Alan-Jowett / sonde / 24690448068

82%

push

github

fix(pair): skip enforce_lesc for node connections (ND-0904) (#766)

* fix(pair): call createBond() before GATT connect latch to ensure LESC Numeric Comparison on Android

The modem calls `ble_gap_security_initiate()` immediately in its
`on_connect` callback, sending a BLE Security Request to Android.
The previous code called `createBond()` only after waiting for the
GATT connection to complete. By that point the Security Request had
already arrived, and Android — not yet in bonding mode — handled the
incoming SMP as a background pairing, advertising NoInputNoOutput IO
capabilities. This forced Just Works (PAIRING_VARIANT_CONSENT) instead
of LESC Numeric Comparison, causing `enforce_lesc()` to reject the
connection (PT-0904).

Fix: move `createBond()` to immediately after `connectGatt()` returns,
before waiting for the connect latch. `connectGatt()` is asynchronous —
the LE link is not yet established when it returns — so `createBond()`
races ahead of the modem's Security Request. Android enters bonding mode
with KeyboardDisplay IO capabilities before the Security Request arrives,
ensuring LESC Numeric Comparison is negotiated.

Windows was unaffected because `BtleplugTransport::pairing_method()`
returns `None`, which `enforce_lesc()` treats as OS-enforced LESC,
bypassing the check.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(pair): preserve lastError when setting up bonding to avoid masking GATT connection errors

When createBond() is called in step 2 (now before waiting for the GATT
connection latch), the onConnectionStateChange() callback may race ahead
and set lastError with a GATT status code. Resetting lastError = null in
the bonding setup block would lose this error, causing step 3 to report
the generic 'connection failed' instead of the actual GATT status.

Fix: do not reset lastError in the bonding setup block. The error will
only be cleared later at step 5, after the GATT connection has completed
successfully.... (continued)

5 of 5 new or added lines in 2 files covered. (100.0%)

24456 of 30264 relevant lines covered (80.81%)

203.57 hits per line