When the check itself fails, the only option was Retry. Add two more:
Install proceeds as if stock flashable firmware had been detected (straight
to the flash page), and Skip proceeds as if compatible custom firmware had
been detected — routed through the same classification path, so the
phone-side G2 font extraction still runs when the fonts are missing. Both
appear only in the error phase; Back and Retry keep their places.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
Another developer reports Samsung devices truncating inbound notifications
larger than 64 bytes, and the missing-settings-reply issue indeed reproduces
only on the Samsung tablet. Every frame the tablet DID deliver was <=55 bytes;
a reply would only trip a 64-byte limit if it were bigger -- e.g. a fuller
first-reply snapshot, or a reply coalesced into one notification value with
one of the lens's own sid-0x80 notifies (11+55 = 66 bytes).
The coalescing case also exposed a latent app-side bug: parseFrame only ever
read the FIRST envelope in a notification value, so a second frame packed into
the same value was silently dropped even on healthy devices.
- BleProtocol.splitFrames(): split a notification value into its individual
aa21/aa12 envelope frames (self-describing via the length byte), returning
a trailing partial frame as-is so callers can log stack truncation.
- Device-info probe, flash-prompt communicator, and flasher control channel
now process every frame in a value. The probe additionally logs multi-frame
values (with the raw value size) and marks frames whose actual length falls
short of the declared length as TRUNCATED, so the next tablet run can
confirm whether the stack truncates or drops oversized values.
The existing first-request retries remain the workaround for replies the
Samsung stack drops outright. Known residual exposure, not addressed here:
the custom firmware's settings reply carries the capability string and is
well over 64 bytes in a single frame, so on affected Samsung devices the
main communicator would never see CFW capabilities; the durable fix is
firmware-side (chunk lens-to-phone notifications at <=64 bytes), since an
app cannot recover bytes the stack never delivers.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
The device-info frame trace settled the remaining firmware-check mystery: the
right lens is the correct settings endpoint, but on stock 2.2.9 a request
written immediately after the prelude ack is sometimes silently dropped while
the lens emits its own sid-0x80 notifications (flag 0x01, cmd 4, device-chosen
magic, result {f1:1}, ~2 s apart). The probe's retry answers within ~100 ms
every time, so the check now passes; the left-lens fallback never fired.
The flash-prompt page-create goes out at exactly the same post-prelude moment,
so give it the same single retry. Finding documented in
../notes/ble-connections-2.2.9.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
On stock 2.2.9 the settings read on the right lens goes unanswered even after
a successful security-auth exchange and fresh pairing. Since no 2.2.9-era
capture of a settings read exists (g2flash only exercises auth + OTA), turn
the probe into an experiment that a single on-device run can settle:
- Log every control-channel frame (sid/flag/type/magic/frag/payload prefix)
and every write under the FaceclawDeviceInfo logcat tag.
- Retry the settings read once before concluding a lens is silent.
- Accept a sid-0x09 frame carrying firmware versions on the device's own
magic (an unsolicited push) as an answer, not just the awaited ack.
- If the right lens stays silent, run the same connect/auth/prelude/read
against the LEFT lens and use its answer, logging which lens replied, to
test whether 2.2.9 moved the settings endpoint.
The probe now takes both arm addresses; the firmware-check page passes them.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
Firmware 2.2.9 requires the sid-0x80 DEVICE_SETTINGS/AUTHENTICATION(4)
exchange shortly after each GATT connection: it gates query responses on
completing it over an encrypted link, and closes unauthenticated links after
~30 s. On a phone with no existing bond (no inherited stock-app pairing) the
exchange is also what triggers SMP pairing. This is why the onboarding
firmware check failed with "no response to the device-info query" on a fresh
device: the probe connected, the prelude was acked, but the settings query was
never answered and no pairing ever happened. (See
../notes/ble-connections-2.2.9.md, validated by g2flash on hardware.)
- BleProtocol: buildAuthenticationRequest / isAuthenticationSuccess (strict:
command 4, matching magic, empty result message), SID_SECURITY_AUTH.
