dangernoodle-io / breadboard / 27518378721

100%

push

github

feat(bb_update_check): spawn the TLS worker on-demand instead of holding it persistent (B1-240) (#432)

- replace the persistent semaphore+worker task model with a one-shot
  xTaskCreatePinnedToCore per timer tick / kick; the 8 KB mbedTLS stack
  is dynamically allocated only during a fetch and freed via vTaskDelete(NULL)
- add atomic_bool s_check_in_flight concurrency guard: atomic_compare_exchange
  before every spawn ensures at most one one-shot task runs at a time; if
  heap is too fragmented for the 8 KB alloc at spawn time, warn + skip (next
  interval retries) rather than crashing
- remove s_kick semaphore and s_worker handle; s_timer survives unchanged
- bb_update_check_run_blocking polls last_check_us advance as before; kick()
  now calls try_spawn() instead of xSemaphoreGive
- bb_update_check_set_task_core / set_task_priority take effect on the next
  spawn rather than at init; docs updated accordingly
- host stub kick() gains the same in-flight guard so host tests can verify
  the skip-on-concurrent-kick path without a real FreeRTOS task
- add bb_update_check_set_in_flight_for_test / get_in_flight_for_test test
  hooks (BB_UPDATE_CHECK_TESTING) to the internal header; on esp-idf they
  toggle s_check_in_flight in bb_update_check_espidf.c, on host they toggle
  the mirrored s_in_flight in bb_update_check_common.c
- add four new host tests: kick_clears_in_flight_after_run,
  kick_in_flight_is_noop, in_flight_resets_on_reset_world,
  kick_returns_ok_on_host (updated comment)
- no Kconfig gate: on-demand is the only model (spawn cost for a
  minutes-apart check is negligible; persistent model never needed on
  boot-strategy boards which keep AUTOREGISTER=n)

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

1667 of 1667 branches covered (100.0%)

Branch coverage included in aggregate %.

12 of 12 new or added lines in 1 file covered. (100.0%)

2786 of 2786 relevant lines covered (100.0%)

1110.76 hits per line