dangernoodle-io / breadboard / 25061954445

100%

push

github

refactor(bb_http): plug esp_http_server leaks across bb_* components (Phase 1) (#120)

Six bb_* components included esp_http_server.h directly and called raw
httpd_* APIs, breaking the portability invariant that bb_http should be
the sole consumer of esp_http_server. Each leak propagated
esp_http_server as a transitive REQUIRES through every consumer of the
leaky component, and the inconsistent handler signatures (httpd_req_t *
vs bb_http_request_t *) blocked future registry work.

bb_http API additions (all thin wrappers; ESP-IDF + Arduino + host
backends):
- bb_http_resp_sendstr(req, str)
- bb_http_resp_send_chunk(req, buf, len)  // SSE; len<0 = strlen, NULL+0 = end
- bb_http_req_sockfd(req)                  // SSE eviction tracking
- bb_http_req_query_key_value(req, key, out, out_len)
- bb_http_req_async_handler_begin/complete // SSE handoff to worker task
- bb_http_unregister_route(server, method, path)

Components migrated (handler sigs, register calls, response helpers,
HTTPD_* status macros → integer codes; esp_http_server dropped from
REQUIRES):
- bb_ota_pull, bb_ota_push, bb_log_routes, bb_openapi, bb_manifest, bb_prov

Notes:
- bb_openapi previously stashed its meta pointer in req->user_ctx
  (raw httpd field). Replaced with a module-level static — safe since
  bb_openapi_register is called once at server init.
- bb_ota_push referenced HTTPD_SOCK_ERR_TIMEOUT (-3) directly; defined
  a local BB_OTA_RECV_TIMEOUT instead of pulling the header.

Verified: all three smokes SUCCESS, host tests 295/295, audit grep
returns zero hits outside bb_http itself.

Public API unchanged for downstream consumers (TM, snugfeather).
Required for Phase 2 (bb_registry component) and Phase 3 (route
provider migration).

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

272 of 272 branches covered (100.0%)

Branch coverage included in aggregate %.

431 of 431 relevant lines covered (100.0%)

146.06 hits per line