dangernoodle-io / breadboard / 28638603985

100%

push

github

feat(bb_event_routes,bb_pool): SSE pool idle-reclaim + reuse-deferred-503 telemetry (B1-492) (#683)

- add bb_event_routes_start() (PRE_HTTP tier, mirrors bb_pub_start()): arms a
  deferred periodic timer that reaps ready SSE task-bundle corpses and, once
  0 active clients + 0 acquired bundles + 0 pending corpses remain, destroys
  the lazy heap-backed task pool (0-byte standing cost); recreated lazily on
  next connect. No-op under CONFIG_BB_EVENT_ROUTES_POOL_STATIC=y. Idempotent
  — a second call while the reclaim timer is already armed is a no-op.
- add bb_pool_slots_reap_ready()/bb_pool_slots_pending_count()/
  bb_pool_slots_acquired_count() (SLOTS mode, read-only + GC pass) so a
  consumer can safely determine full pool idleness without acquiring any slot
- add bb_pool on_destroy per-slot finalizer (SLOTS mode, optional, NULL =
  no-op): bb_pool_destroy() invokes it over EVERY slot before freeing the
  backing arena, so a consumer whose acquire callback lazily creates a
  per-slot OS resource (e.g. sse_bundle_on_acquire's done_sem) can release
  it — fixes a done_sem leak on every idle-reclaim destroy cycle
- extract the keep-vs-destroy decision into a pure, host-tested predicate
  (sse_pool_reclaim_decide) — requires all three of active_clients,
  acquired_slots, and pending_corpses to be zero, closing a use-after-free
  window where a bundle is mid-teardown (client already released, task still
  executing) but neither "active" nor yet "pending"
- add slot_reuse_deferred counter to GET /api/diag/events, distinguishing
  transient reap-gate 503s from max_clients-exhaustion 503s
- new Kconfig BB_EVENT_ROUTES_IDLE_RECLAIM_MS (default 20s)
- document the bb_pool acquire-release lifecycle idiom, the transport-neutral
  in-process-subscriber rule, the idle-reclaim vs POOL_STATIC trade-off, and
  the bb_event_routes_start() public-header ESP_PLATFORM-guard exception in
  CLAUDE.md

2049 of 2049 branches covered (100.0%)

Branch coverage included in aggregate %.

10 of 10 new or added lines in 2 files covered. (100.0%)

3543 of 3543 relevant lines covered (100.0%)

1239.83 hits per line