Skip to content

omamobile-shell 0.2.132: the fingerprint glyph appears when the fingerprint stack starts after the shell - #425

Merged
derekross merged 7 commits into
derekross:mainfrom
W4HO-ham:shell-fp-glyph-late-stack
Nov 15, 2026
Merged

derekross merged 7 commits into
derekross:mainfrom
W4HO-ham:shell-fp-glyph-late-stack

Conversation

@W4HO-ham

@W4HO-ham W4HO-ham commented Nov 4, 2026

Copy link
Copy Markdown
Contributor

On the palmtop this morning fingerprint unlock worked but the lock screen showed no fingerprint glyph. The fingerprint stack had started a few minutes after login, and the plugin only reads the bridge's state file at start and through a file watch; with no file and no directory yet, the watch never fires, and the timed re-read runs only while the glyph shows. A shell restart brought the glyph back.

The re-read timer now also runs while there is no state (every 2 s), so the glyph appears once the stack starts, whenever that is. The shell's test suite and tools/pr-check pass.

…rprint stack starts after the shell

The fingerprint plugin reads the bridge's state file once at start and then follows it with a file watch, re-reading on a timer only while the glyph shows. When the fingerprint stack starts after the shell, the file and its directory don't exist yet, so the watch has nothing to follow and the glyph never appears, although fingerprint unlock works. On the palmtop on 2026-11-04 the stack started a few minutes after login and the lock screen showed no glyph until the shell was restarted. The re-read timer now also runs while there is no state, every 2 s, until the file appears.
@derekross

Copy link
Copy Markdown
Owner

Thank you, @W4HO-ham, for chasing this down. The diagnosis is spot on: with no state file and no directory when
the shell starts, the plugin has nothing to watch, and the timed re-read never runs.

I think the fix covers only the first read, though. In Quickshell's FileView (src/io/fileview.cpp, the
version on our phones), the watch is set up once, when the path or watchChanges is set. It adds the file and
its directory to a QFileSystemWatcher, and if either doesn't exist yet, that addPath just fails and is never
retried. reload() doesn't set the watch up again. So with this change the timer picks up the idle state when
the stack starts, sensor becomes non-null and the timer stops, and nothing is watching the file. When fprintd
then starts verifying at the lock, the bridge's "auth" is never read, and the glyph still doesn't appear. One
way out is to re-arm the watch once the file exists, for example watchChanges: root.sensor !== null on the
FileView. Changing watchChanges re-runs the watch setup, and your timer covers the time before that.

Two more things:

  • The plugin is on for every phone, including ones with no fingerprint stack (the 8 Pro, the 6 Pro, a 7 Pro with
    omamobile-fp masked). There sensor stays null forever, so the timer would re-read a missing file every 2
    seconds for the life of the shell. Could you limit it, for example to when the fingerprint stack is installed
    (the Setup menu tests /usr/lib/omamobile-fp/omamobile-fp-chroot), or to while the session is locked?
  • Could you add the phone you tested on, the uname -r and a tools/phone-smoke report, plus a run of the real
    case: a boot where the stack starts after the shell, then the lock with fprintd verifying, showing the glyph and
    omarchy-shell shell call omamobile.fingerprint state ''. Please reproduce it with a reboot or the harness's
    OMAMOBILE_FP_UI_STATE, not by stopping and starting omamobile-fp in the same boot. Trusty doesn't take that
    well.

One housekeeping note: #421 also takes shell 0.2.101, so whichever lands second will move to the next version.
I'll sort that out at merge.

…e only where the stack is installed

The plugin is on for every phone, and on one without the fingerprint stack the state file never appears, so the re-read every 2 s ran for the life of the shell. It now runs with no state only where omamobile-fp-chroot is installed, the same test the Setup menu's Fingerprint entry uses.
@W4HO-ham

W4HO-ham commented Nov 6, 2026

Copy link
Copy Markdown
Contributor Author

Thanks. On the watch: in the Quickshell the phones run (0.3.2.r7.g2b4501a), FileView::reload() is updatePath(), and that ends in updateWatchedFiles(). It deletes the watcher and sets it up again on the file and its directory (src/io/fileview.cpp, lines 379 and 479-491 at 2b4501a). So once the timer's reload finds the file, the watch is armed on it. I checked that on the same version on the laptop, offscreen. A FileView pointed at a path whose directory didn't exist yet was re-read every 300 ms while there was no state, as this plugin does. The directory and file appeared 5 s later, the timer stopped once the file loaded, and the next rename(2) replacement came through fileChanged with no more timed reloads. So I left watchChanges as it was.

The limit is in. With no state, the timer now runs only where the fingerprint stack is installed, tested the way the Setup menu tests it (test -x /usr/lib/omamobile-fp/omamobile-fp-chroot; OMAMOBILE_FP_RUNNER overrides it, as in onboarding). In the same offscreen check, with the runner present it re-read until the file appeared; without it, it never re-read.

