Skip to content

omamobile-omarchy 0.1.38: omamobile-update refuses to run as root, and passes OMARCHY_PATH on - #548

Merged
derekross merged 2 commits into
mainfrom
omamobile-update-omarchy-path
Nov 9, 2026
Merged

derekross merged 2 commits into
mainfrom
omamobile-update-omarchy-path

Conversation

@derekross

@derekross derekross commented Nov 9, 2026 •

Copy link
Copy Markdown
Owner

Two changes to omamobile-update, which omarchy update runs on the phone:

  1. It refuses to run an update as root (run, start, run-migration; exit 5, with the reason). Omarchy records the migrations it has run in the caller's home (~/.local/state/omarchy/migrations). On 2026-11-09 the release test phone was updated from a root login over SSH. Root's record was empty, so Omarchy ran 59 migrations again as root:

    • one swapped quickshell for an older quickshell-git (Omarchy's later migration that swaps it back was also recorded as done, so it didn't run);
    • others wrote browser and systemd configs into /root and restarted systemd-oomd.

    No package moved apart from that swap (and vulkan-headers, which it pulled in) and the normal updates, and nothing else in /etc changed. It also left /tmp/omarchy-update.log owned by root. Omarchy's update logs to that fixed path through script, so every later update as the user stopped at once ("script: cannot open /tmp/omarchy-update.log: Permission denied") until the file was removed. check, status and channel still answer as root.

  2. It passes OMARCHY_PATH on when the environment has none, as on a TTY login, because Omarchy's scripts read it under set -u. This change alone made the root path reachable. Before it, the root run stopped early on "unbound variable", which is why (1) was needed.

Tests: test/update.sh covers both. With the guard removed, the four root checks fail.

🤖 Generated with Claude Code

derekross and others added 2 commits November 9, 2026 00:52
Omarchy's update scripts read OMARCHY_PATH under set -u, and a session
gets it from Omarchy's env-bootstrap. A root login over SSH or a TTY has
none, so `omarchy update` there stopped after updating, in
omarchy-update-dev ("OMARCHY_PATH: unbound variable"; the 2026.11.08
test install on the Pixel 7 Pro, over tools/phone-ssh). omamobile-update
already knew the default, /usr/share/omarchy, but passed it only to the
migrations it runs; it now sets it for Omarchy's update too when the
environment has none.

test/update.sh runs the sheet's update once more with OMARCHY_PATH
removed from the environment and checks that omarchy-update receives
Omarchy's path; against the previous omamobile-update that check fails.
test/all: 233 ok.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Omarchy records the migrations it has run in the caller's home
(~/.local/state/omarchy/migrations). On 2026-11-09 the release test
phone was updated from a root login over SSH: root's record was empty,
so Omarchy ran 59 migrations again as root. One swapped quickshell
for an older quickshell-git (Omarchy's own later migration swaps it
back, but that one counted as done too), others wrote configs into
/root and restarted systemd-oomd. Before the OMARCHY_PATH default in
the previous commit, the same run stopped early on "unbound variable";
the default alone made the root path reachable.

run, start, __inner and run-migration now refuse as root with exit 5
and say why; check, status and channel still answer. The tests stand
in for root with OMAMOBILE_UPDATE_TEST_EUID.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@derekross derekross changed the title omamobile-omarchy 0.1.38: omarchy update without a session environment omamobile-omarchy 0.1.38: omamobile-update refuses to run as root, and passes OMARCHY_PATH on Nov 9, 2026
derekross added a commit that referenced this pull request Nov 9, 2026
…an back

From review: the Installing section still named the 2026.11.03 bundle
and URL; the flat-battery item claimed a recovery the model-load retry
hasn't shown on a real flat battery (#169 is open); "and sound" named
nothing; the kernel-update claim now rests on this release's own test
install; Hyprland is 0.57.0-3.2 on stable; bare issue numbers are links.

