("Voice reconnect failed", err);
scheduleVoiceReconnect(); scheduleVoiceReconnect();
} }
}, Math.min(4000, 800 * voiceReconnectAttempts)); }, 800);
} }
function endVoiceSession() { function endVoiceSession() {
@ -2706,27 +2583,24 @@ function endVoiceSession() {
micStream = null; micStream = null;
} }
if (micSourceNode) { micSourceNode.disconnect(); micSourceNode = null; } if (micSourceNode) { micSourceNode.disconnect(); micSourceNode = null; }
if (audioEncoder) { try { audioEncoder.close(); } catch(e){} audioEncoder = null; }
if (voiceWsHeartbeat) { clearInterval(voiceWsHeartbeat); voiceWsHeartbeat = null; }
if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; } if (voicePingTimer) { clearInterval(voicePingTimer); voicePingTimer = null; }
if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; } if (voiceRosterRetryTimer) { clearInterval(voiceRosterRetryTimer); voiceRosterRetryTimer = null; }
if (voiceRosterPollTimer) { clearInterval(voiceRosterPollTimer); voiceRosterPollTimer = null; }
peerLastSeen = {}; peerLastSeen = {};
deadPeers.clear(); deadPeers.clear();
lastVoiceKeyRequestAt = 0; lastVoiceKeyRequestAt = 0;
voiceReconnectAttempts = 0;
voiceActiveUsers = []; voiceActiveUsers = [];
voiceHadRemotePeers = false; voiceHadRemotePeers = false;
voiceRttFastMs = 0;
voiceRttSamples = [];
voiceRttBaseMs = 0;
voiceAudioHoldUntil = 0;
voiceOutstandingPings.clear();
voiceLastEchoAt = 0;
voiceLastRttSampleAt = 0;
for (let id in audioDecoders) {
try { audioDecoders[id].close(); } catch(e){}
}
for (let id in audioDecryptChains) { for (let id in audioDecryptChains) {
if(audioDecryptChains[id]) audioDecryptChains[id].clear(); if(audioDecryptChains[id]) audioDecryptChains[id].clear();
} }
audioDecoders = {};
audioDecryptChains = {}; audioDecryptChains = {};
remoteAudio = {}; remoteAudio = {};
@ -2798,11 +2672,8 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
screenSenderNode = new AudioWorkletNode(audioContext, 'audio-sender'); screenSenderNode = new AudioWorkletNode(audioContext, 'audio-sender');
screenSenderNode.port.onmessage = async (e) => { screenSenderNode.port.onmessage = async (e) => {
if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
// Ten sam zawór co mikrofon (limit bajtów + korek RTT) - audio // Same backpressure valve as the mic sender.
// ma pierwszeństwo przed wideo na wspólnym sockecie. if (currentVoiceWs.bufferedAmount > 256 * 1024) return;
if (currentVoiceWs.bufferedAmount > VOICE_WS_AUDIO_QUEUE_LIMIT_BYTES) return;
if (voiceCongestionDelay() > 450) voiceAudioHoldUntil = Date.now() + 700;
if (Date.now() < voiceAudioHoldUntil) return;
const iv = nextVoiceIv(); const iv = nextVoiceIv();
const cipher = await crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, e.data.audioData); const cipher = await crypto.subtle.encrypt({name: "AES-GCM", iv}, voiceMySenderCryptoKey, e.data.audioData);
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
@ -2832,24 +2703,7 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
return { width, height, fps, bitrate }; return { width, height, fps, bitrate };
}; };
let currentConfig = getCurrentConfig(), isRunning = true, lastKeyframeAt = 0; let currentConfig = getCurrentConfig(), isRunning = true, frameCount = 0;
//AIMD + slow-start: start na 40% celu (pełną prędkością waliłby w nieznany
//uplink, a bufory TCP w kernelu to MB-y = sekundy kolejki zanim cokolwiek
//zmierzymy). Sygnał korka: skipy (zawór 12 KB) lub nadwyżka RTT.
