fix: correct memcmp misplaced parentheses (size argument in _crcError CRC field detection)

[m-mcgowan]

Summary

• Fix misplaced parenthesis in _crcError where the memcmp return value is not checked correctly — != 0 was intended to test whether memcmp found a match, but instead it's parsed as part of the size argument, reducing a 7-byte comparison to 1 byte
• Add regression tests using a realistic hub.status response that triggers the false detection

The Bug

_crcError detects CRC fields by checking 22 bytes before the closing } for the pattern "crc":". The memcmp call was:

memcmp(..., CRC_FIELD_NAME_TEST, ((sizeof(CRC_FIELD_NAME_TEST) - 1)) != 0))

The != 0 was intended to test the memcmp return value, but the closing parentheses are misplaced — it's parsed as part of the third argument: (7 != 0) = 1. So memcmp only compared 1 byte — the opening " quote — which matches ANY JSON field boundary at that position.

When does this trigger?

Only when talking to a Notecard that does not support CRC. note-c always adds CRC to outgoing requests speculatively. When the Notecard supports CRC, it responds with a real ,"crc":"..." field at the expected offset, so the 1-byte compare accidentally matches the correct field and everything works.

When the Notecard does not support CRC, responses lack a CRC field. If the response happens to place a " character at the CRC detection offset (22 bytes before the closing }), the 1-byte compare falsely matches it.

For example, {"connected":false,"status":"connecting"} (41 bytes) — a real hub.status response during connection — has " at position 19 (= 41 - 22), the opening quote of "status". This is a realistic trigger because hub.status only returns connected (boolean) and status (string) with no other fields. This triggers false CRC detection, which:

1. Permanently sets notecardFirmwareSupportsCrc = true
2. Every subsequent response without a real CRC field is then flagged as a CRC error
3. Triggers retries until failure, effectively bricking communication

Fix

// Before (buggy — != 0 applied to size arg, evaluates to 1):
memcmp(..., CRC_FIELD_NAME_TEST, ((sizeof(CRC_FIELD_NAME_TEST) - 1)) != 0))
// After (fixed — != 0 applied to memcmp return value):
memcmp(..., CRC_FIELD_NAME_TEST, sizeof(CRC_FIELD_NAME_TEST) - 1) != 0)

Test plan

• Existing _crcError_test tests pass (CRC detection, validation, edge cases)
• New regression test: hub.status response does NOT trigger false CRC detection when firmware does not support CRC
• New regression test: notecardFirmwareSupportsCrc is not falsely set to true
• When firmware supports CRC, missing CRC is correctly flagged as error
• CI passes