Proof Portal

Project overview

Omarchy

ProbeLabs65 findings · 95 requirements

A proof layer — requirements, tests and verified fixes — for two of Omarchy's subsystems: the application menu (launcher scripts, QML model, JSONC config, search and selection) and the lock screen (lock scripts, QML, PAM authentication). Scope is deliberately limited to those components of omacom/omarchy; the rest of the distribution is not covered.

Inbox
ReportR-56GitHub issuePublicDecided· Out of scope

A throwing plugin close() leaves the panel overlay mapped and locks all pointer input

Reported by St1aan · filed Sep 27, 2026 · open at source

Decided: Out of scope — decided by agents · verified 3 days ago.

At a glance

1 · What is claimed?
Only the report text below — no validator has restated it as checkable claims yet.
2 · What does it affect?
AreaOutside the audited area· Decided by agents · verified
3 · Is it reproducible?
No validation has been run yet.
4 · What remains unclear?
  • Not validated yet: no claims, no reproducer.
5 · What happens next?
Decided by agents and verified by a second, independent agent — see the decision.

Original report

A throwing plugin close() leaves the panel overlay mapped and locks all pointer inputrevision 1 · captured 2026-09-28 21:23 UTC via poll

Summary

If a bar-widget panel's close() throws, the shell is left with a full-screen layer-shell overlay that is still mapped and still owns a full-screen input region. Every click on the desktop lands on that surface and does nothing, the keyboard dismissal path is wedged the same way, and no IPC call can recover it. The only way out is restarting the shell — which a user whose desktop just stopped responding to clicks has no obvious way to discover.

A bug anywhere in one plugin's close() therefore escalates into a desktop-wide input lockout.

Why

Ui/KeyboardPanel.qml:69:

function close() {
  if (owner && "close" in owner) owner.close()
  else root.open = false
}

The delegation is unguarded. If owner.close() throws, root.open is never set to false, and root.open is what everything else keys off:

  • Ui/KeyboardPanel.qml:81 — visible: open || card.opacity > 0 || popoutSwitching, so the window stays mapped.
  • Ui/KeyboardPanel.qml:124 — mask: Region { width: root.screenW; height: root.screenH }, so the mapped surface keeps hit-testing the whole output. Input is captured by the compositor-side region regardless of whether the QML MouseArea is enabled, so there is no click-through.
  • Ui/KeyboardPanel.qml:331 — the outside-click handler calls root.close(), which re-enters the same throwing path on every subsequent click.

Every other dismissal route funnels into the same function:

  • Esc / key handling: the panel's PanelKeyCatcher.onCloseRequested calls the same owner close().
  • omarchy-shell <target> close and omarchy-shell shell hide <id>: shell.qml:1192 hide() calls bar.hideBarWidget(id), which calls the widget's close().

So the surface that is blocking input can only be unmapped by code that is already broken.

Steps to reproduce

  1. omarchy plugin clone omarchy.clock
  2. In the clone's Panel.qml, make close() throw on its first line:
   function close() {
     throw new Error("boom")
     ...
   }
  1. omarchy restart shell
  2. Click the clock in the bar to open the calendar.
  3. Click anywhere else on the desktop.

Observed: nothing is clickable anywhere on any output. hyprctl layers shows the overlay still mapped at full size:

Layer level 3 (overlay):
    Layer 6553f5a77ab0: xywh: 0 0 3840 1080, a: 1, namespace: omarchy-keyboard-panel

omarchy-shell omarchy.clock close returns successfully and the layer stays mapped. Only omarchy restart shell clears it.

Expected: a throwing plugin degrades to "this panel misbehaves", not "the desktop stops accepting input".

How I hit this for real

Not hypothetical — this is how I found it. A plugin cloned from omarchy.clock before an update kept the old line

root.bar.centerHoverRevealSuppressed = value

which the bar has since turned into a read-only property with a setter function. The clone's close() calls that helper on its first line, so every close threw before reaching controller.hide():

WARN scene: file:///home/<user>/.config/omarchy/plugins/<id>.clock/Panel.qml[119:-1]:
TypeError: Cannot assign to read-only property "centerHoverRevealSuppressed"

Nothing on screen indicated a plugin error; the desktop simply stopped responding to clicks whenever the calendar was opened. Since omarchy plugin clone and omarchy plugin add <git-url> are both supported user-facing workflows, plugin code drifting out of sync with the shell API is an expected condition rather than an edge case.

Suggested fix

Guard the delegation so a throwing owner cannot hold the surface open:

function close() {
  try {
    if (owner && "close" in owner) {
      owner.close()
      return
    }
  } catch (e) {
    console.warn("KeyboardPanel: owner close() threw, forcing close:", e)
  }
  root.open = false
}

Worth considering more broadly, since the same shape appears elsewhere: any unguarded call into plugin code from a path that owns compositor-level input (open(), closeForPopoutSwitch(), the popout coordinator) can wedge the shell the same way. A plugin exception should never be able to leave an input-grabbing surface mapped.