The OMARCHY_PATH known issue was wrong about where root's update
stops (before updating anything), and incomplete: with the variable
set by hand the 2026-11-09 test run went on and ran Omarchy's
migrations again as root. It now says to run the update as the user,
and that omamobile-omarchy 0.1.38 (#548) refuses root.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 9, 2026
…ests

The identifiers stage dumped /etc, /root, /usr/share/omamobile and
/var/lib/omamobile from the root image into one directory. rdump names
each by its last component, so the second "omamobile" collided with the
first ("File exists while making directory") and debugfs skipped it,
behind >/dev/null and || true: /var/lib/omamobile was never scanned.
Each is now dumped at its own path, and the stage stops if a directory
the image has didn't arrive. Checked on a scratch ext4 image built with
mke2fs -d: the old commands lose var/lib/omamobile, the new ones dump
all three.

Two tests close mutants the review found surviving: a dhcp-option value
in another file is found, and a path that only ends in etc/dnsmasq.conf
(xetc/dnsmasq.conf) gets no exemption.

The test-install runbook now says to run omarchy update on the phone as
its user, never as root over USB SSH (#548).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 9, 2026
…asq's example addresses in its one file (#550)

* release: the identifier scan skips group MACs and dnsmasq's option bytes

The 2026.11.08 identifiers stage stopped on 15 findings, all in the root
image's /etc/dnsmasq.conf: dnsmasq (an omamobile-shell dependency since
#525, for the Hotspot tile) ships Arch's example configuration, with
11:22:33:44:55:66 in 14 commented dhcp-host examples and
e4:45:74:68:00:00 as a commented dhcp-option=128 value.

Neither is anyone's address. 0x11 has the group (multicast) bit set, and
no network card's own address does, so group addresses join the locally
administered ones the scan already treats as made up. The second is
option data ("\xe4Eth\0\0"), so in etc/dnsmasq.conf a value on a
dhcp-option line, commented or not, is skipped. A universally
administered unicast address still counts everywhere else, a dhcp-host
line in dnsmasq.conf included.

Checked with tools/release scan: dnsmasq 2.93-1's shipped dnsmasq.conf
passes (the previous scan reports it); the same file with an added
dhcp-host=a4:5e:60:12:34:56 line, and a plain file naming
a4:5e:60:12:34:58, are both still reported.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release: the dnsmasq exemption covers its one file only, with tests

Review of the previous commit: a group-bit exemption across the whole
image also passes a quarter of Bluetooth LE static random addresses
(top bits 11, any low bits), which AGENTS.md counts as identifiers.
The only group address the release hit is dnsmasq's 11:22:33:44:55:66,
so the group bit is now skipped in /etc/dnsmasq.conf alone, as are
dhcp-option values there, up to a comment on the same line.

tests/release: the shipped file's lines pass; a unicast dhcp-host in
it, an address in a trailing comment after an option value, and a
group-looking address in another file are all found. Test addresses
come from RFC 7042's documentation range and are put together at run
time, so the test file holds none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release: the identifier scan reads /var/lib/omamobile, and two more tests

The identifiers stage dumped /etc, /root, /usr/share/omamobile and
/var/lib/omamobile from the root image into one directory. rdump names
each by its last component, so the second "omamobile" collided with the
first ("File exists while making directory") and debugfs skipped it,
behind >/dev/null and || true: /var/lib/omamobile was never scanned.
Each is now dumped at its own path, and the stage stops if a directory
the image has didn't arrive. Checked on a scratch ext4 image built with
mke2fs -d: the old commands lose var/lib/omamobile, the new ones dump
all three.

Two tests close mutants the review found surviving: a dhcp-option value
in another file is found, and a path that only ends in etc/dnsmasq.conf
(xetc/dnsmasq.conf) gets no exemption.

