Repository navigation
Conversation
parse() accepted a fixed32/fixed64 field that ran past the end of the buffer, and fmt() then crashed in struct.unpack. A payload in the Gen 1 capture CAP-081 hit this, so the decoder could not list that log at all. A truncated fixed field now makes parse() return None (not a message), like a truncated length-delimited field already did. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
redact_btsnoop.py prepares a contributor's btsnoop log for publication (CONTRIBUTING.md, "the PII exception") without changing any frame number: other devices' ACL data and advertising reports become an empty vendor event, addresses become placeholders, link keys are zeroed, and given strings (serials, names) are replaced, also across ACL fragments. It fails if anything it was told to remove is still present. decode_maestro_frames.py decodes text dumps of pw_hdlc frames (used by GEN1_SUPPORT_INVESTIGATION.md for PixelBudsMacOS's third-party frames). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…enControl A third-party capture (Group BQ) of a Gen 1 on release_5.11_signed, so the Gen 1 support can be judged on evidence (PROJECT.md: other models only when "demonstrably identical", verified, never assumed). Part 1, the official app on a Pixel 4 XL: every Gen 1 setting written. 22 of its Maestro WriteSetting frames equal this app's encoder output. The mode and multipoint go over signed Message Stream commands instead. Settings 27/28/29 answer status 2. Find left/right only. Part 2, OpenControl on a Pixel 9 Pro: 43 writes, 43 OK; the official app then read back the mode and multipoint this app set. Logs redacted with scripts/redact_btsnoop.py; films, bugreports and the unredacted logs stay local (hashes listed, ADR-066). PROTOCOL.md gets a section 0.1 row and section 4.6, all HYPOTHESIS, proposed for FACT; no promotion and no ADR here (AGENTS.md section 6). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Gen 1 owners can use the app instead of being held in Safe Mode, with only the writes CAP-081 shows the official app making the same way. - DeviceProfiles: per model the firmware, device type (GetHardwareInfo field 1), Model ID, read sweep and allowed write targets. SafeModeGate checks the model and the target. Gen 1 = release_5.11_signed and device type 4 read on the same connection; Find also needs Model ID 6edaf7. A Pro 2 firmware with a read device type other than 6 is now refused. - Gen 1 writes: settings 2, 4, 7, 11, 12 (three booleans), 15, 16, 17, 19, 22, the mode 1-3, ring left/right. The mode and multipoint carry TODO(verify): the official app uses signed Message Stream commands. - Press and hold is written one bud at a time, as the official app does. - Gen 1 UI: no Adaptive, head gestures, case sounds or Ring both; the Info tab shows the model. Debug: "Read Fast Pair model ID". - Tests: Gen1OfficialAppWritesTest compares 22 official-app frames of CAP-081 with the encoders byte for byte; the Gen 1 fixtures are CAP-081 frames (serials redacted, CRC re-sealed). Depends on the draft ADRs in the pull request; none is accepted yet. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GEN1_SUPPORT_INVESTIGATION.md is the 2026-11-01 analysis behind the change, with a status note on what CAP-081 superseded and its external references turned into links. CHANGELOG [Unreleased], TODO.md section 5 (the open Gen 1 items) and the session log 0102 with the three draft ADRs for the maintainer. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The below is AI generated, and should adhere to the contribution guidelines for this project. I've tested the change locally quite extensively with my own Pixel Bud Pro 1, and everything seems to work, and the payloads are essentially identical to those in the official Pixel Buds app.
Thank you for making this project, by the way. I was very happy to see this as an alternative to the official app, as I'm trying to de-google my life a bit.
Adds support for the original Pixel Buds Pro (Gen 1), firmware
release_5.11_signed, next to the Pro 2. A Gen 1 is identified by its firmware and itsGetHardwareInfodevice type (4); everything unknown stays in Safe Mode, and the Pro 2's behaviour is unchanged (plus one extra check, below).This is a third-party contribution and it needs your decision:
PROJECT.mdlists other models as a non-goal "unless the protocol is demonstrablyidentical — this must be separately verified". The evidence for that is a new capture,
CAP-081. The three draft ADRs are below and inai-sessions/0102_FEATURE_RESULT_2026_11_02.md. They are not inDECISIONS.md, and nothing inPROTOCOL.mdis promoted (AGENTS.md§6).Evidence —
CAP-081(Group BQ, my own Gen 1, redacted)Part 1, the official Pixel Buds app (1.0.996096625, Pixel 4 XL, Android 13): every Gen 1 setting changed and changed back.
WriteSettingframes are byte for byte this app's encoder output, CRC included. They are the fixtures of the newGen1OfficialAppWritesTest.Part 2, this branch's build (Pixel 9 Pro, GrapheneOS, Android 17): the same actions.
RESPONSE OK, only on fields 2, 4, 7, 11, 12, 13, 15, 16, 17, 19 and 22.The Gen 1 has no settings 27, 28 or 29: the official app reads them and gets status 2.
Two settings the official app writes differently. For the noise-control mode and multipoint it sends signed Message Stream commands (
08 12,07 12, with 16 trailing bytes that need the account key). This branch usesWriteSetting 4:{13:n}(ADR-067) and4:{11:b}(ADR-053) instead.TODO(verify)inDeviceProfiles.kt. D-3 asks you to choose between accepting them and making them read-only.Redaction (new
scripts/redact_btsnoop.py; frame numbers are kept):The films, the bugreports and the unredacted logs stay with me; their hashes are in
CAP-081-sha256sums.txt(ADR-066).What changed in code
DeviceProfiles(new): a profile per model with its firmware, device type, Model ID, read sweep and allowed writes.SafeModeGatenow checks the model andthe write target.
scripts/pwrpc_decode.py: fixed a crash on a truncated fixed32/fixed64 field. It hit this onCAP-081.Draft ADRs (proposed, not accepted; numbers are yours to assign)
PROJECT.md's non-goal gets a dated note.release_5.11_signedand device type 4 read on the same connection. Find also needs Model ID6edaf7.until then.
Not in this PR (in
TODO.md§5)README.md,PROJECT.md,ARCHITECTURE.md§5a andRELEASING.mdare not updated. They depend on your decision on D-1.Checks
cd android && ./gradlew assembleDebug testDebugUnitTest test lint(plus the release compile, manifest and lint steps ofandroid.yml): green.INTERNETpermission in the source manifests or in any merged manifest.python3 scripts/lint_docs.pyin a full checkout: exit 0. Its output is identical tomain's..ktfiles carry the AGPL-3.0 header.git merge-tree): the code merges cleanly, andCAP-080/Group BP are left to 1.4.1: the fixes after CAP-079 — do not merge before 1.4.0 is released and CAP-080 is analysed #70.CAPTURE_BLUETOOTH_HCI_SNOOP.md,CHANGELOG.mdandid_registry.csvconflict only where both PRs append rows.🤖 Generated with Claude Code