- Device-info probe: authenticates before the prelude with a 30 s window (the
user may need to accept an OS pairing prompt; the check page says so). Soft
on non-response since the probe also runs against the custom firmware —
but a query timeout after unconfirmed auth now names the pairing problem.
- Flash-prompt communicator: authenticates each arm before the prelude
(mandatory — it only ever talks to stock firmware).
- Firmware flasher: control-channel notifications are now mandatory, each
lens authenticates before OTA BEGIN, and the OTA-time heartbeat is gone —
official captures show no control traffic during a transfer, heartbeats
demonstrably do not satisfy the 2.2.9 deadline, and a concurrent control
writer interleaves with the marker/data transactions. The OTA envelope
counter now resets after bring-up, mirroring the stock sequence.
- Main communicator: authenticates both arms after connecting, before the
prelude, with a short soft wait (logs and continues on timeout so firmware
that never answers cannot stall reconnects).
Hardware test pending: fresh-device onboarding against stock 2.2.9 (check,
prompt, flash), and a regression pass on custom-firmware connect + probe.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
Correctness:
- Pair-glasses flows no longer strand the app in the manual-disconnected
state: the pre-scan disconnect is followed by resumeAutoReconnect(), and
the config page's scan hand-off drops the live link too.
- The mixed-arms serial warning is reachable now: pairs() groups by the
canonical serial key, and a lone left + lone right with different serials
get the mismatch warning on both rows (it previously required two
different serials inside one serial-keyed group, which cannot happen).
- A selected pair that degrades to one arm (prune while the page is open)
is deselected, and Save never pads a missing arm from stored addresses.
- DiscoveryAggregator.ingest no longer lets a bonded re-emit or split
advertisement overwrite the note/name/flag learned from a full
advertisement, and reclassifies split reports using remembered fields so
a nameless half doesn't drop the arm.
- estimateProximity ignores the advertised TX Power Level (radiated power,
not measured-power-at-1m); the calibration table is the only reference.
- Saving a ring without a glasses pair updates the stored identity's ring
name/address instead of leaving the previous ring on record.
- Ring detection requires the stock "EVEN R1" name prefix in both the TS
classifier and the Java pre-filter; a bare "R1" substring matched
unrelated devices, and ring rows are always selectable.
- One-shot scans keep every sample (buildAddressSet merges them through
DiscoveryAggregator), and flash-flow address resolution takes left/right
from one serial-joined complete pair instead of latest-per-role.
Lifecycle:
- The pairing scan and its refresh timer pause on app suspend and resume
with the app, and the model's app-event hooks are detached when the page
is left for good.
Cleanup:
- One MAC normalizer (even-advertisement.normalizeMacAddress, re-exported
by device-addresses), used for the ring "Yours" comparison too.
- One serial-equality definition (serialKey/serialsMatch); the containment
fallback in isPreviouslyPaired is gone.
- Shared formatErrorMessage replaces five per-view-model copies; shared
hex-util replaces the duplicate TS hex codecs; one Java bytesToHex in
FaceclawFirmwareUtil replaces three hand-rolled encoders.
- glassesRow/ringRow share a presentation-base helper; dead row address
fields, the constant glassesSectionVisibility, the identical-branch
status ternary, and zoneRank/ZONE_ORDER are removed.
- .gitignore now actually covers .test-build/ as tests/README claims.
UI:
- Dark-mode styles for the pairing page: readable instructions, dark-gray
cards with consistent light text; light mode keeps its current look.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016ZqzwmRQ3HQxasBz5rYyZV
Removes unused keyword-spotting model. Makes on-device voice model a
runtime download rather than bundling it. Removes native-code library
versions for 32-bit and x86_64 processors.
Stream BLE advertisements into a dedicated pairing flow with arm, ring, serial, and hardware variant detection.
Add proximity ranking, device artwork, pairing safeguards, and unit coverage for discovery and candidate grouping.
The flasher was validating that there are exactly 5 firmware segments
(which was true of 2.2.4 but not 2.2.6), causing it to reject the stock
firmware in the uninstall flow.
(This doesn't actually prevent uninstalling as you can still use the
stock Android app's OTA updater.)