The test-install runbook now says to run omarchy update on the phone as
its user, never as root over USB SSH (#548).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 9, 2026
* release notes 2026.11.08

The notes for the first public release: kernel-2026.11.08 (1a9d1d4e9289),
Omarchy held at 4.0.6, what changed since the 2026.11.03 candidate
(double tap to wake, the reboot reason, the flat-battery charger
re-check, the touch, Wi-Fi and RTC fixes, the shell's changes), and the
known issues the test install found: Google's firmware still a separate
provisioning step, omarchy update as root over SSH (#548), the flat
battery recovery (#169), the modem stop (#532), Omarchy 4.5 phones,
systemd-firstboot. tools/release scan: no identifiers.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release notes 2026.11.08: the right bundle, and claims this release can back

From review: the Installing section still named the 2026.11.03 bundle
and URL; the flat-battery item claimed a recovery the model-load retry
hasn't shown on a real flat battery (#169 is open); "and sound" named
nothing; the kernel-update claim now rests on this release's own test
install; Hyprland is 0.57.0-3.2 on stable; bare issue numbers are links.

The OMARCHY_PATH known issue was wrong about where root's update
stops (before updating anything), and incomplete: with the variable
set by hand the 2026-11-09 test run went on and ran Omarchy's
migrations again as root. It now says to run the update as the user,
and that omamobile-omarchy 0.1.38 (#548) refuses root.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release notes 2026.11.08: the hook ran on a reinstall, not a kernel update

The test install's hook run came from reinstalling the release's own
linux-omamobile: four images matched and vendor_kernel_boot was
rewritten because the hook adds the owner's SSH key to its initramfs.
Say so, and point at the 2026-11-02 update and rollback for a real
kernel update through the same omamobile-boot-update 0.1.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 9, 2026
… as root (#560)

* device/cheetah: NOTES, the 2026.11.08 test install, and an update run as root

The release record §11.6 asks for: the signed bundle's install, the
hand provisioning, a kernel update through omamobile-boot-update's hook,
and a clean omarchy update as the user. It also records what a root
update did to the phone and how it was put back (#548).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* device/cheetah: NOTES, the hook ran on a reinstall; the root run had OMARCHY_PATH set

From review: the 00:42 hook run came from reinstalling the release's
own linux-omamobile, and vendor_kernel_boot differed from the bundle's
because the hook adds the owner's authorized_keys to the initramfs.
The root update had OMARCHY_PATH set by hand; vulkan-headers came and
went with the quickshell swap; and the 2026-11-03 gotcha that gave the
recipe now says not to.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
@derekross
derekross merged commit 09b7eb2 into main Nov 9, 2026
2 checks passed
@derekross
derekross deleted the omamobile-update-omarchy-path branch November 9, 2026 16:32
derekross added a commit that referenced this pull request Nov 9, 2026
Round 46's confirmed candidates from the reviews of the open
omarchy-mobile and linux-omarchy-mobile pull requests, the round's own
process slips (shared refs and scratch files deleted by glob, a
maintainer fixup that skipped the package's test/all, an amend on a
contributor branch) and the 2026.11.08 release PRs (#547, #548, #549,
#550, #560). Repeats of existing lessons extend them instead: L-45
(kernel #222), L-183 (#409), L-426 (the round's version collisions,
now set by hand at merge) and L-550 (#471 with #420). C-34 and C-35
correct our #290 shell suggestion and round 45's claim that
tests/dipper-probe guards anything.

Stacked on #551 (round 45's lessons).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Nov 9, 2026
Round 46's confirmed candidates from the reviews of the open
omarchy-mobile and linux-omarchy-mobile pull requests, the round's own
process slips (shared refs and scratch files deleted by glob, a
maintainer fixup that skipped the package's test/all, an amend on a
contributor branch) and the 2026.11.08 release PRs (#547, #548, #549,
#550, #560). Repeats of existing lessons extend them instead: L-45
(kernel #222), L-183 (#409), L-426 (the round's version collisions,
now set by hand at merge) and L-550 (#471 with #420). C-34 and C-35
correct our #290 shell suggestion and round 45's claim that
tests/dipper-probe guards anything.

