Y2Linux 1.0 Beta 3 (Baseline 52)

Status: PUBLISHED as GitHub release baseline-52 (2026-10-15). Release 1.0.0-beta.3, Reborn 1.0.0-beta.3, kernel 6.18.56-y2linux-1.0.0-beta.3, update sequence 5200, Linux 2731169, Reborn 8cd4362, git tag baseline-52 in both repositories. Built with tools/production/release.py (--public): container suite 739 OK, all seal checks PASS, no MediaTek firmware in any image. Milestone audit: BETA-3 in roadmap-gap-audit. First release of the new pipeline: build and all checks 119 s (build 76 s incl. BlueALSA and Reborn, checks in parallel 27 s, signing/firmware scan/assets 15 s), 158 s with the clean-up scan.

Image	SHA-256
BOOTIMG	`ca90dda51376b0fa58da5d6950936e3faf75eccf855e0725a08e98cc05acaebd`
Y2ROOT	`1e71564d7e1e80e81c293a6f5c0dc6e8eb4cb716479c2d9c7c95fb114e01e221`
RECOVERY	`20c665edab509fb26a164ca7480e9fb9f497c664b22f29fbe73bc146420b142d`
Y2DATA (no SSH key)	`01a73f767a8883d1327c43d4cc0d05c0ea5b762092bdaca6c51b0771a091b078`

Evidence behind the changes (Beta 2 player, 2026-10-15, read-only)

AirPods Pro connected, AAC 44.1 kHz. BlueZ's cached endpoint c0 01 8c 83 e8 00: the AirPods accept AAC **VBR, up to
256 kbit/s; Beta 2 sent BlueALSA's default 220 kbit/s CBR** without the FDK afterburner (60-s HCI capture: 2588 packets of 651 bytes ≈ 224 kbit/s).
Link at close range, Wi-Fi associated and idle: median acknowledgement 46 ms, worst 78 ms, ≤ 4 packets in flight,
no gap above 100 ms. Encoder load ≈9 % of one core at 598 MHz (Reborn ≈14 %).
BlueALSA keeps a three-packet socket buffer and blocks while the radio cannot deliver; only LDAC adapts its rate
(ABR). A weak link therefore starved AAC headphones at the full rate.
Coexistence: our WMT start sends the stock shared-antenna setting (01 10 02 00 01 01, coex_wmt_ant_mode=1 from the
owner's WMT_SOC.cfg), identical to the stock sequence for this chip; nothing missing was found there.
Page timeout: no page timeout … slots line appeared on Beta 2, so the controller already used 0x2000 (5.12 s); the
~18 s "Host is down" of Beta 1 comes from BlueZ's own sequence, and Beta 2's 45-s reconnect deadline is the fix.

Since Beta 2

Change	Detail
AAC quality	`--aac-afterburner`, `--aac-bitrate=320000`, capped by the peer (AirPods 256 kbit/s) and by one packet per frame (`y2_aac_mtu_bitrate`: a 672-byte MTU keeps the tested 220 kbit/s; headsets that cannot reassemble fragmented frames keep working). CBR, not VBR (Android's default for compatibility). Setting `aac_quality` high (default) / standard.
AAC adaptive rate	BlueALSA patch `0002-y2-aac-adaptive-bitrate.patch`: per frame the socket queue is read (as for LDAC ABR); two queued packets or a blocked write step the CBR rate down 8/8 → 6/8 → 5/8 → 4/8 (not below 128 kbit/s, at most every ~0.2 s), ~10 s of an empty queue steps it back up. FDK re-initialises its configuration between frames, no encoder restart. Decisions in `y2-aac-abr.h`, host-tested (`tests/test_bluetooth_aac_abr.py`). Setting `aac_abr` on (default) / off.
Observation	`EncoderStats` now covers AAC; platform status reports `bitrate_bps`, `aac_abr_enabled`, `aac_abr_level`, `aac_abr_adjustments`.
SBC XQ	XQ / XQ+ selectable without the experimental switch (default stays high).
Reborn	Service menu rows AAC Quality and AAC Auto Rate; SBC XQ codec choice no longer tied to the experimental switch.
Pipeline	First release through `release.py` (one workspace, parallel checks, unchanged sources linked, prune).

Device checks owed (after the owner's over-the-air install, read-only)

uname -r = 6.18.56-y2linux-1.0.0-beta.3; with the AirPods playing: y2-platform status bluetooth → encoder.bitrate_bps 256000 (or the MTU-capped value), aac_abr_enabled true, aac_abr_adjustments 0 at close range; bluealsad CPU (was ≈9 %); a listening comparison against Standard; a walk-away test watching aac_abr_level and audible dropouts.

Known gaps

Battery in sleep ≈9.4 mA vs stock ≈1.3 mA; the AAC changes are not yet heard or range-tested on the device; no headphone jack detection; rare slow display start after a restart; full install from stock run on one player only.