Task: review-and-test track for birkskyum's upstream Snapdragon PRs (#8672, #129, #221–#223)

[JimmayVV]

Question

Keep birkskyum's five upstream PRs alive through review. They are omacom/omarchy#8672, omacom#129, and omacom/omarchy-pkgs#221, #222 and #223. This map can do two things nobody else can. Check the split against the map's decisions, and test each PR's artifacts on the HP EliteBook Ultra G1q so every PR carries a second-machine Testing line.

Standing items as of 2026-08-27.

• #222 drops the pkgbase shim. Flagged in review. Zesko !64 finds a kernel without pkgbase, but the hook then runs mkinitcpio --kernel <version> with no --kernelimage, and mkinitcpio searches only /lib/modules/<ver>/vmlinuz, /lib/modules/<ver>/vmlinux and /boot/vmlinuz-*. It never looks at /boot/Image. Wait for birkskyum's answer. If he keeps the shim, its retirement note stays as written in Task: omarchy-pkgs aarch64 boot packages — limine-entry-tool !63 patch + pkgbase shim #18. If he points mkinitcpio at /boot/Image instead, verify that a clean install builds a UKI without the shim.
• Encrypted-install gate. Task: encrypted installs on Snapdragon — keyboard driver and a visible LUKS prompt in the initramfs #25's finding, a dark panel and no keyboard driver at the LUKS prompt, went to him on #8672. It is his declared draft gate on Boot and install Omarchy on Snapdragon X ARM64 systems omacom/omarchy-iso#129.
• Rule for the track. Review and test. Do not grow the stack. Audio compat (Task: audio on the G1q — ADSP boot, topology name, UCM match, qcom-firmware-compat #26), USB-C root (Task: USB-C root disk drops when the DSPs boot with real firmware (UCSI port reset) — fix or keep the DSPs off on USB installs #27) and the encrypted path (Task: encrypted installs on Snapdragon — keyboard driver and a visible LUKS prompt in the initramfs #25) become follow-up PRs after these land.
• When the maintainers are silent, do nothing. That is the norm. A re-ping is not.

Resolves when the PRs merge, are closed, or birkskyum hands the track back.

[JimmayVV]

Re-review of birkskyum's PRs, 2026-08-28 00:30Z, about three hours after the endorsement.

He answered every point the same evening.

• The shim is back on #222. "Good catch. Confirmed: MR 64 discovers the module directory but mkinitcpio still cannot find Arch Linux ARM's /boot/Image." He restored the three linux-aarch64-pkgbase-shim commits with Jimmy's authorship, bumped it to 1-3, and added accept-shim-pkgbase.patch to limine-mkinitcpio-hook 1.37.1-4. That patch is Task: omarchy-pkgs aarch64 boot packages — limine-entry-tool !63 patch + pkgbase shim #18's 03-alarm-pkgbase-ownership.patch rebased onto !64, same pacman -Qqo "${pkgbase_file%/pkgbase}" check, credited to Jimmy in its header. A native ARM build of the corrected stack passed (run 33122253609).
• Boot and install Omarchy on Snapdragon X ARM64 systems omacom/omarchy-iso#129 gained four fixes. Copilot found that archiso-copy-boot-efi.patch was added but never applied in build-iso.sh, so Boot and install Omarchy on Snapdragon X ARM64 systems omacom/omarchy-iso#129's ISO had no live UKI; fixed in 5572e51 with a fail-fast in the builder and a CI check for omarchy-live.efi. amd-ucode removed from packages.aarch64. The shim added to the ISO's package list. archlinuxarm-keyring kept in the offline mirror with a build-time require_offline_package gate. That last one is the same fix as this fork's unpushed ab0b594, plus the gate; ab0b594 is now redundant.
• The encrypted gate passed on the Yoga. Full-disk LUKS2 install boots through Limine. The LUKS prompt is invisible because there is no DRM framebuffer before unlock, but blind passphrase entry works. He documents that as a separate early-display fix. Recorded on Task: encrypted installs on Snapdragon — keyboard driver and a visible LUKS prompt in the initramfs #25.
• Yoga speakers are silent too. He corrected his validation: microphones work, internal speakers produce no sound. Recorded on Task: audio on the G1q — ADSP boot, topology name, UCM match, qcom-firmware-compat #26.
• State. #8672, #221, #222, #223 are ready for review. Boot and install Omarchy on Snapdragon X ARM64 systems omacom/omarchy-iso#129 is still draft. No maintainer has commented anywhere.

Nothing for this ticket to do right now. The one standing item, the shim, is resolved.

[JimmayVV]

2026-08-28: Jimmy let the PRs take over. Coverage was checked commit by commit against the three snapdragon branches. Everything functional is in the five PRs; the omissions are the fork's own package pins, qcom-firmware-compat as a follow-up, and the diag branches. The fork's branches are frozen from here. This ticket's work is now only two things: test omacom#129's ISO on the G1q when one exists, and answer review questions. Ticket stays open until the PRs merge or close.

