dangernoodle-io / breadboard
97%

DEFAULT BRANCH: main
Repo Added
Files 212
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

LAST BUILD ON BRANCH main
branch: main
CHANGE BRANCH
x

coverage: 96.776% (+0.002%) from 96.774%
30707415682

push

github

web-flow 
fix: gate bb_ws_server WS handshake and frame delivery on provisioning state (#1182)

bb_ws_server registers its own httpd_uri_t directly via
httpd_register_uri_handler, bypassing bb_http.c's three existing
bb_http_prov_gate_check() insertion points entirely -- a WS endpoint
registered through this primitive was reachable, handshake and all
subsequent frames, by an unauthenticated SoftAP client regardless of
provisioning state (B1-1348). No in-tree component ships a WS endpoint
today, so nothing shipped is currently exposed, but there was no
lint/fence/test to catch a future consumer falling into the hole.

Two insertion points, not one:

- ws_pre_handshake_gate_cb, wired via httpd_uri_t.ws_pre_handshake_cb,
  runs before ESP-IDF ever sends the WS handshake response (confirmed
  against the vendored ESP-IDF 5.5.4 httpd_uri() source) -- a denied
  handshake never upgrades the connection at all. It also sends an
  explicit httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, NULL) before
  returning ESP_FAIL: verified against httpd_txrx.c that httpd_resp_send
  has no dependency on WS handshake state (the request is still an
  ordinary, not-yet-upgraded HTTP request at this point), so a plain
  HTTP response is exactly as valid here as from any other handler. NULL
  usr_msg makes it byte-identical to what a genuinely-unregistered WS
  path gets from httpd_uri()'s own uri==NULL branch (also NULL usr_msg,
  via httpd_req_handle_err's no-custom-handler default) -- without this,
  a bare rejected handshake (zero bytes, socket just closes) was
  trivially distinguishable on the wire from that genuine miss, letting
  an unauthenticated SoftAP client fingerprint that a WS endpoint exists
  at a path without completing a handshake.
- ws_shim_handler re-checks the gate on every subsequent call (the
  handshake-completion call AND every real data frame after it), because
  a WS session is long-lived: it can complete its handshake while
  provisioning is inactive, then h... (continued)

8105 of 8678 branches covered (93.4%)

Branch coverage included in aggregate %.

11 of 11 new or added lines in 1 file covered. (100.0%)

13751 of 13906 relevant lines covered (98.89%)

2580.02 hits per line

Relevant lines Covered
13906 RELEVANT LINES 13751 COVERED LINES
2580.02 HITS PER LINE
Source Files on main

Recent builds

Builds Branch Commit Type Ran Committer Via Coverage
30707415682 main fix: gate bb_ws_server WS handshake and frame delivery on provisioning state (#1182) bb_ws_server registers its own httpd_uri_t directly via httpd_register_uri_handler, bypassing bb_http.c's three existing bb_http_prov_gate_check() insertion poin... push web-flow github
96.78
30704714615 main fix: init-marker-gated-srcs lint misses in-file Kconfig-gated definition bodies (B1-1347) (#1181) - extend init-marker-gated-srcs to catch a // bbtool:init marker's fn whose SRCS entry is unconditional but whose definition BODY is wrapped in an i... push web-flow github
96.77
30689651483 main refactor: remove BB_MQTT_CLIENT_AUTOREGISTER knob (B1-1331) (#1180) Composition knobs die per northstar; wiring is codegen-or-handwire only, never a self-registration Kconfig toggle. The knob was default y with nothing overriding it anywhere in-t... push web-flow github
96.77
30682570778 main feat: lint guard for // bbtool:init markers naming a Kconfig-gated-only fn (B1-1337) (#1179) Add lint rule init-marker-gated-srcs: hard-errors when a // bbtool:init marker's fn= names a function whose ONLY definition lives in a .c file conditiona... push web-flow github
96.77
30680483108 main fix: unconditionally compile bb_system_boot_banner.c, gate body internally (#1178) Another instance of the Kconfig-gated-SRCS defect class (companion to B1-1336): bb_system's CMakeLists.txt had a second, separate if(CONFIG_BB_SYSTEM_BOOT_BANNER) ... push web-flow github
96.77
30679710228 main fix: scan CMakeLists.txt for di_legacy AUTOREGISTER/AUTO_ATTACH usage (#1177) The di_legacy fence's _config_usages scanner (backing autoregister_usage/auto_attach_usage) globbed only *.c/*.h/*.cpp, so a Kconfig-gated usage form in a build file wa... push web-flow github
96.77
30679398676 main fix: unconditionally declare bb_diag_routes_init/bb_diag_sections_init (B1-1336) (#1176) bb_diag_http's CMakeLists.txt Kconfig-gated bb_diag_http_routes.c and bb_diag_http_section_dispatch.c on CONFIG_BB_DIAG_ROUTES/ CONFIG_BB_DIAG_SECTIONS, whil... push web-flow github
96.77
30678983253 main refactor: remove ungated bb_mdns_cache_init() codegen no-op (B1-1340) (#1175) PR #1173 (B1-1329) removed BB_MDNS_CACHE_AUTOREGISTER and collapsed bb_mdns_cache_init() to an unconditional no-op returning BB_OK, but the bbtool:init marker on its de... push web-flow github
96.77
30678118823 main refactor: remove BB_SYSTEM_ROUTES_AUTOREGISTER Kconfig gate (#1174) Composition knobs die per northstar: codegen and handwire are the only sanctioned composition paths, and BB_SYSTEM_ROUTES_AUTOREGISTER was the last of the six declared *_AUTOREGI... push web-flow github
96.77
30677787772 main refactor: remove BB_MDNS_CACHE_AUTOREGISTER composition knob (B1-1329) (#1173) Composition knobs die per northstar: composition is hand-wire or codegen only, never component self-registration. The symbol defaulted n and nothing in-tree set it, so... push web-flow github
96.77
See All Builds (2617)