Stacked on #551 (round 45's lessons).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…asq's example addresses in its one file (derekross#550)

* release: the identifier scan skips group MACs and dnsmasq's option bytes

The 2026.11.08 identifiers stage stopped on 15 findings, all in the root
image's /etc/dnsmasq.conf: dnsmasq (an omamobile-shell dependency since
derekross#525, for the Hotspot tile) ships Arch's example configuration, with
11:22:33:44:55:66 in 14 commented dhcp-host examples and
e4:45:74:68:00:00 as a commented dhcp-option=128 value.

Neither is anyone's address. 0x11 has the group (multicast) bit set, and
no network card's own address does, so group addresses join the locally
administered ones the scan already treats as made up. The second is
option data ("\xe4Eth\0\0"), so in etc/dnsmasq.conf a value on a
dhcp-option line, commented or not, is skipped. A universally
administered unicast address still counts everywhere else, a dhcp-host
line in dnsmasq.conf included.

Checked with tools/release scan: dnsmasq 2.93-1's shipped dnsmasq.conf
passes (the previous scan reports it); the same file with an added
dhcp-host=a4:5e:60:12:34:56 line, and a plain file naming
a4:5e:60:12:34:58, are both still reported.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release: the dnsmasq exemption covers its one file only, with tests

Review of the previous commit: a group-bit exemption across the whole
image also passes a quarter of Bluetooth LE static random addresses
(top bits 11, any low bits), which AGENTS.md counts as identifiers.
The only group address the release hit is dnsmasq's 11:22:33:44:55:66,
so the group bit is now skipped in /etc/dnsmasq.conf alone, as are
dhcp-option values there, up to a comment on the same line.

tests/release: the shipped file's lines pass; a unicast dhcp-host in
it, an address in a trailing comment after an option value, and a
group-looking address in another file are all found. Test addresses
come from RFC 7042's documentation range and are put together at run
time, so the test file holds none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release: the identifier scan reads /var/lib/omamobile, and two more tests

The identifiers stage dumped /etc, /root, /usr/share/omamobile and
/var/lib/omamobile from the root image into one directory. rdump names
each by its last component, so the second "omamobile" collided with the
first ("File exists while making directory") and debugfs skipped it,
behind >/dev/null and || true: /var/lib/omamobile was never scanned.
Each is now dumped at its own path, and the stage stops if a directory
the image has didn't arrive. Checked on a scratch ext4 image built with
mke2fs -d: the old commands lose var/lib/omamobile, the new ones dump
all three.

Two tests close mutants the review found surviving: a dhcp-option value
in another file is found, and a path that only ends in etc/dnsmasq.conf
(xetc/dnsmasq.conf) gets no exemption.

The test-install runbook now says to run omarchy update on the phone as
its user, never as root over USB SSH (derekross#548).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
* release notes 2026.11.08

The notes for the first public release: kernel-2026.11.08 (1a9d1d4e9289),
Omarchy held at 4.0.6, what changed since the 2026.11.03 candidate
(double tap to wake, the reboot reason, the flat-battery charger
re-check, the touch, Wi-Fi and RTC fixes, the shell's changes), and the
known issues the test install found: Google's firmware still a separate
provisioning step, omarchy update as root over SSH (derekross#548), the flat
battery recovery (derekross#169), the modem stop (derekross#532), Omarchy 4.5 phones,
systemd-firstboot. tools/release scan: no identifiers.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release notes 2026.11.08: the right bundle, and claims this release can back

From review: the Installing section still named the 2026.11.03 bundle
and URL; the flat-battery item claimed a recovery the model-load retry
hasn't shown on a real flat battery (derekross#169 is open); "and sound" named
nothing; the kernel-update claim now rests on this release's own test
install; Hyprland is 0.57.0-3.2 on stable; bare issue numbers are links.

