You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
fix(devbox): give long-poll requests a client read timeout above the server hold
The wait_for_status and exec await long-poll POSTs forwarded a timeout_seconds
to the server telling it how long to hold the connection, but left the client
read timeout at the global 30s default. The devbox wait_for_status endpoint
clamps its hold to 30s as well, so the two race: when the client read fires
first it aborts the stream (RST_STREAM) and surfaces a connection error instead
of the server's graceful 408.
Set a per-request read timeout of (server hold + 5s) on every long-poll sender,
capping the forwarded timeout_seconds at the server's clamp (30s status, 25s
exec) and preserving a 5s connect timeout. The server now always returns 408
first and the poll loop re-polls cleanly.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
0 commit comments