When no other peer is present, heartbeats, force-sync, and episode auto-sync
are now fully suppressed (previously the keepAlive heartbeat, force-sync, and
episode lobby were still broadcast to an empty room):
- keepAlive heartbeat gated on a live peer count
- force-sync is a no-op when solo (no pause-then-timeout freeze, no traffic)
- episode auto-sync skips the lobby entirely and just plays the next episode
The solo state is recomputed live from the current peer list on every event
(never cached), so the instant another peer joins syncing resumes. On that
solo->not-solo transition the background also requests an immediate heartbeat
from the content script so the newcomer sees the current position without
waiting for the next interval.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Room ID, password, and username were resurrected from synced storage on a
fresh install (sync survives an uninstall in the user's Google account), which
made the extension silently auto-connect to a dead room and appear permanently
connected. getSettings() and all settings reads (popup, content, audio-options,
applyAudioSettingsToTab) are now local-only, and legacy keys are actively
purged from sync on install/update/startup. Only onboardingComplete and
dismissedHints remain in sync.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Address client-side rate-limit edge cases that could trip the server's
per-socket event budget or per-IP connection limit:
- Pace the offline event-queue flush after (re)connect in batches
(10 per 3s) instead of one synchronous burst, so a reconnect after a
long outage no longer dumps the whole backlog and gets kicked.
- Wrap socket.send() in try/catch and re-queue on failure (race with a
server-side disconnect) instead of losing the event.
- Ping liveness: tolerate one missed PONG; only force a reconnect after
2 consecutive misses (~20s) instead of a single hard 5s timeout.
- Reconnect backoff: tune to ~8 attempts/60s (under the connection limit)
and add +/-20% jitter to de-synchronize reconnect herds.
All magic numbers extracted into named constants.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
3-way merge of Kaia-Alenia's tab-manager refactor (PR #11) onto current
main (v2.4.3). Only the tab-related service-worker listeners move out of
background.js into extension/modules/tab-manager.js via initTabManager({...})
dependency injection; all other recent main changes are preserved untouched.
Behavior-preserving; listeners still register synchronously at top level.
The unrelated website/assets/NewLogoIcon_64.webp change from the PR was
excluded to keep main clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Switch the Ko-Fi support links to https://support.koalastuff.net across the
README badge, the extension footer (popup.html/template.html, label
"Support KoalaSync") and remove the now-unused KOFI_URL constant. Stop
generating/sending the anonymous uninstall token: initUninstallURL() now
registers the feedback URL without the `t` parameter or storage write.
CHANGELOG updated to match.
Also guard the client CMD_ACK path: only ACK a command sender that is still
a known peer in the room, so we don't emit ACKs to peers that already left
(which the server now drops as absent-peer ACKs anyway).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Media players fire bursts of native play/pause events (source swaps, ABR,
ads, teardown). Each was relayed as a distinct command, spamming peers and
tripping the relay's event rate limit. Add leading+trailing coalescing in
the content script: emit the first event instantly (zero added latency) and,
on a 150ms window, collapse a burst to its final state. Echo-suppression and
seek-flush stay synchronous on event arrival; the trailing send is never
deduped against the leading one (a remote command may change shared state
mid-window — re-sending is the safe, idempotent choice).
Server: split the EVENT_ACK handler's logging so only a genuine different-room
ACK is logged as [SECURITY]; an ACK to a peer that already left is logged
quietly as [ACKDROP] (verbose-only), so real signals are no longer drowned out.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add two new languages (uk, zh / zh_CN) across the extension and website:
register them in SUPPORTED_LANGUAGES, the Chrome _locales map, the website
build/lang-init/app routing, hreflang/og/schema tags and language selectors.
