Skip to content

device-bluejay 0.1.0: the Pixel 6a's first-light package, nothing enabled - #346

Draft
PeterXMR wants to merge 1 commit into
derekross:mainfrom
PeterXMR:device-bluejay-package
Draft

PeterXMR wants to merge 1 commit into
derekross:mainfrom
PeterXMR:device-bluejay-package

Conversation

@PeterXMR

@PeterXMR PeterXMR commented Nov 2, 2026 •

Copy link
Copy Markdown
Contributor

What and why

Stage 0, part 2 of the Pixel 6a port (#2): packages/omamobile-device-bluejay 0.1.0, the package device/bluejay/device.conf names, so tools/build-rootfs can build a 6a image with the [omamobile] repository configured. It follows the merged first-light DTS (linux-omarchy-mobile#179) the same way: nothing has booted it, and it stays a draft until a 6a owner posts the first tools/phone-smoke report on #2.

It is felix's first-light shape without the fold's lid and cover-touch parts, plus the suspend-to-idle drop-in the #179 review asked for: the storage udev rule, the usb0 address with USB as the last gateway, NetworkManager off usb0, logind ignoring the power key, [Sleep] on s2idle with no hibernation (raven's file byte for byte: gs101's firmware offers PSCI SYSTEM_SUSPEND, the kernel lists deep, and Linux has no gs101 sleep sequence to wake from it; bluejay's bootargs, like oriole's, don't set mem_sleep_default), the mirrorlist and pacman hook, reboot-to and a copy of the descriptor. provides/conflicts=(omamobile-device), depends on omamobile-keyring.

Not in it, on purpose: no service is enabled, no modprobe policy ships, no omamobile-bootctl or boot-success marker (marking slot B on the gs101 phones is on hold, #252; bin/omamobile initramfs builds its freestanding helpers without the source, as for oriole), no USB rescue, Wi-Fi, charger, haptics, display options or cameras. Each comes in when a booted 6a has given it a runbook and the evidence. The package is also not added to tools/build-packages' ORDER or to tests/all's package list: husky, tegu and dodge aren't there either, so joining them is the maintainer's call when a package is published. A default tools/build-packages still builds it, through the trailing glob after the listed packages.

Every file is a copy of one that runs on the Pixel 7 Pro (cheetah) or the Pixel 6 Pro (raven, the same SoC), and test/all pins each copy with cmp, the descriptor against device/bluejay/device.conf, and checks that the staged tree holds no system unit, no modprobe policy and nothing that marks a slot, loads a driver or starts hardware. The cmp pins run only from the repository (tests/packages, CI): makepkg's check() on a copied package directory skips them, as #222 asked.

Two caveats the README records: a replug may need a reboot, because nothing closes the Type-C chip's USB2 switch again (raven's package does for its MAX77759; the 6a's chip, bus and address are unread), and on a first boot reboot-to bootloader most likely just boots slot B again, because the descriptor lists no INITRAMFS_MODULES (raven's records the PMIC and reboot-mode modules it needs), so fastboot is by the buttons.

Device and build

  • Phone: none: no Pixel 6a has booted our kernel; the package is prepared for the first boot
  • Kernel release (uname -r): n/a
  • Packages changed, with versions: omamobile-device-bluejay 0.1.0 (new)

Checks (no phone needed)

  • tools/pr-check origin/main passes (7 ok)
  • tests/all passes: all 34 suites ok, run in CI's archlinux:latest container with Docker on this Mac (the host shell alone can't: bash 3.2, BSD sed), and CI's Tests job is green
  • tests/packages passes: omamobile-device-bluejay 17 ok, 24 of 26 suites ok in the same container; omamobile-audio and omamobile-omarchy fail there only under x86_64 emulation (a 2 s watchdog wait, a missing exec reported as 127 instead of ENOENT) and pass in CI
  • Each changed package has its version bumped, and the new version is in the commit subject
  • Commits have area prefixes and bodies that say why; any AI help is a Co-Authored-By: or Assisted-by: trailer, never a Signed-off-by:

On the phone

tools/phone-smoke report

Not run: no Pixel 6a has booted this build. The first boot's report goes here and to #2.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Peter. This is a tidy, conservative package: every file is a copy of one that runs on a Pixel today, the test pins each copy, and leaving out bootctl, the USB rescue and every service until a 6a has booted is exactly right. Including raven's s2idle drop-in, since bluejay's bootargs don't set mem_sleep_default, is a good catch.

