mirror of
https://github.com/screentinker/screentinker.git
synced 2026-08-13 22:03:13 -06:00
dcd3a05a7e
Replace the resume-only hide-duration heuristic as the AUTHORITATIVE half-open detector with a real server-silence watchdog, so the .wgt self-heals a dead-but-connected socket from ANY cause (network drop, NAT idle timeout, transport death while foregrounded), not just resume. - Central receive-path liveness: markAlive() refreshes lastServerMsgAt on EVERY inbound server message — app events via socket.onAny, and the server's ~15s engine ping via socket.io 'ping' (both client-only signals the server already sends; no server change; heartbeats get no ack). - Watchdog (10s cadence): if socket.connected && authenticated && silent > 35s (2+ missed pings, under engine.io's own ~45s close), treat as half-open and reconnect via the teardown-first connect() -> exactly one socket, #118 re-registers once. - #148 discipline: fires ONLY while socket.connected===true (the state socket.io can't see), so it never races socket.io's own down-socket auto-reconnect; connect() resets liveness so the watchdog and the resume fast-path can't double-fire. Resume path kept as the fast suspend path. - Timers cleared on exit. Client-only; keep-awake+lifecycle remain the leading flap candidate, NOT a confirmed cause. |
||
|---|---|---|
| .. | ||
| app.js | ||
| device-control.js | ||
| pip-overlay.js | ||
| player.js | ||
| schedule-eval.js | ||
| socket.io.min.js | ||