Fix systematic machine-translation word-sense errors in the new zh/uk
locales (e.g. zh "Status"=地位→状态, "Leave Room"=留出空间→离开房间,
"Play"=玩→播放, "Seek"=寻找→跳转, audio Attack/Release/Knee; uk "Play"=Грати
→Відтворити, "Clear"=ЯСНО→Очистити, "Open"=ВІДЧИНЕНО→Відкрити, peers
"Однолітки"→"Учасники"), translate remaining English leftovers (Room ID,
Custom) and normalise terminology. All 15 locales pass test-locales.cjs.
Note: popup.html and website/template.html also carry the language-selector
additions here; their Ko-Fi→Support label text is part of the branding change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- "No peers yet" state now shows a localized "Copy invite link" button (reuses BTN_COPY_INVITE, present in all 13 locales) so users can share without hunting for the field
- Add v2.4.1 changelog entry covering the onboarding tour improvements, empty-state action, and the mobile landing-page fixes from this session
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Add the previously-orphaned 5th onboarding step (ONBOARDING_5) so the tour ends on a closing card instead of stopping at the username field
- Make step 1 a centered welcome card instead of spotlighting the logo; guard querySelector for target-less steps
- Fix stale static placeholders in popup.html (Step 1 of 3 -> 5, progress 33% -> 20%)
Verified ONBOARDING_5/DONE exist across all 13 locales; rendered with the longest (German) copy.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Move extractEpisodeId() and sameEpisode() to shared episode-utils.js.
background.js imports via ES module. content.js (IIFE) receives injected
copy via build-extension.js marker replacement, eliminating duplication.
- background.js: remove isConnecting=false at line 608 (was set before
WebSocket handshake completed, defeating the guard for the entire
CONNECTING phase). isConnecting is already properly reset in onclose,
onopen/40 handler, catch block, and socket guard path.
- background.js: restore isConnecting=false in forceDisconnect() so
subsequent connect() calls can proceed after intentional disconnect.
- popup.js: remove hardcoded applyConnectionStatus('disconnected') that
caused a guaranteed red flicker on every popup open. Status is now
set only by the async GET_STATUS response.
If user has a room configured but the background is not connected
(e.g. service worker was killed, startup race, or stale state), the
popup now sends CONNECT on open. Previously it only called GET_STATUS
and showed 'disconnected' with 0 logs — the user had no way to know
whether the extension was working.
- Prevent concurrent connect() by removing isConnecting reset from
forceDisconnect(); the guard in connect() now reliably blocks
re-entry since no external caller can defeat it.
- Reset isConnecting at end of connect() so subsequent calls can
proceed without waiting for the async '40' handler.
- Clear connectIntent and cancel reconnect timer on server ERROR
when no currentRoom exists, preventing infinite reconnect loops
after failed join attempts (e.g. wrong password via invite link).
- Move connectIntent assignment after !roomId guard in
WEB_JOIN_REQUEST to avoid clobbering existing intent on invalid
input.
- Broadcast JOIN_STATUS failure to popup and bridge tabs when
WEB_JOIN_REQUEST receives an invalid room ID, so the website
join page receives feedback instead of hanging forever.
- Clear joinBtnTimeout on CONNECTION_STATUS connected/disconnected
to avoid stale timeout error messages in the popup.
After LEAVE_ROOM or idle-leave, reconnectAttempts was not reset.
GET_STATUS would return 'reconnecting' instead of 'disconnected',
showing wrong status in popup and leaving Retry button visible.
Lazy-connect introduces a new state where the user clicks 'Raum erstellen'
and the WebSocket takes 1-2s to establish (previously always pre-connected).
The old 15s blind timeout would re-enable the button while background was
still connecting, causing silent no-op clicks. Now polls GET_STATUS and
extends the window if still connecting.
- Add connectIntent flag to gate all reconnect attempts
- Only auto-connect on startup if roomId exists in storage
- Close socket + stop reconnect after leaveRoom or idleLeave
- Preserve existing behavior 1:1 when actively in a room
- Guard force-sync state and peer list clearing during transient disconnects
- Guard WEB_JOIN_REQUEST against empty sanitized roomId
KoalaDev