lekt8 9aa2445ec7 Reconnect after a failed SEARCH ALL so the email poller doesn't desync IMAP (#1613) (#1748) ...

On a large Gmail mailbox the email-summary poller's SINCE scan often finds
nothing (INTERNALDATE/date-header quirks), so it falls back to SEARCH ALL. That
returns one enormous UID line; the socket read can time out mid-response, and the
exception was swallowed — leaving the unread '* SEARCH 325188 …' bytes on the
socket. The next command (the downstream re-select) then read those leftover
bytes and failed with 'EXAMINE => unexpected response: b'325188 …''.

Extract the fallback into _latest_inbox_fallback_uids(conn, reconnect): on a
failed SEARCH ALL it logs out the poisoned connection and reconnects, returning
the fresh connection for downstream use. Reconnecting is correct by construction
— a new connection cannot carry the old one's leftover bytes — so the re-select
always runs on a clean socket.

The same SEARCH ALL + reuse pattern also exists in mcp_servers/email_server.py
and routes/email_routes.py; left for a separate change to keep this surgical.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-03 13:28:53 +09:00

2.6 KiB

Raw Blame History

View Raw