The OMARCHY_PATH known issue was wrong about where root's update
stops (before updating anything), and incomplete: with the variable
set by hand the 2026-11-09 test run went on and ran Omarchy's
migrations again as root. It now says to run the update as the user,
and that omamobile-omarchy 0.1.38 (derekross#548) refuses root.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* release notes 2026.11.08: the hook ran on a reinstall, not a kernel update

The test install's hook run came from reinstalling the release's own
linux-omamobile: four images matched and vendor_kernel_boot was
rewritten because the hook adds the owner's SSH key to its initramfs.
Say so, and point at the 2026-11-02 update and rollback for a real
kernel update through the same omamobile-boot-update 0.1.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
… as root (derekross#560)

* device/cheetah: NOTES, the 2026.11.08 test install, and an update run as root

The release record §11.6 asks for: the signed bundle's install, the
hand provisioning, a kernel update through omamobile-boot-update's hook,
and a clean omarchy update as the user. It also records what a root
update did to the phone and how it was put back (derekross#548).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* device/cheetah: NOTES, the hook ran on a reinstall; the root run had OMARCHY_PATH set

From review: the 00:42 hook run came from reinstalling the release's
own linux-omamobile, and vendor_kernel_boot differed from the bundle's
because the hook adds the owner's authorized_keys to the initramfs.
The root update had OMARCHY_PATH set by hand; vulkan-headers came and
went with the quickshell swap; and the 2026-11-03 gotcha that gave the
recipe now says not to.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…d passes OMARCHY_PATH on (derekross#548)

* omamobile-omarchy 0.1.38: omarchy update without a session environment

Omarchy's update scripts read OMARCHY_PATH under set -u, and a session
gets it from Omarchy's env-bootstrap. A root login over SSH or a TTY has
none, so `omarchy update` there stopped after updating, in
omarchy-update-dev ("OMARCHY_PATH: unbound variable"; the 2026.11.08
test install on the Pixel 7 Pro, over tools/phone-ssh). omamobile-update
already knew the default, /usr/share/omarchy, but passed it only to the
migrations it runs; it now sets it for Omarchy's update too when the
environment has none.

test/update.sh runs the sheet's update once more with OMARCHY_PATH
removed from the environment and checks that omarchy-update receives
Omarchy's path; against the previous omamobile-update that check fails.
test/all: 233 ok.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* omamobile-omarchy 0.1.38: omamobile-update refuses to run as root

Omarchy records the migrations it has run in the caller's home
(~/.local/state/omarchy/migrations). On 2026-11-09 the release test
phone was updated from a root login over SSH: root's record was empty,
so Omarchy ran 59 migrations again as root. One swapped quickshell
for an older quickshell-git (Omarchy's own later migration swaps it
back, but that one counted as done too), others wrote configs into
/root and restarted systemd-oomd. Before the OMARCHY_PATH default in
the previous commit, the same run stopped early on "unbound variable";
the default alone made the root path reachable.

run, start, __inner and run-migration now refuse as root with exit 5
and say why; check, status and channel still answer. The tests stand
in for root with OMAMOBILE_UPDATE_TEST_EUID.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
Round 46's confirmed candidates from the reviews of the open
omarchy-mobile and linux-omarchy-mobile pull requests, the round's own
process slips (shared refs and scratch files deleted by glob, a
maintainer fixup that skipped the package's test/all, an amend on a
contributor branch) and the 2026.11.08 release PRs (derekross#547, derekross#548, derekross#549,
derekross#550, derekross#560). Repeats of existing lessons extend them instead: L-45
(kernel derekross#222), L-183 (derekross#409), L-426 (the round's version collisions,
now set by hand at merge) and L-550 (derekross#471 with derekross#420). C-34 and C-35
correct our derekross#290 shell suggestion and round 45's claim that
tests/dipper-probe guards anything.

Stacked on derekross#551 (round 45's lessons).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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.

1 participant