The palmtop run you asked for still has to happen: a reboot where the stack starts after the shell, then the lock with fprintd verifying, the glyph and the state call, and a smoke report. The palmtop is in a suspend series now, so that comes after.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @W4HO-ham, and you're right about the watch. I went back to fileview.cpp at 2b4501a: reload() is updatePath(), which ends in updateWatchedFiles(), so the first reload that finds the file sets the watch up on it and its directory. I'd read updateWatchedFiles() and missed that reload() gets there. I'm sorry for the detour, and thanks for checking it on the real Quickshell rather than taking my word for it.

The new commit is just what I hoped for. The re-read with no state now runs only where the runner is installed, using the same test as the Setup menu and onboarding. A phone with the package but no running stack still checks every two seconds, but that's cheap and I'm happy with it.

The only thing left is the palmtop run you described: a reboot where the stack starts after the shell, the lock with fprintd verifying, the glyph, the state call, and a smoke report, with this head installed as a package. No rush while the suspend series runs.

On the version: several shell PRs are open at 0.2.101 right now, so I'll set the number and the title myself when I merge. You don't need to push anything for that.

…tack

# Conflicts:
#	packages/omamobile-shell/version
Main took 0.2.126 for another change; the number is the maintainer's to set at merge.
@derekross

Copy link
Copy Markdown
Owner

Thanks for keeping this merged up with main. I went through it again against today's tree, now that the phones are on Omarchy 4.5 and the shell is at 0.2.126:

  • It still merges cleanly, and the plugin on main differs from your branch only by your change. CI is green on this head.
  • Omarchy 4.5's new lock screen doesn't affect it. The glyph isn't part of the lock's files: it's our own layer, drawn above the lock by the above_lock rule in hypr/phone.lua, and none of our 4.5 lock patches reads the bridge's state file. So there's nothing to port.
  • The Quickshell that [omarchy] ships is still 0.3.2.r7.g2b4501a, the commit where we both read reload() → updatePath() → updateWatchedFiles(), so your reasoning about the watch holds for what the phones run.

So the code is where we left it on 11-07, and the one thing still open is the same: a run on a phone with this head installed as a package. A boot where the fingerprint stack starts after the shell, then the lock with fprintd verifying, the glyph, the output of omarchy-shell shell call omamobile.fingerprint state '', and a tools/phone-smoke report with the phone, uname -r and pacman -Q omamobile-shell quickshell omarchy. As before, please get the late start from a reboot (or the harness's OMAMOBILE_FP_UI_STATE), never by stopping and starting omamobile-fp in one boot. If the palmtop is still tied up in its sleep series, tell me and I'll run the same on my Pixel 7 Pro instead.

Two notes, nothing to do for either:

  • The version. The branch says 0.2.127 and another open pull request carries the same number, so I'll set it (and the title) when I merge. You don't need to keep merging main for that.
  • omamobile.home has a second FileView on the same state file, for its long-press rule, with the same gap. Your plugin's desktop guard sits above the home screen and will follow the state after this, so nothing is exposed; if you'd like to give it the same re-read in a follow-up, I'd take it.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The palmtop run, on this head installed as a package (8bc5910, built on the palmtop with tools/phone-install-package packages/omamobile-shell).

7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty
omamobile-shell 0.2.127-1, quickshell 0.3.2.r7.g2b4501a-1.1, omarchy 4.5.0-3, omamobile-fingerprint 0.1.17-1

The late start, from a reboot. The 9a still starts its fingerprint stack from the bring-up unit (tegu-fp-stage: the goodix-fp module, then omamobile-fp-chroot start bridge), not from omamobile-fp. I disabled that unit, rebooted, and started it once after the shell was up; nothing was stopped and started in one boot. The unit is enabled again.

MST what
07:29:11 boot
07:29:52 the shell answers; state: "op":"none", "spot":null, no /run/omamobile-fp-ui/state.json
07:30:01.81 systemctl start tegu-fp-stage
07:30:10.82 the bridge is ready and the state file is born
07:30:11.0 state (asked every 0.25 s): "op":"idle", "spot":{"x":202.5,"y":658.5,"r":37.875}

So the plugin had the sensor within a poll of the file appearing, 9 s after a start that came 50 s into the boot, with no shell restart.