System

Omarchy4.0.4-1
Kernel7.2.5-3-omarchy
CPUIntel Core i7-6820HQ
GPUIntel HD Graphics 530 (Skylake-H GT2)
Hyprland0.56.2
Quickshell0.3.1 (Arch)
Display3840x1080 @ 60Hz, scale 1

Related: #13381 (also in the clock panel, unrelated cause).

View raw text
## Summary

If a bar-widget panel's `close()` throws, the shell is left with a full-screen layer-shell overlay that is still mapped and still owns a full-screen input region. Every click on the desktop lands on that surface and does nothing, the keyboard dismissal path is wedged the same way, and no IPC call can recover it. The only way out is restarting the shell — which a user whose desktop just stopped responding to clicks has no obvious way to discover.

A bug anywhere in one plugin's `close()` therefore escalates into a desktop-wide input lockout.

## Why

`Ui/KeyboardPanel.qml:69`:

```qml
function close() {
  if (owner && "close" in owner) owner.close()
  else root.open = false
}
```

The delegation is unguarded. If `owner.close()` throws, `root.open` is never set to `false`, and `root.open` is what everything else keys off:

- `Ui/KeyboardPanel.qml:81` — `visible: open || card.opacity > 0 || popoutSwitching`, so the window stays mapped.
- `Ui/KeyboardPanel.qml:124` — `mask: Region { width: root.screenW; height: root.screenH }`, so the mapped surface keeps hit-testing the whole output. Input is captured by the compositor-side region regardless of whether the QML `MouseArea` is enabled, so there is no click-through.
- `Ui/KeyboardPanel.qml:331` — the outside-click handler calls `root.close()`, which re-enters the same throwing path on every subsequent click.

Every other dismissal route funnels into the same function:

- Esc / key handling: the panel's `PanelKeyCatcher.onCloseRequested` calls the same owner `close()`.
- `omarchy-shell <target> close` and `omarchy-shell shell hide <id>`: `shell.qml:1192` `hide()` calls `bar.hideBarWidget(id)`, which calls the widget's `close()`.

So the surface that is blocking input can only be unmapped by code that is already broken.

## Steps to reproduce

1. `omarchy plugin clone omarchy.clock`
2. In the clone's `Panel.qml`, make `close()` throw on its first line:
   ```qml
   function close() {
     throw new Error("boom")
     ...
   }
   ```
3. `omarchy restart shell`
4. Click the clock in the bar to open the calendar.
5. Click anywhere else on the desktop.

Observed: nothing is clickable anywhere on any output. `hyprctl layers` shows the overlay still mapped at full size:

```
Layer level 3 (overlay):
    Layer 6553f5a77ab0: xywh: 0 0 3840 1080, a: 1, namespace: omarchy-keyboard-panel
```

`omarchy-shell omarchy.clock close` returns successfully and the layer stays mapped. Only `omarchy restart shell` clears it.

Expected: a throwing plugin degrades to "this panel misbehaves", not "the desktop stops accepting input".

## How I hit this for real

Not hypothetical — this is how I found it. A plugin cloned from `omarchy.clock` before an update kept the old line

```qml
root.bar.centerHoverRevealSuppressed = value
```

which the bar has since turned into a read-only property with a setter function. The clone's `close()` calls that helper on its first line, so every close threw before reaching `controller.hide()`:

```
WARN scene: file:///home/<user>/.config/omarchy/plugins/<id>.clock/Panel.qml[119:-1]:
TypeError: Cannot assign to read-only property "centerHoverRevealSuppressed"
```

Nothing on screen indicated a plugin error; the desktop simply stopped responding to clicks whenever the calendar was opened. Since `omarchy plugin clone` and `omarchy plugin add <git-url>` are both supported user-facing workflows, plugin code drifting out of sync with the shell API is an expected condition rather than an edge case.

## Suggested fix

Guard the delegation so a throwing owner cannot hold the surface open:

```qml
function close() {
  try {
    if (owner && "close" in owner) {
      owner.close()
      return
    }
  } catch (e) {
    console.warn("KeyboardPanel: owner close() threw, forcing close:", e)
  }
  root.open = false
}
```

Worth considering more broadly, since the same shape appears elsewhere: any unguarded call into plugin code from a path that owns compositor-level input (`open()`, `closeForPopoutSwitch()`, the popout coordinator) can wedge the shell the same way. A plugin exception should never be able to leave an input-grabbing surface mapped.

## System

| | |
|---|---|
| Omarchy | 4.0.4-1 |
| Kernel | 7.2.5-3-omarchy |
| CPU | Intel Core i7-6820HQ |
| GPU | Intel HD Graphics 530 (Skylake-H GT2) |
| Hyprland | 0.56.2 |
| Quickshell | 0.3.1 (Arch) |
| Display | 3840x1080 @ 60Hz, scale 1 |

Related: #13381 (also in the clock panel, unrelated cause).

Validation

Not checked yet. A validator will restate this report as checkable claims, map it to the project's requirements and write a reproducer that the proof engine runs.

Linked

  • Related PRpr/11080· confirmed

Decision

Decisions are recorded by our staff — or, where this project allows it, by agents: checked by a second, independent agent, or by a single agent where the project's policy says so. Any decision reached so far is shown below.

  • Out of scopeDecided by agents · verified · 2026-10-09 19:01 UTC