dangernoodle-io / breadboard / 28694352616

100%

push

github

fix: fail-soft the SSE task-pool ensure failure to a retryable 503 (B1-561) (#703)

- events_handler's task-stack-pool BB_ERR_NO_SPACE branch now maps to
  HTTP 503 (same retryable idiom as the sibling payload-pool and
  max_clients acquire failures) instead of a hard 500, since heap
  pressure on sse_task_bundles_ensure() is transient
- factor the bb_err_t -> HTTP status decision into a pure,
  host-testable helper, sse_connect_error_status()
  (components/bb_event_routes/src/sse_connect_error_decision.c),
  mirroring the sse_bundle_decision/sse_pool_reclaim_decision pattern
- unify the client_acquire_ex acquire-failure branch through the same
  helper (behavior-preserving, no status/body change)
- add host tests covering the NO_SPACE->503 and other->500 mapping
- give this task-pool-ensure 503 a distinct body ({"error":"busy"})
  instead of reusing "max_clients", so it's distinguishable from
  genuine max_clients exhaustion and from the reap-gate 503; update
  the OpenAPI enum accordingly
- add a dedicated pool_ensure_deferred counter mirroring the existing
  slot_reuse_deferred idiom exactly (storage, atomicity, reset,
  bb_event_routes_note_pool_ensure_deferred()), surfaced alongside it
  on GET /api/diag/events, with host coverage for the counter and
  updated CLAUDE.md documentation distinguishing all three 503 causes

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

2262 of 2262 branches covered (100.0%)

Branch coverage included in aggregate %.

9 of 9 new or added lines in 2 files covered. (100.0%)

3793 of 3793 relevant lines covered (100.0%)

1203.53 hits per line