Y2Linux 1.0 Beta 8.7 (Baseline 95)

Status: PUBLISHED as GitHub release baseline-95 (2026-10-18). Release 1.0.0-beta.8.7, Reborn 1.0.0-beta.8.7 (identity only), kernel 6.18.56-y2linux-1.0.0-beta.8.7, update sequence 6900, Linux f7b4fb5, Reborn 82fe089. Built with tools/production/release.py (151 s): container suite 778 OK, host suite 805 OK; all seal checks PASS, no MediaTek firmware in any image. Milestone audit BETA-8.7 in roadmap-gap-audit.

Image	SHA-256
BOOTIMG	`81130491bf75928c06579e6820da1dd9b0298f83b09a858af601bacb9f787ce6`
Y2ROOT	`1543fbcfbfcbe8f228c0c330d75688af1c0c4ce8ed54ae6ec644f5a44ab6402d`
RECOVERY	`6a646b689d1d643c67e71fb8558d363c7124cbbed4a883e7f2c16478f71fa874`
Y2DATA (no SSH key)	`7b9c5c7b907142726468c5fb5f50ee9a8c67a18ffb828cb180ac641c0a12eb93`

What changed

Part Change

Panel identification (#45, #46) REA of the stock LK (v3.2.0 lk.bin, bb1a93b4…): auto-detect 0x81e0af60 walks [gc9503v, st7701] (0x81e3821c) and names the first driver whose lcm_compare_id matched, else the last one (st7701); the lcm= digit is a separate DCS 0x0A read. The GC9503V compare (0x81e1eb60) only matches DCS 0x04 = xx 95 03/04, so lcm=0-gc9503v… is a positive identification. kernel/platform/boot-policy.h trusted only lcm=1-, so the #45 reporter's GC9503V (lcm=0- on a cold boot, DSI read lag at the ID read) was never identified. Now a gc9503v name counts with either digit; st7701 still needs lcm=1-. Linux 7046833, tests test_power_policy (mutation caught) and test_power_final_27.

Behaviour ST7701S players: unchanged. GC9503V players: identified and recorded; the power-down stays opt-in (CONFIRMED_PANELS) until a GC9503V owner confirms it.

Part	Change
Panel identification (#45, #46)	REA of the stock LK (v3.2.0 `lk.bin`, `bb1a93b4…`): auto-detect 0x81e0af60 walks `[gc9503v, st7701]` (0x81e3821c) and names the first driver whose `lcm_compare_id` matched, else the last one (`st7701`); the `lcm=` digit is a separate DCS 0x0A read. The GC9503V compare (0x81e1eb60) only matches DCS 0x04 = xx 95 03/04, so `lcm=0-gc9503v…` is a positive identification. `kernel/platform/boot-policy.h` trusted only `lcm=1-`, so the #45 reporter's GC9503V (`lcm=0-` on a cold boot, DSI read lag at the ID read) was never identified. Now a `gc9503v` name counts with either digit; `st7701` still needs `lcm=1-`. Linux `7046833`, tests `test_power_policy` (mutation caught) and `test_power_final_27`.
Behaviour	ST7701S players: unchanged. GC9503V players: identified and recorded; the power-down stays opt-in (`CONFIRMED_PANELS`) until a GC9503V owner confirms it.

Device results

None on a GC9503V yet (the reporter's test is requested on #45). The owner's ST7701S player installs it by OTA; its panel path does not change.

Known gaps

The reporter's LK comes from the 3.2.1 updater; its hash is requested on #45/#46 to confirm the same driver order.
Not done on purpose: skipping the power-down on boots with DSI register-read lag (the reporter's 7.10 run showed the
lag clears after the first screen-off).
Unchanged from 8.6: battery current with a profile, output measurements, sleep drain (#42).