Two things before it can land:

  1. Order with device-bluejay: the factory image settles AVB, boot gets a footer #350. test/all compares the package's device.conf with device/bluejay/device.conf, and device-bluejay: the factory image settles AVB, boot gets a footer #350 adds the three AVB lines there. I'm merging device-bluejay: the factory image settles AVB, boot gets a footer #350 now, so please refresh the package's copy (and its NOTES row) on top of it.
  2. Draft until the first boot. Your description says the package stays a draft until a 6a owner posts the first smoke report, which matches what CONTRIBUTING asks of a package. Could you mark the PR as a draft? A first boot needs only slot B's images, not a root image, so nothing waits on this package before then.

A small one: raven's and tegu's READMEs carry the SPDX header line; please add it here too.

@PeterXMR
PeterXMR marked this pull request as draft November 3, 2026 21:57
@PeterXMR
PeterXMR force-pushed the device-bluejay-package branch from c717b3b to e12583a Compare November 3, 2026 22:06
@derekross

Copy link
Copy Markdown
Owner

Thank you, Peter, and sorry for the slow reply. Everything I asked for on the 3rd is in: the package's device.conf carries #350's AVB lines and still matches the descriptor byte for byte, the README has its SPDX line, and the PR is a draft. Pointing the first boot's report at the 6a's own thread, #348, is a good change too.

I re-ran the package's test today on top of current main. It passes (17 ok), and every file it pins still matches its source.

Nothing more is needed from you until a Pixel 6a boots. Two things I'll take care of when this lands, so you don't need to chase them:

  • docs/DEVICES.md now conflicts with main, because the rows around the 6a's line moved. It's a one-line resolution, and I'll do it at merge.
  • If the reboot-to --help fix (reboot-to: print usage on an option instead of rebooting #459) merges first, the package's copy of reboot-to.c needs refreshing to match, because your test compares it with the Pixel 7 Pro's. I'll fold that in too.

Thanks again for keeping this one so conservative.

…bled

The package device/bluejay/device.conf names, so tools/build-rootfs can
build a 6a image with the [omamobile] repository configured: felix's
first-light shape without the fold's lid and cover-touch parts, plus
the suspend-to-idle drop-in. The storage udev rule, the usb0 address
with USB as the last gateway, NetworkManager off usb0, logind ignoring
the power key, [Sleep] on s2idle with no hibernation, the mirrorlist
and pacman hook, reboot-to (cheetah's source) and a copy of the
descriptor; provides and conflicts omamobile-device, depends on
omamobile-keyring.

The sleep drop-in is raven's, byte for byte: gs101's firmware offers
PSCI SYSTEM_SUSPEND, so the kernel lists deep, and Linux has no gs101
sleep sequence to wake from it (device/raven/SUSPEND-PLAN.md).
Bluejay's bootargs, like oriole's, don't set mem_sleep_default, so the
file is what keeps systemctl suspend on s2idle, as asked in the review
of linux-omarchy-mobile#179.

No service is enabled, no modprobe policy ships, and there is no
omamobile-bootctl: marking slot B on the gs101 phones is on hold
(derekross#252), and bin/omamobile initramfs builds its freestanding helpers
without the source, as for oriole. No Pixel 6a has run it; every file
is a copy of one that runs on raven or cheetah, and test/all pins each
copy. The descriptor copy carries the AVB lines the factory image
settled (derekross#350), the same bytes as device/bluejay/device.conf. The
first boot's phone-smoke report goes to derekross#348, the 6a's own thread, and
NOTES.md.

test/all and tests/packages pass on the laptop. The README carries the
SPDX header line raven's and tegu's do.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@PeterXMR
PeterXMR force-pushed the device-bluejay-package branch from e12583a to 8d0e50f Compare November 11, 2026 17:26
@latentoperator

Copy link
Copy Markdown
Contributor

The first Pixel 6a boot/smoke result is now posted in #348, with the persistent desktop/dependency notes already merged through #561. This is the historical development stack, not a release-kernel or passive-package qualification.

