Addresses Copilot review suggestion on #15720: when the handler aborts a
previously interrupted OTA session because a new upload arrived, also fire
the OTA_ABORT state notification so user-facing on_abort: automations and
other OTA state listeners observe the teardown. Without this, the abort
would be silent to listeners (distinct from OTA_ERROR, which signals a
failed session rather than a superseded one).
The previous fix tracked the AsyncWebServerRequest pointer to distinguish
web_server_idf's double index==0 callbacks (Start marker with data==nullptr,
then the first real data chunk) from a retry after an interrupted upload.
That is unreliable: AsyncWebServerRequest is a stack-allocated object in the
httpd task, so a new request that happens to reuse the same stack address
as an interrupted one compares equal and silently skips the abort. Closing
a browser tab mid-upload and starting a fresh upload from another window
could then concatenate partial data from the first upload with the new
image.
Gate the init block on 'index == 0 && len > 0' instead. This uniquely
identifies the first real byte of an upload on both IDF (start-marker has
len==0) and Arduino, no identity tracking needed.
web_server_idf invokes handleUpload twice with index==0 at the start of every
upload (once as a 'Start' signal with nullptr data, then with the first data
chunk). The previous fix treated the second call as a stale-session retry,
wastefully aborting and re-initializing the backend on every single upload.
Store the AsyncWebServerRequest pointer that owns the current session and only
tear down the backend when a different request arrives at index==0. Clear the
tracked pointer on every backend reset path so a stack-reused request pointer
from a later upload cannot collide.
If a captive_portal or web_server OTA upload is interrupted mid-stream (e.g. TCP
connection reset), the shared OTARequestHandler's ota_backend_ is left open. When
the browser resends the multipart POST, the guard 'index == 0 && !ota_backend_'
skipped re-initialization, so new bytes were written at the previous session's
offset. The Updater's end() then reports success with a concatenated image in
flash, bricking the device on reboot.
Abort and reset any in-progress backend whenever a new multipart upload starts
(index == 0) so the fresh upload begins a clean OTA session.
Mark SubscribeLogsResponse with (speed_optimized) = true so its
encode() and calculate_size() are compiled with -O2. Log messages
are among the most frequent messages sent over the API, especially
during debug sessions.
Add __attribute__((optimize("O2"))) to the main loop functions
(loop_task on ESP32, codegen loop() on other platforms) so GCC
inlines scheduler helpers and loop bookkeeping more aggressively.
Under -Os, GCC outlines small functions (Scheduler::call helpers,
millis conversions, etc.) that are called every loop iteration.
With -O2, these get inlined into the loop body, reducing call
overhead on the hottest code path in the firmware.
ESP32: loop_task grows from 303 to 416 bytes (+113 bytes).
ESP8266: no change (already fully inlined via ESPHOME_ALWAYS_INLINE).
Add __attribute__((optimize("O2"))) to the main loop functions
(loop_task on ESP32, codegen loop() on other platforms) so GCC
inlines scheduler helpers and loop bookkeeping more aggressively.
Under -Os, GCC outlines small functions (Scheduler::call helpers,
millis conversions, etc.) that are called every loop iteration.
With -O2, these get inlined into the loop body, reducing call
overhead on the hottest code path in the firmware.
ESP32: loop_task grows from 303 to 416 bytes (+113 bytes).
ESP8266: no change (already fully inlined via ESPHOME_ALWAYS_INLINE).
Add a new (speed_optimized) message option that emits
__attribute__((optimize("O2"))) on the generated encode() and
calculate_size() methods. Under -Os, GCC does not inline the small
ProtoEncode helpers (write_raw_byte, encode_varint, etc.) into the
generated methods, causing significant overhead on hot paths.
Apply to SensorStateResponse and BluetoothLERawAdvertisementsResponse
which are the highest-frequency encode paths.
Add a new (speed_optimized) message option that emits
__attribute__((optimize("O2"))) on the generated encode() and
calculate_size() methods. Under -Os, GCC does not inline the small
ProtoEncode helpers (write_raw_byte, encode_varint, etc.) into the
generated methods, causing significant overhead on hot paths.
Apply to SensorStateResponse and BluetoothLERawAdvertisementsResponse
which are the highest-frequency encode paths.
CodSpeed results show only Noise benchmarks improved with -O2 —
the speedup comes from libsodium's crypto primitives (Curve25519,
ChaCha20, Poly1305), not noise-c's protocol layer. Narrow the
optimization to libsodium only.
J. Nick Koston