dangernoodle-io / breadboard / 28603832147

100%

push

github

feat(bb_mqtt): add receive primitive with fragmented-payload reassembly (#666)

Adds bb_mqtt_on_message()/bb_mqtt_subscribe() reconnect-safety so a
consumer can receive inbound MQTT messages (needed by the upcoming
bb_sub ingress core):

- bb_mqtt_on_message(h, cb, ctx): per-handle receive callback slot — a
  race-free redesign from the original process-wide slot (two independent
  handles, e.g. a telemetry client and an ingress client, can never splice
  bytes across each other's messages).
- fragment-reassembly is now a pure, host-compilable state machine
  (bb_mqtt_reasm_step, components/bb_mqtt/src/bb_mqtt_reassemble.c) shared
  verbatim by the espidf MQTT_EVENT_DATA glue and new host tests covering
  single-shot, 3-fragment concat, oversized-total drop, zero-length
  payload, and mid-message overflow — no test-only mirror of the logic.
- the rx reassembly buffer is per-handle and lazily heap-allocated
  (SPIRAM-preferred) on the first bb_mqtt_on_message(cb != NULL) call, so
  publish-only handles pay zero standing cost.
- bb_mqtt_subscribe() now tracks filters per handle and re-subscribes them
  on every MQTT_EVENT_CONNECTED (reconnect-safe); the tracked-filter
  compare is now truncation-aware (strcmp against the would-be-stored
  form) instead of a raw sizeof-bound strncmp.
- callback contract documented: cb runs synchronously on the esp-mqtt
  event task and must not block or take a lock that could be held
  elsewhere.
- Bump CONFIG_BB_MQTT_RX_BUFFER_BYTES default 256 -> 1024 (also sizes the
  reassembly buffer; real meta/stats payloads exceed 256B).
- Host stub: bb_mqtt_on_message + bb_mqtt_host_inject_message/
  bb_mqtt_host_inject_fragment test hooks (per-handle, same reassembly
  path as device), plus bb_mqtt_host_set_subscribe_fail to cover a
  consumer's subscribe-failure branch (needed by the upcoming
  bb_sub_mqtt commit) without a real broker.

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

1884 of 1884 branches covered (100.0%)

Branch coverage included in aggregate %.

33 of 33 new or added lines in 1 file covered. (100.0%)

3226 of 3226 relevant lines covered (100.0%)

1295.34 hits per line