I kept the follow-up on your current 8d0e50fa657f0bf3942aa70bf1ee53b8a10645e3 instead of opening a competing package PR. The focused coordination branch adds only this follow-up commit: descriptor BOOT_UPDATE_AUTO_WRITE=no, installed AUTO_WRITE=no as a pacman backup file, tests and qualification/dependency wording. It is labeled 0.1.1 for the follow-up; feel free to fold the policy into your initial version if preferred. #563 must land first for the descriptor policy to control new installations.

The earlier local patch's retained-display trigger is intentionally excluded. This branch enables no service, driver, experimental display refresh or slot-success writer. Historical startup/configuration stays in #565.

Validation: the complete package fixtures and policy checks pass, independent source review found no remaining blocker, and the candidate builds as an aarch64 package. The shared reboot-to source already matches current main, so no extra source-refresh delta is needed.

The current development handset reports device-bluejay 0.1.1 too, but its older package owns active check/service/modprobe files that this passive candidate omits. I have not replaced it and do not claim a qualified upgrade: those files need an explicit preserved/migrated startup arrangement first. No package agreement or hardware migration is assumed here. The automatic boot-success mask and AUTO_WRITE=no remain in force.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Peter, for the rebase: it resolved the docs/DEVICES.md conflict I'd said I would do at merge, and the package itself is byte for byte the one I checked on the 8th. I ran its test on your head merged onto today's main: 17 ok, and every pinned copy still matches its source, reboot-to.c included, so nothing needs refreshing there.

One thing is needed from you. CI is red on a lint that landed on main after you wrote the test: no script under pipefail may pipe into grep -q, because grep -q exits early and the writer gets a SIGPIPE. It's line 122 of packages/omamobile-device-bluejay/test/all. The other device packages have the same line in the fixed form:

! sed -n '/^\[Network\]/,/^\[/p' "$net" | grep -c '^Gateway=' >/dev/null || fail "usb0 sets no default route in [Network] (metric 0)"

Because tests/all stops there, CI never got to the package suites on this head, so the fix also gets your own test running in CI again.

On leaving draft: not yet, and it isn't waiting on you. A Pixel 6a boots now (thank you, @latentoperator, for the report in #348), but on a development kernel with a private package of the same name that owns start-up services and a module blacklist, and this package hasn't been installed on a 6a. What I'd like to see before it leaves draft is a smoke report from a 6a where pacman -Q lists omamobile-device-bluejay 0.1.0-1.

@latentoperator, three things on how that comes together:

  • The version on your handset. It reports device-bluejay 0.1.1 from a local package. The repository's 0.1.0 would never install over that, and the first published version above yours would replace it on the next omarchy update and remove the files it owns: the blacklist and the ordered start-up units. On your development kernel those drivers would then load on their own at the next boot. So please don't let that happen by accident: until the move is planned, keep IgnorePkg = omamobile-device-bluejay in that phone's pacman.conf, or rename the local package.
  • The order. device/bluejay: keep pinned startup outside the passive package #565 first, so the start-up and the blacklist are bring-up files that no package owns. Then replacing the local package with this one is its own attended step with a short runbook in device/bluejay/ (which hand copies go where, the RAM rescue image ready, one boot, pacman -Q, tools/phone-smoke). Note that this package's 20-usb-gadget.network differs from your bring-up copy: it adds the host as the last gateway (metric 2000) and public DNS.
  • The 0.1.1 follow-up. Thank you for coordinating here rather than opening a competing package. I'd keep it out of this pull request: it depends on install-phone: preserve development boot-update policy and serial pin #563, and it makes the package own /etc/omamobile/boot-update.conf, a file that so far only the installer writes. Without that file the boot-update hook already does nothing, so let's settle it in install-phone: preserve development boot-update policy and serial pin #563's review.

Peter, when you push the one-line fix, two sentences could also catch up with the first boot: the description's "no Pixel 6a has booted our kernel" and the README's "merged without a boot" are now "a 6a boots a development kernel (#348); the published kernel and this package haven't run on it yet".