The lock. The owner pressed Power to lock and again to wake. He saw the glyph about a quarter of a second after the press, put his finger on it and was let in. The plugin's state through it, asked every 0.25 s:

07:30:56.220 op auth, target true, guard true
07:30:56.666 the omamobile-fingerprint layer opens
07:30:58.329 finger true
07:30:58.759 op idle
07:30:59.581 op auth
07:31:00.821 op idle, target false, the layer closed (unlocked)

omarchy-shell shell call omamobile.fingerprint state '' after it:

{"op":"idle","finger":false,"screen":"DSI-1 405x909","screens":1,"spot":{"x":202.5,"y":658.5,"r":37.875},"transform":0,"target":false,"osk":false,"oskWatch":true,"lastLayerEvent":"closelayer omamobile-fingerprint","contact":false,"holdGuard":false,"inhibit":"/run/omamobile-fp-ui/inhibit/osk-mobile","guard":false,"enroll":"closed","authPending":false,"enrollWindow":false,"helper":false,"progress":0,"hint":"","fingers":[]}

One thing I didn't record: a line from fprintd itself for the verify. The auth operation and the unlock are what I have.

tools/phone-smoke, the same boot

phone-smoke: tegu

Kernel 7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty, tools at 1d4df112, 2026-11-12T14:31Z.

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty, slot b, up 140s
storage guard (required) ok 62 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 88G free
kernel health (required) ok no oops or BUG; 1 WARN, 209 error lines in dmesg (185 known, of them cpif message for a channel nobody opened x9, 24 new)
new dmesg errors not ok nci: nci_ntf_packet: unsupported ntf opcode N x21; pcieport N:N:N: of_irq_parse_pci: failed with rc=N x1; brcmfmac: brcmf_c_process_txcap_blob: TxCap blob found, loading x1
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card1-DSI-1 connected, 1080x2424
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked false)
touch ok Synaptics TouchComm Touchscreen
buttons ok gpio-keys
Wi-Fi ok wlan0 on brcmfmac, link up
Bluetooth ok hci0, powered true
battery ok 88% Unknown, 33 °C
sound cards ok googleteguaocsn
sensors ok Accelerometer AmbientLight Proximity
cameras ok 3 listed by libcamera
thermal ok hottest zone 48 °C

Packages: hyprland 0.57.0-3.3, libcamera 0.7.2-4.23, mesa 1:26.2.4-1.4, omacam 0.3.3-1, omamessages 0.2.2-1, omamobile-apps 0.2.2-1, omamobile-audio 0.5.18-1, omamobile-device-tegu 0.2.16-1, omamobile-fingerprint 0.1.17-1, omamobile-firstboot 0.1.4-1, omamobile-gnss 0.2.9-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-mmd 0.1.4-1, omamobile-modem 0.1.20-1, omamobile-modem-sit 0.3.37-1, omamobile-omarchy 0.1.43-1, omamobile-owl-widgets 0.1.2-1, omamobile-sensors 0.2.1-1, omamobile-shell 0.2.127-1, omamobile-telephony-tegu 0.1.0-1, omaphone 0.2.3-1, omarchy 4.5.0-3, omatel-agent 0.2.4-2, quickshell 0.3.2.r7.g2b4501a-1.1

On omamobile.home's second FileView: yes, I'll send the same re-read as a follow-up.

@derekross

Copy link
Copy Markdown
Owner

Thank you, that is exactly the run I asked for, and done the careful way: the late start came from a reboot with the bring-up unit held back, nothing was stopped and started in one boot, and the unit is enabled again.

It shows both halves. The plugin had the sensor within a poll of the state file appearing, 9 s after a start that came 50 s into the boot, with no shell restart. And at the lock, auth came through after the timer had already stopped, the layer opened, the owner saw the glyph and was let in. That second half is the part I had doubted on 11-05, so it's good to see it on a phone as well as in your laptop check. A line from fprintd itself would have been nice to have, but op auth followed by the unlock is enough for this.

The smoke report is clean and from the same boot, and the package was built from this head. Nothing more is needed from you here.

I'm merging it. The branch says 0.2.127, which main took yesterday with the Do Not Disturb change, so I'll push the next free number to your branch first and merge once CI has run on it; the title gets the same number.

Thanks also for offering the same re-read for omamobile.home's FileView. I'll take that as its own pull request whenever it suits you.

…tack

# Conflicts:
#	packages/omamobile-shell/version
…tack

# Conflicts:
#	packages/omamobile-shell/version
@W4HO-ham W4HO-ham changed the title omamobile-shell 0.2.101: the fingerprint glyph appears when the fingerprint stack starts after the shell omamobile-shell 0.2.132: the fingerprint glyph appears when the fingerprint stack starts after the shell Nov 15, 2026
@derekross

Copy link
Copy Markdown
Owner

Thank you for merging main again and setting 0.2.132 yourself: that is the next free number, one above main's 0.2.131, so the version is settled without my push. The plugin file is byte for byte the one from your Pixel 9a run, CI is green on this head and it merges cleanly, so I'm merging it now. Sorry it took two more days after I said I would.

@derekross
derekross merged commit b7cfc6a into derekross:main Nov 15, 2026
2 checks passed
angelrenefm added a commit to angelrenefm/omarchy-mobile that referenced this pull request Nov 15, 2026
The version numbers for this pull request. omamobile-shell takes 0.2.133
because derekross#425 took 0.2.132 on main.

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.

2 participants