dangernoodle-io / breadboard / 25980529271

100%

push

github

fix(bb_log): detect SSE peer disconnect within 500ms (#241)

sse_task previously only noticed a closed peer on the next write, up
to CONFIG_BB_LOG_SSE_KEEPALIVE_MS (5s default) after disconnect.
Combined with LWIP's 20s FIN_WAIT_2 timeout, each closed SSE session
held a socket for ~25s — exhausting the 8-socket LWIP pool after a
handful of dashboard navigations and causing httpd_accept_conn to
fail with errno 23 (ENFILE).

Add a non-blocking MSG_PEEK recv() at the top of each drain iteration
so peer FIN / ECONNRESET is caught within the 500ms drain interval.
CLOSE_WAIT residency drops from ~5s to ~500ms. Companion TM-side
reduction of CONFIG_LWIP_TCP_FIN_WAIT_TIMEOUT shrinks the FIN_WAIT_2
half.

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

752 of 752 branches covered (100.0%)

Branch coverage included in aggregate %.

1340 of 1340 relevant lines covered (100.0%)

1118.15 hits per line