[JimmayVV]

Correction, 2026-08-28 01:40Z. The note above calling the fork's ab0b594 redundant was wrong, and it cost an install. birkskyum's keyring fix lives in omacom#129, not on this fork's ISO. This fork's post-install/pacman.sh still installs archlinuxarm-keyring from the offline mirror, so dropping the package from configs/packages.aarch64 made run 33130679877's ISO fail at "Configuring system" with target not found: archlinuxarm-keyring, after the disk was written. Exactly the failure his commit message describes. Restored as 3ae7074; run 33133495201 is the rebuild. Lesson: redundant with upstream is not redundant for the fork's own ISO while the fork still builds its own.

[JimmayVV]

G1q test of birkskyum's ISO, 2026-08-28 (HITL over Tailscale). ISO from his integration/encrypted-yoga branch at 5c3bce42, run 33128799299, artifact omarchy-iso-aarch64-5. Unencrypted install to the external NVMe on USB-A.

Verified: boots via Limine 12.6.1 into omarchy_linux-aarch64.efi; Snapdragon kernel parameters on the cmdline; zero failed units; both DSPs running; all five vendor blobs staged from Windows during install; GPU card0 and renderD128 with Adreno firmware loaded; Wi-Fi wcn7850 hw2.0 on ath12k, connected from the bar; keyboard, touchpad, touchscreen; ALARM package sources configured and pacman -Sy working out of the box.

Deep suspend and resume works. mem_sleep = [deep], entered 23:07:52 local, returned 23:08:52 on lid open, suspend_stats success 1 / fail 0, devices resumed in 0.7 s. Root on USB survived; Wi-Fi reconnected on its own; touchpad works after an "incomplete report" warning on resume. This closes the map's "suspend/resume validation" fog item.

Not working: audio (no card; expected without qcom-firmware-compat), battery capacity (devices register, values empty), media keys (#28). Bluetooth powered but qca/wcnhpbtfw20.tlv missing (#19 scope); pairing untested. Caps Lock "not working" turned out to be Omarchy's default kb_options = "compose:caps,shift:both_capslock_cancel": the key sends KEY_CAPSLOCK, the LED works, both Shifts toggle caps. Not a bug.

Packages on this ISO are the fork lineage (shim 1-2, hook 1.37.1-1.1), not the #222 versions. Hostname on the tailnet is now omarchy-snappy-1.

[JimmayVV]

Correction: battery reporting works. The earlier "capacity reads empty" was a misread on my side. qcom-battmgr-bat exposes energy_now/energy_full/voltage_now, not a capacity attribute, and upower computes the percentage from those: state fully-charged, 56.96 Wh of 59.16 Wh design, 100%, health 96.3%. The bar shows the battery with a charging bolt. Screenshot taken over SSH with grim confirms the bar: Bluetooth, Wi-Fi, volume, display, battery all present. Battery is fully working, not half-working; the map's fog entry saying otherwise was based on the same misread.

[JimmayVV]

Testing line posted on omacom#129 (omacom#129 (comment)), 2026-08-28. The map's one standing obligation to the stack is met: two verified X1E machines. Remaining on this ticket: answer review questions if any come; the encrypted path on the G1q is optional extra evidence.

[JimmayVV]

2026-08-28: posted on omacom/omarchy#8672 (issuecomment-5448755706) that the shared speaker silence is ALARM's missing CONFIG_RESET_GPIO, with the #26 resolution linked; asks nothing of birkskyum and changes nothing in the stack. The fix is #32 (ALARM config PR), outside these PRs.

[JimmayVV]

2026-08-28 13:44Z: birkskyum validated the RESET_GPIO fix on the Yoga and submitted archlinuxarm/PKGBUILDs#2217 crediting Jimmy. omacom#129 left draft. Still no maintainer comment on any of the six PRs. Jimmy's remaining action: a G1q validation line on #2217 (tracked on #32).

[JimmayVV]

Status 2026-09-14 — two of the five merged, into dragon.

PR	State
omacom/omarchy#8672	merged 2026-09-13 into dragon (7 of Jimmy's commits, his authorship)
omacom#129	merged 2026-09-13 into dragon (6 of Jimmy's commits)
omacom/omarchy#8673 (Yoga board)	merged 2026-09-13 into dragon
omacom/omarchy-pkgs#221 firmware extract	open
#222 kernel image across updates	open
#223 repo workflow	open
#380 Limine/mkinitcpio drop-ins on aarch64	open
omacom#171 native ARM builds	open

birkskyum merged omacom#129 and #8672 himself. omacom#121 and #8039 closed as superseded. dragon is 19 commits ahead of basecamp/omarchy quattro, with no dragon→quattro PR yet. omacom#129's body says a release image needs #221, #222 and #380.

This ticket resolves when the PRs merge or close, so it stays open for the package PRs. The standing obligation, a G1q Testing line on omacom#129's image, is blocked on the G1q booting again: last recorded state is black screen on the fork-built 7.2.1-2 kernel (#32). ALARM's own 7.2.5-1 now ships CONFIG_RESET_GPIO, so rolling back and then taking ALARM's kernel is the path.