let brScale = 0.4, encTries = 0, encSkips = 0, brWindowAt = performance.now(), lastBrReconfAt = 0;
//Korek trwa mimo podłogi bitrate'u → FPS spada do 1/2, 1/3
let fpsDiv = 1, lowStreak = 0, cleanStreak = 0, tickCount = 0;
let activeBitrate = Math.round(currentConfig.bitrate * brScale);
const applyEncoderBitrate = (force = false) => {
const target = Math.max(120000, Math.round(currentConfig.bitrate * brScale));
if (!force && Math.abs(target - activeBitrate) / activeBitrate < 0.08) return;
if (!force && performance.now() - lastBrReconfAt < 800) return;
activeBitrate = target;
lastBrReconfAt = performance.now();
try {
encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: activeBitrate, latencyMode: 'realtime'});
} catch (e) {}
};
const encoder = new VideoEncoder({ const encoder = new VideoEncoder({
output: async (chunk) => { output: async (chunk) => {
if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return; if (currentVoiceWs?.readyState !== WebSocket.OPEN || !voiceMySenderCryptoKey) return;
@ -2859,9 +2713,9 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
const payload = new Uint8Array(14 + cipher.byteLength); const payload = new Uint8Array(14 + cipher.byteLength);
payload[0] = typeId; payload[1] = chunk.type === 'key' ? 1 : 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14); payload[0] = typeId; payload[1] = chunk.type === 'key' ? 1 : 0; payload.set(iv, 2); payload.set(new Uint8Array(cipher), 14);
currentVoiceWs.send(payload.buffer); currentVoiceWs.send(payload.buffer);
}, error: (e) => { voiceLogError('videoencoder', e); stopFunction(); } }, error: (e) => { stopFunction(); }
}); });
encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: activeBitrate, latencyMode: 'realtime'}); encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: currentConfig.bitrate, latencyMode: 'realtime'});
// Transient capture-mode wobbles must not thrash the encoder - // Transient capture-mode wobbles must not thrash the encoder -
// commit only after ~300 ms stable. Camera size/fps changes arrive // commit only after ~300 ms stable. Camera size/fps changes arrive
@ -2879,62 +2733,20 @@ async function startHardwareVideo(btnEl, streamPromise, typeLabel, typeId) {
pendingCfg = newConfig; pendingCfg = newConfig;
pendingAt = performance.now(); pendingAt = performance.now();
} else if (performance.now() - pendingAt > 300) { } else if (performance.now() - pendingAt > 300) {
//Zmiana geometrii źródła - uczony scale łącza zostaje, currentConfig = newConfig; forceKeyframe = true; frameCount = 0; pendingCfg = null;
//resetujemy tylko redukcję FPS encoder.configure({codec: 'av01.0.04M.08', width: currentConfig.width, height: currentConfig.height, framerate: currentConfig.fps, bitrate: currentConfig.bitrate, latencyMode: 'realtime'});
currentConfig = newConfig; forceKeyframe = true; lastKeyframeAt = 0; pendingCfg = null;
activeBitrate = Math.round(currentConfig.bitrate * brScale);
fpsDiv = 1; lowStreak = 0; cleanStreak = 0;
applyEncoderBitrate(true);
} }
} else if (pendingCfg) { } else if (pendingCfg) {
pendingCfg = null; // wiggle reverted on its own - ignore pendingCfg = null; // wiggle reverted on its own - ignore
} }
//Zawór kolejki socketu (12 KB) ustępuje DALEJ przed progiem audio if (encoder.state === 'configured' && encoder.encodeQueueSize <= 2 && currentVoiceWs.bufferedAmount < 1024 * 256 && videoEl.readyState >= 2) {
//(14 KB) - przy korku najpierw znika obraz, głos płynie dalej.
tickCount++;
const policySkip = fpsDiv > 1 && (tickCount % fpsDiv !== 0);
let skipped = false;
if (policySkip) {
//celowe tempo (fpsDiv) - nie liczy się do skipRatio
} else if (encoder.state === 'configured' && encoder.encodeQueueSize <= 2 &&
currentVoiceWs.readyState === WebSocket.OPEN &&
currentVoiceWs.bufferedAmount < VOICE_WS_VIDEO_QUEUE_LIMIT_BYTES && videoEl.readyState >= 2) {
let frame = null; let frame = null;
try { try {
const nowMs = performance.now(); frame = new VideoFrame(videoEl, {timestamp: Math.trunc(performance.now() * 1000)});
frame = new VideoFrame(videoEl, {timestamp: Math.trunc(nowMs * 1000)}); frameCount++;
//Keyframe wg zegara co ~1 s (nie licznika klatek - zawór je pomija) encoder.encode(frame, {keyFrame: forceKeyframe || (frameCount % (currentConfig.fps * 2)) === 0});
const wantKeyframe = forceKeyframe || (nowMs - lastKeyframeAt >= 1000);
encoder.encode(frame, {keyFrame: wantKeyframe});
if (wantKeyframe) lastKeyframeAt = nowMs;
} catch (err) { } finally { if (frame) frame.close(); } } catch (err) { } finally { if (frame) frame.close(); }
} else if (encoder.state === 'configured') {
skipped = true;
}
encTries++;
if (skipped) encSkips++;
const winNow = performance.now();
if (winNow - brWindowAt >= 800) {
const skipRatio = encTries > 0 ? encSkips / encTries : 0;
const congestionMs = voiceCongestionDelay();
if (congestionMs > 450 || skipRatio > 0.3 || congestionMs > 250) {
brScale = Math.max(0.15, brScale * (congestionMs > 450 ? 0.6 : 0.75));
applyEncoderBitrate();
if (brScale <= 0.17) lowStreak++; else lowStreak = 0;
if (lowStreak >= 4 && fpsDiv < 3) { fpsDiv++; lowStreak = 0; }
cleanStreak = 0;
} else if (skipRatio === 0 && congestionMs < 80) {
cleanStreak++;
lowStreak = 0;
if (cleanStreak >= 5) {
if (fpsDiv > 1) { fpsDiv--; }
else if (brScale < 1) { brScale = Math.min(1, brScale * 1.2); applyEncoderBitrate(); }
cleanStreak = 0;
}
}
encTries = 0; encSkips = 0; brWindowAt = winNow;
} }
setTimeout(encodeLoop, 1000 / currentConfig.fps); setTimeout(encodeLoop, 1000 / currentConfig.fps);
}; };