Y2Linux 1.0 Beta 8.17 (Baseline 114)
Status: PUBLISHED as GitHub release baseline-114 (2026-10-21). Release 1.0.0-beta.8.17, Reborn 1.0.0-beta.8.17, kernel 6.18.56-y2linux-1.0.0-beta.8.17, update sequence 8100, Linux 8719042, Reborn 6d5458c. Built with tools/production/release.py (135 s, resumed workspace, update based on baseline-113): container suite 857 OK, host suite 884 OK, Reborn workspace tests, Clippy -D warnings and rustfmt clean; all seal checks PASS, no MediaTek firmware in any image. Not installed on a player before publishing; the owner updates by OTA. Milestone audit BETA-8.17 in roadmap-gap-audit.
| Image | SHA-256 |
|---|---|
| BOOTIMG | 722d7aed7348482faf0b3157418f008500bdd0cae35583ea784ef2d5eaf32214 |
| Y2ROOT | 4e2d06d5d01341edf48d73453cc94ee71ac59f120866975bff14290b90d4ffd3 |
| RECOVERY | 4d81a13a9f0b1a6784c31b893de7a21ebe4e658690aff6f88d81bbb642a1d1f1 |
| Y2DATA (no SSH key) | b02a224358a429addba6cb50117edcd8d43675be5b94daa8bd55737c56f9da17 |
Sources
Issue #51 (with a report), nine uploaded reports of 2026-10-20/21 (8.10–8.16) and four Reddit posts without reports. Every finding was checked against the 8.16 source before a change; nothing 8.13–8.16 had already fixed was changed again. Ledger: UserReports/REVIEWED.md (F12–F22, maintainer only).
What changed since Beta 8.16
| Finding | Evidence | Change |
|---|---|---|
#51: headphones silent after pause, lock and Next; ALSA -110 on every Play until a reboot | #51 report (install 2813306c, the #48 player, 8.16): 12 failed opens, both crystal tries XTAL_READY -110 each time, also after runtime rail cycles | Stock starts the CS43131 crystal with the codec and never stops it between streams (REA on the stock kernel); the upstream driver stopped it at every hw_free and restarted it at the next hw_params. Now a started crystal runs until the codec runtime-suspends; a start forces its writes to the chip, clears a stale ready flag, logs the live clock state when it fails, and cycles the codec rails once before an open fails. y2_hp/xtal counts failures and recoveries (kernel overlay 0030) |
| F12: SD card "needs repair, use a PC" though the PC reads it | c75aae95 (8.16, 7 791-song card refused at every insertion), Reddit (fixed by reformatting) | fsck.fat -n findings Windows mounts over are admitted: stale free count, label mismatch or invalid/lowercase label, unused clusters, backup boot sector differences, with or without the dirty bit. Each check records its classes and time; a slow check is a timeout, not "needs repair"; the last 10 admissions are in reports. Reborn says "needs repair" or "took too long" instead of "Reinsert the SD card" |
| F20: "Couldn't complete that" on every button until a restart | 61caa81e/297f1032, 319dab53, 26db176e (8.12) + Reddit | A disconnect seen in one of two queued Bluetooth snapshots was overwritten by the next; output stayed on Bluetooth. Kept now, and a Play refused for a missing peer pauses with "Bluetooth disconnected" |
| F21: Bluetooth search fails the first presses | 685c003a, 09cabe08, dbe84d46 | StartDiscovery 15 s; InProgress and a late NoReply count as scanning |
| F14: Connect refused while auto-reconnect pages a switched-off headset | 70cce0f1, 319dab53, 69635cdc | Reborn's intents went to /run/y2/bt-control.ini, the reconnect owner reads /run/y2/bt-intent/ (since Oct 5): never seen. Fixed; a user Connect cancels the automatic page (lock free in ≈ 0.2 s on the test bus), Connect retries a busy lock or a busy BlueZ answer for 45 s; a failed Connect, pending or codec intent holds auto-reconnect only until its 30 s deadline |
| F22: SBC soundbar stutters | Reddit (no report) | BlueALSA steps the SBC bitpool 53 → 45 → 35 when packets queue and back after 10 s clear (patch 0003, --sbc-abr, on by default); bitpool and steps in reports |
F15: pause_failed on every Bluetooth pause | 70cce0f1, d8152ea6 | BlueALSA's drain timeout that leaves the PCM ready counts as a pause |
| F16: a failed update check delayed the next by a day | ef22a23d (8.15, never offered 8.16) | Only a successful check's time counts |
| F5: an earlier release's panic reported again after an update | 19a41489, d48362a7 | Recorded, not reported, when its version is not the running one (versions.json) |
| F19: own radio stations (UK m3u) don't play | Reddit; no report shows a My Stations play | Mostly BBC, HLS/DASH only (not supported; planned for a later beta). Playlists served as text or bytes are recognised; station_failed logs reason, HTTP status and content type |
| F13: Wi-Fi does not find the saved network after a wake | 70cce0f1 (dbe84d46, 3 boots) | Cause unproven; wifi-failure.json now records scan counts, the saved network's signal and channel, and the country (no names) |
| F18: hang right after a wake from a sleep requested before Reborn was ready | 4fed1b6a (8.12) | Sleep waits up to 60 s for application_ready early in a boot (Power cancels); journal rows around the screen restore |
| F17: GC9503V wake stays dark (DSI reads lag after a screen-off) | 51f3b06d (8.12) | After 3 power cycles still lag, every DSI register is read twice (the lag is one read deep); Reborn's next frame after a failed wake frame tries one screen off/on |
Known gaps
- Not run on a player before publishing. Owed on the owner's player: #51 (pause 3 min, lock, Next, 50 quick Nexts,
y2_hp/xtal), codec rails idle after stop, sleep drain with the crystal running while the codec is active, a Connect while auto-reconnect pages, SBC steps at range, the F17/F18 paths (rare). - The rail cycle and the DSI double read are untested on hardware; both only run on an already failing path.
- HLS radio (BBC) and an on-device SD repair: later betas (owner decision 2026-10-21).
- Unchanged: #42, #45, #47 confirmation, WAF rule and billing alert, USB DAC check, video not for 1.0.