Repository navigation
Conversation
[Medium impact] The changes appear safe to merge; no actionable issues were found. SummaryThe lock now uses compositor input to resume fingerprint retries after the displays blank, including keys that never reach the lock view.
No actionable issues found. Reviews (1) · Last reviewed commit: "Resume lock fingerprint on compositor in..." · Reviewed by Greptile |
|
GPT-6 in Codex (reasoning effort not exposed) read the diff, description and existing review, and checked the base branch's shell guidance. I classify this as a bug fix, but the configured second classifier, deepseek-v4p1-flash, returned no usable classification, so the classification remains unsettled. Worker tests and full verification have not run, and no labels or commits were added. This is waiting for classification to be resolved before verification can continue. |
scottjones
left a comment
There was a problem hiding this comment.
Thanks, this works, and it fixes something I hit this morning. Pressing the Touch ID / power key on a blanked lock now restarts the fingerprint reader in about half a second.
I tested on an M4 Pro (J616s) with Hyprland 0.57.2 and quickshell 0.3.2.r7, with this Service.qml bind-mounted over the one from the installed omarchy-dev build (same file as this PR's base).
Before the PR, the first fingerprint attempt timed out, and pressing the key lit the panel but started nothing. The reader only came back when I typed into the password field, about 35 s later. The key is a locked Hyprland bind, so the lock view never sees it.
With the PR, I let the first attempt time out at 08:19:57, then pressed the key at 08:22:22.99. The fingerprint session started at 08:22:23.55 (+0.56 s), and my finger unlocked at 08:22:25.03.
The wake still goes through resumeFingerprint(), so a struggling reader keeps its backoff. Holding the blank until the input monitor goes idle makes sense too, since otherwise the next press would never register as a wake.
Two non-blocking notes:
- The blank now waits for 1 s with no input from any device on the seat, not just the lock surface. A mouse with sensor jitter keeps a locked screen lit.
mouse_move_enables_dpmsalready re-lit the panel for pointer noise, so the real difference is small, but it's worth a line in the description. - Both tests stub the monitor, and the new Node test fails before this PR only because
IdleMonitoris missing from the source. Nothing exercises a real compositor event, so the hardware run above is the only proof that a locked bind's key press counts as activity.
Separate from this PR: on Macs the Touch ID sensor is the power key, so the press that wakes the lock also opens the system menu, which is waiting on screen after unlock. Laptops with the reader in the power button get the same. A small follow-up drops locked = true from the power-menu bind in default/hypr/bindings/utilities.lua: #19439
What changes
The lock pauses new fingerprint attempts when it blanks the displays, as described in #15294. Hyprland can wake the display for a locked binding, such as the power or brightness key, without passing the event to the lock view. The display comes back on, but the fingerprint reader waits for another key or pointer event.
The lock service now watches compositor input while locked, even when an idle inhibitor is active. When input arrives after the display blanked, it uses the existing wake path to resume fingerprint retries with the existing backoff policy. When the display is already lit, compositor input rearms the blank timer without starting another fingerprint attempt. The lock waits for a quiet input window before blanking so that the next input can trigger a wake, and keeps the current countdown interval, including the longer post-suspend interval.
Related work includes #10130 and #18595 for reblanking after compositor input, #7471 and #8930 for keyboard wake and focus recovery, #17143 for fingerprint recovery when the lid opens, #17205 for sensor touch feedback and display wake, and #16840 for opt-in verification with the display blanked.
Verification
The original report was investigated on an M3 MacBook Air running Omarchy. During a recorded idle-lock reproduction, the first key lit the display but left fingerprint verification paused. A subsequent Shift press restarted verification about 250 ms later, and a finger unlocked the session. The investigation also confirmed that the Touch ID button reports as the power key and that Hyprland consumes its locked binding. These observations were on the existing installation, before applying this patch.
After the rebase, the focused lock tests on an M4 Mac mini running macOS passed 82 assertions. The service test exercised the QML handlers with simulated timer and process boundaries, including compositor-only wake at the existing 250 ms retry interval, quiet-input blanking, and preservation of the post-suspend countdown. Bash syntax, ShellCheck, and Git whitespace checks passed. Four Wayland-dependent checks skipped, including the real Quickshell service fixture. The rebased patch has not been retested on Linux/Wayland or with the physical Touch ID button.
Screenshots
The lock screen keeps the same layout and controls. The change affects when the existing fingerprint flow resumes after input. There is no recording of the rebased behavior on hardware.
Opened by GPT-6.2 Sol on @satchlj's behalf.