POWER-FINAL-49 — firmware-free public images, one-time flash package, public codec set

Status: PUBLISHED as GitHub release baseline-49 and installed over Wi-Fi (2026-10-16): the update files, a first-install package for SP Flash Tool and the GPL source collection. Linux b4610ea, Reborn ceb6313, kernel 6.18.56-y2linux-platform-v1-power-final-49, release …candidate.49, update sequence 4900. Container suite 727 OK, all seal checks PASS (public distribution gate included). Packages out/y2linux-platform-v1-power-final-candidate-49/ (update; -ota-release/ signed with platform-v1-release-20261015) and out/y2linux-49-first-install/ (public flash package), receipts out/deep-sleep-final/build49-receipts/. Owner-private variant with aptX: …-candidate-49-private/ (-ota/), receipts build49p-receipts/, not published.

Image	SHA-256
BOOTIMG	`c7ece1080fe5dd76f01b02d671dfbe860e5339901b9526f291b44e9f2a27f1a5`
Y2ROOT (public)	`f1a073c8014163db53781cb3a339bd92acf09d5d17b90c15610cc074760c0d97`
Y2ROOT (private, aptX)	`93405c27600a51a26430822fb1ea52b2c7e03f6f7c5febc3a42026e8bfa48c23`
RECOVERY	`7097cce59b90c55196fa581db03dd2d1538a37e48fcf3bc92a69c1b8e3478418`
Y2DATA (public, no SSH key)	`1b741e60dd08d4596bcb89294404ed6c7b8341734aabac0b95a95a01f9c8a262`

Why

The owner asked for release 49 to carry all images (the one-time flash setup for other developers) and an explanation of the recovery. Preparing it showed that the system image of releases 46–48 (rootfs.ext4.gz) contained MediaTek's radio firmware, which may not be redistributed (owner: permission from Innioasis, not from MediaTek). Those assets were removed (notes added; never in Git). The public-distribution gate also blocked AAC, aptX, aptX HD and LDAC.

Change	Detail
No firmware in any image	`/lib/firmware/mediatek/mt6582` → `/data/firmware/mt6582`. Seal check refuses firmware files; the final images were searched for the firmware's bytes: none.
Existing players	The 49+ recovery copies the reviewed files from the old root to Y2DATA (hash-checked) before it writes the new root.
New players	`y2_platform.radiofw` reads the owner's stock `system.img` (Android sparse ext4) itself, checks the five files (+ reader library) by SHA-256, derives the two defaults, installs them, deletes the copy and starts the radios. While missing, MTP shows a Firmware drive (one empty Y2DATA folder; responder allowlist extended); Reborn explains on Wi-Fi/Bluetooth. `y2-platform firmware status
Codecs	Owner decision 2026-10-16: public SBC, AAC from fdk-aac-free 2.0.3 (AAC-LC only, as Fedora ships), LDAC; aptX/aptX HD only private (`build.py --public`).
Public package	`package.py --public`: no SSH key (developers personalize Y2DATA with `first_owner.py`), no local paths, one scatter writing BOOTIMG, RECOVERY, ANDROID, USRDATA.
GPL sources	Buildroot legal-info (package sources, patches, licences; Reborn = its tagged repository) and the pinned Linux archive are release assets.
Build hygiene	Every build installs into an empty target (a reused workspace had kept aptX and the full FDK in an image); two licence-file hashes of bumped packages recorded.
Contributions	Y2Linux #38 (cs43130 on-demand load measurement, two driver fixes), Y2Reborn #6 (artwork worker), #7 (A–Z sort), #8 (genre grid title).
Docs	INSTALL (eight-step release install), Recovery and updates, BUILDING.

Device results

Installed over Wi-Fi from the release (2026-10-16, archive 20261016/153, boot 1723225f): PASSED. The
recovery logged radio firmware kept on Y2DATA (seven files, reviewed hashes) before writing the new root; install 54 s; recovery screen lit (vblank 16 at 1.28 s; the DSI read lag appeared again and the pipeline ran, as on 47/48); Acknowledged, taint 0. /lib/firmware/mediatek/mt6582 → /data/firmware/mt6582, firmware status Ready. Radios from the new location: MD1 calibration boot result 0, both WMT patches applied, Wi-Fi Online, Bluetooth powered; codec inventory public (SBC, AAC, LDAC), only libfdk-aac.so.2.0.3, no aptX.
Found on the way: after the cancelled private 49, Reborn kept showing Update Didn't Finish and never offered the
release a later check had found (journal Failed: owner cancelled won over everything). Fixed in Reborn 476e281 (a cancel is no failure; a later check supersedes a failure), ships with the next release. Owner unblocked by staging the release over the player's Wi-Fi via SSH.
Firmware drive (new-player path), owner test (archive 20261016/154): works, two bugs found. Firmware backed up
and removed at runtime → Missing; the Firmware MTP drive appeared after ~5 s (README only). The owner copied the stock system.img (518,590,232 bytes); the installation took ~2 s on the player, all seven hashes identical to the original, the copy deleted, space freed, the drive removed. Bugs: (1) the watcher had seen the firmware at service start and never re-checked, so the copy was taken only by a manual firmware scan (a fresh install is not affected); (2) the combined connectivity restart was cut off by a 10 s limit after its stop half, so the radios stayed off until started by hand (then Wi-Fi Online, Bluetooth powered). Both fixed in Linux d1f1195 (next release). The owner's file manager moved instead of copied: system.img restored from the Innioasis zip (hash verified); INSTALL and the drive's README now say copy, don't move.
Still to do: AAC playback with the owner's headphones.