@latentoperator

Copy link
Copy Markdown
Contributor

November 13 coordination update: #562, #563 and #565 have landed, and this 6a is now running Omarchy 4.5.0-3 with its mobile migrations completed. The revised kernel's headless RAM/touch/pool/PCIe/Wi-Fi stages also passed; the current graphical desktop is restored on the historical development kernel (details in #348).

The package gate remains exactly as requested: the passive 0.1.0-1 package has not been installed or smoke-qualified. The locally active 0.1.1 package remains protected by IgnorePkg=omamobile-device-bluejay. Its startup/module-blacklist ownership and the USB network/DNS differences still need the separate attended migration runbook before replacement.

I am not asking to fold the earlier boot-update.conf-owning 0.1.1 coordination follow-up into this PR; your requested ownership boundary stands. The author's lint fix and passive-package phone qualification remain separate from the new kernel results.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @latentoperator, for the update, and for keeping the local package held and leaving the boot-update.conf question to its own review. That's exactly the order I hoped for.

@PeterXMR, two one-line changes are needed when you next push, and the second one is my fault:

  1. The lint fix from last time. Line 122 of packages/omamobile-device-bluejay/test/all still pipes into grep -q:

    ! sed -n '/^\[Network\]/,/^\[/p' "$net" | grep -c '^Gateway=' >/dev/null || fail "usb0 sets no default route in [Network] (metric 0)"
  2. reboot-to.c needs refreshing after all. On the 12th I wrote that every pinned copy still matched, reboot-to.c included. A few hours later reboot-to: print usage on an option instead of rebooting #459 changed the Pixel 7 Pro's reboot-to.c (an option now prints the usage instead of rebooting), so on today's main your test stops at "reboot-to.c drifted from the cheetah package". Please copy packages/omamobile-device-cheetah/src/reboot-to.c from main over your package's copy. I checked that with that copy the test passes again (17 ok) on your branch merged with main.

While you're in there, the two sentences from last time can catch up too: "no Pixel 6a has booted our kernel" in the description and "merged without a boot" in the README are now "a 6a boots a development kernel (#348); the published kernel and this package haven't run on it yet".

Nothing else has changed: the pull request stays a draft until there's a smoke report from a 6a with omamobile-device-bluejay 0.1.0-1 installed. That comes after the attended move from the local package, which gets its own short runbook in device/bluejay/. Thank you both.

latentoperator added a commit to latentoperator/omarchy-mobile that referenced this pull request Nov 14, 2026
The boot-update.conf ownership follow-up was not accepted on derekross#346. Keep the package passive and describe the attended migration and smoke report that still gate it.

Assisted-by: OpenAI Codex
latentoperator added a commit to latentoperator/omarchy-mobile that referenced this pull request Nov 14, 2026
The boot-update.conf ownership follow-up was not accepted on derekross#346. Keep the package passive and describe the attended migration and smoke report that still gate it.

Assisted-by: OpenAI Codex
latentoperator added a commit to latentoperator/omarchy-mobile that referenced this pull request Nov 14, 2026
The boot-update.conf ownership follow-up was not accepted on derekross#346. Keep the package passive and describe the attended migration and smoke report that still gate it.

Assisted-by: OpenAI Codex
derekross pushed a commit that referenced this pull request Nov 15, 2026
…ejay UPSTREAM status (#666)

The transfer profile behind linux-omarchy-mobile #225's aggregation-retry fix needs a phone, so it lives here and not in the kernel's selftests. Sixteen fresh `tools/phone-ssh` connections each pull 2 MiB of zeros from the phone, checked by size and SHA-256 under one 20 s deadline, with no retry and no write on the phone. A failure names the transfer and the check and shows ssh's stderr. `tests/ncm-transfer` runs it against a stand-in `ssh`. A pass shows that the link carried the transfers, not that the retry ran, and it covers neither laptop sleep nor a replug. Bluejay's UPSTREAM page is brought up to the November kernel revisions, none of which has booted on the phone, and it records that #346 stays passive.

Assisted-by: OpenAI Codex
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants