Skip to content

Detect FocalTech match-on-chip fingerprint readers - #11387

Open
alanwcurry-dev wants to merge 4 commits into
omacom:quattrofrom
alanwcurry-dev:focaltech-fingerprint-detection
Open

alanwcurry-dev wants to merge 4 commits into
omacom:quattrofrom
alanwcurry-dev:focaltech-fingerprint-detection

Conversation

@alanwcurry-dev

@alanwcurry-dev alanwcurry-dev commented Sep 11, 2026 •

Copy link
Copy Markdown

Fixes #11384.

omarchy-hw-fingerprint misses FocalTech match-on-chip readers, so omarchy setup security fingerprint aborts with "No fingerprint sensor detected." on a laptop whose reader the pinned libfprint-git already supports (upstream lists 2808:a57a as "Focaltech MOC Sensors").

The reader reports FocalTech FT9365 ESS and misses both detection paths: 2808 is not in fingerprint_vendors, and the product string names no function, so none of the fingerprint / biometric / elan:arm-m4 / fpc patterns match it. The vendor ID is the only thing that can carry these readers, and the existing driver check is what keeps FocalTech's touch controllers out — they bind usbhid/i2c-hid, while a match-on-chip reader sits there driverless, exactly the shape has_kernel_driver already tests for.

Because the wizard bails before it installs anything, affected machines end up with no fprintd/libfprint at all, which makes it look like the sensor does not exist. The same detector also gates setup-fingerprint.hook, so those machines never get the post-update invitation either.

Testing

  • test/shell.d/hw-fingerprint-test.sh gains two cases: a driverless 2808:a57a device is detected by vendor, and the same device with usbhid bound is rejected.
  • On the affected laptop (ASUS ExpertBook PM3406CKA, 2808:a57a) the patched script exits 0 against real sysfs where it previously exited 1, and omarchy setup security fingerprint then runs end to end: enrollment, fprintd-verify match, PAM stacks written, and the fingerprint accepted by both sudo and polkit.
  • ./test/all here: test/shell.d/hw-fingerprint-test.sh passes. config-test.sh, snapper-test.sh and unowned-system-paths-test.sh fail identically at the parent commit on this machine because they want an omarchy-pkgs checkout that is not present — unrelated to this change.

Fixes #14089

@johnpippett johnpippett left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Source commit: 26745881380a7217d0dc35cf68de730b71df399e.
Comparison commit: 31bd80daa4613ffdee995ac27467fce5a2990806.

Codex did the isolated tests on 2026-09-12. All 40 independent cases gave the expected results.
The cases included missing device information, driver exclusions, multiple devices, and setup failures.
The upstream fingerprint suite gave the expected results for 11 base assertions and 13 head assertions.
Call records showed the expected package and enrollment steps.

The tests used the actual scripts with synthetic USB device files.
Test programs replaced the package, enrollment, and verification commands.
Bubblewrap used private home, process, runtime, and network namespaces. The source files did not permit writes.
I do not have this reader. These tests did not include physical enrollment or authentication.

I have a question about the device detection rule.
The description reports successful hardware tests for 2808:a57a and explains that FocalTech touch controllers have kernel drivers.
The new vendor rule accepts every device with vendor ID 2808 that has no kernel driver.

What device or driver documentation shows that all FocalTech devices without kernel drivers are fingerprint readers?
If the documentation describes only 2808:a57a, could the rule use that vendor and product pair instead?

This is a question about device coverage. Our tests did not show incorrect detection of a physical device.
They do not establish compatibility with every FocalTech product.

@omarchybot omarchybot added bug Something isn't working compatibility Make Omarchy work better on everything! labels Sep 27, 2026
@omarchybot

Copy link
Copy Markdown
Collaborator

Reviewed at 2674588, on a disposable Omarchy worker. The fix does what it says, and nothing in the diff needs changing.

What was checked

  • My own sysfs reproduction, separate from the PR's tests: a driverless 2808:a57a "FocalTech FT9365 ESS" device is not detected on quattro and is detected on this head, including while fprintd holds it through usbfs. A 2808 device bound to usbhid is rejected on both.
  • test/shell.d/hw-fingerprint-test.sh fails its new FocalTech case with the vendor addition reverted and passes 13/13 on the head. fingerprint-invitation-test.sh and fingerprint-package-test.sh pass, and so does ./test/cli.
  • The full shell suite ran 3229 passed, 2 failed. Neither failure is related to this PR: ascii-test.sh and branding-about-animation-test.sh fail the same way on quattro. locate-test.sh also failed in the full run, but only because an earlier test leaves a bin/__pycache__ behind; it passes once that is removed.
  • The libfprint-git that omarchy setup security fingerprint installs (1.94.100.r10.g6f9479c) lists 2808:a57a under its focaltech_moc driver, so the reader this detects is one the setup can actually drive.

One thing for the maintainer to weigh. That same package also lists several FocalTech readers it knows of but has no driver for (a658, 9338, 6652 and others). Any of them that sits driverless and does not name itself a fingerprint reader will now pass detection and get the setup invitation, and the setup will then find no device. Goodix's vendor entry already behaves the same way, and #8800's a658 was detected before this change through its product string. So this is the existing trade-off of matching by vendor, not something this PR introduces.

Second opinion: none. The configured reviewer (Codex Medium) answers, but today's review budget was spent before this item was reached, so this review is mine alone (Claude Opus 5.5).

What it waits on: this is hardware a worker does not have. The stubbed sysfs proves the detection logic, not the laptop. The only on-device confirmation so far comes from the PR's author, and that cannot double as independent confirmation. It needs a maintainer, or another owner of a 2808:a57a reader, to run omarchy-hw-fingerprint; echo $? on 2674588 and get 0. After that it needs a second-opinion review on that head. It is waiting on the maintainer.

@jddmgt

jddmgt commented Oct 6, 2026

Copy link
Copy Markdown

Tested on an ASUS ExpertBook B9406CAA with a FocalTech 2808:a97a ("FocalTech FT9349 ESS"). On 4.0.4, omarchy-hw-fingerprint exits 1. With this PR's version (2674588) it exits 0, and fprintd enrolls and verifies fine on this reader. 👍

@omarchybot

Copy link
Copy Markdown
Collaborator

Re-checked at 2674588 after @jddmgt's report, this time with a second opinion. Nothing in the diff needs changing.

What's new since the last review

  • @jddmgt's reader is a different model, 2808:a97a ("FocalTech FT9349 ESS"). I added it to my sysfs reproduction on a disposable Omarchy worker. On quattro it is not detected. On this head it is detected, and with the 2808 vendor entry reverted it is not detected again. The same holds for a57a, including while fprintd holds the reader through usbfs. A 2808 device bound to usbhid is rejected on all three. hw-fingerprint-test.sh (13/13), fingerprint-invitation-test.sh, fingerprint-package-test.sh and ./test/cli pass on the head. hw-fingerprint-test.sh fails its FocalTech case with the vendor entry reverted.
  • The libfprint-git the setup installs (1.94.100.r10.g6f9479c) carries udev entries for 22 FocalTech product IDs, a57a and a97a among them. That answers @johnpippett's question about matching only 2808:a57a: two distinct readers have now been seen working, and a product-pair match would miss the second one. The vendor entry follows the same pattern the list already uses for Goodix and Synaptics.
  • quattro has moved on since this branch was cut, but omarchy-hw-fingerprint has not changed there, and the branch still merges cleanly.

Second opinion (Codex Medium): found no defect in the change, and agreed that adding the vendor is the proportionate fix. It confirmed that the usbhid test really exercises the driver check: the fake interface sits exactly where has_kernel_driver looks for it. It also noted that a FocalTech reader libfprint cannot drive would now be detected and offered setup. In that case enrollment fails and exits before PAM is touched, so nothing is left half-configured. It read my earlier comment, so its agreement is not guaranteed to be independent.

What it waits on: the maintainer. This is hardware a worker does not have, and a confirmation on a real device counts toward readiness only when it comes from the issue's reporter or a maintainer. Here the reporter is also the PR's author, so @jddmgt's test is useful corroboration but does not settle it. If a maintainer runs omarchy-hw-fingerprint; echo $? on a FocalTech machine at 2674588 and gets 0, that is the remaining step.

@omarchybot

Copy link
Copy Markdown
Collaborator

Re-checked at 2674588. The diff is unchanged and still correct. What has changed is who can confirm it.

What's new. This pull request now also closes #14089, which @jddmgt opened. A reporter's confirmation on real hardware is what a fingerprint fix like this needs, since a worker has no FocalTech reader. That confirmation can come from the person who opened an issue the fix closes, but not from the pull request's author. So @jddmgt's test now counts, where the last review could only call it corroboration. There is one gap: the rule asks for the full commit the test ran on, and the earlier comment names 26745881.

@jddmgt, could you reply here with the result at the full commit 26745881380a7217d0dc35cf68de730b71df399e, naming that commit in full? If your earlier test was at that commit, saying so is enough. To run it again without installing anything:

curl -fsSL https://raw.githubusercontent.com/alanwcurry-dev/omarchy/26745881380a7217d0dc35cf68de730b71df399e/bin/omarchy-hw-fingerprint -o /tmp/omarchy-hw-fingerprint
bash /tmp/omarchy-hw-fingerprint; echo $?

0 means your 2808:a97a is detected.

What was checked this time. bin/omarchy-hw-fingerprint and its test have not changed on quattro since this branch was cut. The diff still applies exactly to quattro at da2db1a, and the pull request merges cleanly. Nothing was re-run on a worker, because the head is the one already reproduced there twice: not detected on quattro, detected on this head, not detected with the 2808 entry reverted, and usbhid-bound 2808 devices rejected.

Second opinion (Codex Medium): it read the diff against current quattro and found no defect. It agreed that adding the vendor is proportionate, and that matching by vendor can still let through a FocalTech device libfprint cannot drive, a limit of the existing approach rather than of this change. It also agreed that @jddmgt is eligible to confirm and that only the full commit is missing. It had the earlier comments in front of it, so its agreement is not guaranteed to be independent.

What it waits on: @jddmgt's confirmation at the full commit. A maintainer's confirmation on a FocalTech machine would also do.

@jddmgt

jddmgt commented Oct 18, 2026

Copy link
Copy Markdown

Confirmed at the full commit 26745881380a7217d0dc35cf68de730b71df399e: my earlier test was on this head. On an ASUS ExpertBook B9406CAA with a FocalTech 2808:a97a ("FocalTech FT9349 ESS"), omarchy-hw-fingerprint exits 1 on 4.0.4 and exits 0 at this commit, and fprintd enrolls and verifies fine on this reader.

@omarchybot omarchybot added verified Omarchy Triage has verified that this issue is ready for final review ready Good to merge labels Oct 18, 2026
@omarchybot

Copy link
Copy Markdown
Collaborator

Re-checked at 2674588 after @jddmgt's confirmation. The diff is unchanged, and it is now verified on real hardware.

The confirmation. @jddmgt opened #14089, which this pull request closes, and has confirmed at the full commit 26745881380a7217d0dc35cf68de730b71df399e that their FocalTech 2808:a97a reader is detected and that fprintd enrolls and verifies on it. That is the on-device evidence a worker cannot supply. The other half of #14089, unlock failing after resume, was fixed by #7158, so closing #14089 when this merges is correct.

What was checked this time, on a disposable Omarchy worker, with this head merged into current quattro (da2db1a):

  • My own sysfs reproduction: driverless 2808:a57a and 2808:a97a readers are detected, including a97a while fprintd holds it through usbfs. With the 2808 entry removed, which is quattro's version of the script, none of them is detected. A 2808 device with usbhid bound is rejected either way.
  • hw-fingerprint-test.sh (13/13), fingerprint-invitation-test.sh and fingerprint-package-test.sh pass, and so does ./test/cli.
  • The full shell suite failed 12 of 454 files, none of them about fingerprints. 11 fail the same way on plain quattro. The twelfth, background-intro-lifecycle-test.sh, passes and fails from run to run on both quattro and this branch.

Second opinion (Codex Medium): found no defect. It agreed the change is proportionate, and it confirmed the usbhid test really exercises the driver check. It also confirmed that a FocalTech device libfprint cannot drive, if one ever sits driverless, fails at enrollment before any PAM stack is touched. I checked that against bin/omarchy-setup-security-fingerprint. Its agreement is not guaranteed to be independent, since it could read the earlier comments here.

What it waits on: the maintainer. Nothing is left for the author.

@omarchybot omarchybot removed ready Good to merge verified Omarchy Triage has verified that this issue is ready for final review labels Oct 19, 2026
quattro's omacom#14675 added platform-named readers to omarchy-hw-fingerprint and its test. The script merged cleanly; the test conflicted only because both appended cases at the end, so the FocalTech cases now follow the platform ones, after the platform file is removed, and still exercise the USB scan alone.

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

Copy link
Copy Markdown
Collaborator

This branch stopped merging after quattro picked up #14675, which taught omarchy-hw-fingerprint to find readers a platform package names (a Mac's Touch ID). I merged quattro into the branch and pushed it as aad2b220af3ef64115e25dc2b77da452738ec668. The diff against quattro is still exactly this pull request's change: 2808 in fingerprint_vendors and the two FocalTech test cases. The script merged on its own. The test conflicted only because both sides added cases at the end, so the FocalTech cases now run after the platform ones, once the platform fixture file has been removed.

What was checked at aad2b220, on a disposable Omarchy worker:

  • My own sysfs reproduction: driverless 2808:a57a and 2808:a97a readers, including a97a while fprintd holds it through usbfs, are not detected by quattro's script and are detected at this head. A 2808 device with usbhid bound is rejected by both.
  • hw-fingerprint-test.sh passes 25/25. With the 2808 entry reverted, only its FocalTech case fails. fingerprint-invitation-test.sh, fingerprint-package-test.sh and ./test/cli pass.
  • The full shell suite had 8 failures, none of them about fingerprints. Six fail the same way on plain quattro (087f73d). lock-appear-budget-test.sh also fails on quattro when run alone. preinstalls-test.sh failed in the full run but passes alone on both.

Second opinion (Codex Medium): it read the merge and found no defect. It confirmed that nothing from either side was lost, that the test's copy of the script carries the 2808 entry, and that no leftover platform fixture can affect the FocalTech cases. It also found that the new platform check cannot change how a USB reader is detected. It could read this thread, so its agreement is not guaranteed to be independent.

What it waits on: a confirmation on real hardware at the new head. A merge commit is a new head, so @jddmgt's confirmation at 26745881 no longer covers it, even though the fingerprint logic it tested is unchanged. @jddmgt, could you run this and reply with the result, naming aad2b220af3ef64115e25dc2b77da452738ec668 in full?

curl -fsSL https://raw.githubusercontent.com/alanwcurry-dev/omarchy/aad2b220af3ef64115e25dc2b77da452738ec668/bin/omarchy-hw-fingerprint -o /tmp/omarchy-hw-fingerprint
bash /tmp/omarchy-hw-fingerprint; echo $?

0 means your 2808:a97a is detected. A maintainer's confirmation on a FocalTech machine would also do.

@greptile-apps

greptile-apps Bot commented Oct 21, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable issues were found.

Summary

Adds FocalTech fingerprint readers to hardware detection while keeping the existing check that excludes devices with bound kernel drivers.

  • The fingerprint detector recognizes driverless FocalTech readers.

No actionable issues found.

Reviews (2) · Last reviewed commit: "Drop the FocalTech rejection test the ex..." · Reviewed by Greptile

@jddmgt

jddmgt commented Oct 21, 2026

Copy link
Copy Markdown

Confirmed at the full commit aad2b220af3ef64115e25dc2b77da452738ec668. On an ASUS ExpertBook B9406CAA with a FocalTech 2808:a97a ("FocalTech FT9349 ESS"), omarchy-hw-fingerprint exits 0 at this commit and exits 1 with the script shipped in 4.0.5.

@omarchybot omarchybot added verified Omarchy Triage has verified that this issue is ready for final review ready Good to merge labels Oct 21, 2026
@omarchybot

Copy link
Copy Markdown
Collaborator

Thanks, @jddmgt. Your confirmation that omarchy-hw-fingerprint exits 0 at aad2b220af3ef64115e25dc2b77da452738ec668 on the ASUS ExpertBook B9406CAA (2808:a97a), and 1 with the 4.0.5 script, covers the merge head. That was the one thing this was waiting on.

Together with the worker run and the Codex Medium second opinion described in the comment above, both at this same head, it is now labelled verified and ready. Nothing is left for the author. It waits on the maintainer.

omarchybot and others added 2 commits November 18, 2026 14:21
FocalTech joins the same vendor list the file's existing driver-rejection test already exercises, and that block passed on the base branch.

Co-Authored-By: Opus 5.5 High <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@omarchybot omarchybot removed the ready Good to merge label Nov 18, 2026
@omarchybot

Copy link
Copy Markdown
Collaborator

I merged current quattro into this branch; it was 810 commits behind. I also dropped the second test block: FocalTech joins the same vendor list the file's existing driver-rejection test covers, and that block passed before this change.

What was checked, on an Omarchy worker, against the merged branch:

  • hw-fingerprint-test.sh passes, and fails against quattro's code.
  • ./test/cli passes, 136 checks.

Claude Opus 5.5 at high effort reviewed the changes as a second opinion; its findings are folded in. Its independence from my review isn't guaranteed. Greptile scores the new head 5/5 with no open findings.

This waits on the maintainer, starting with approving the workflow run: CI hasn't run on this head yet.

@omarchybot omarchybot added the ready Good to merge label Nov 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working compatibility Make Omarchy work better on everything! ready Good to merge verified Omarchy Triage has verified that this issue is ready for final review

Projects

None yet

4 participants