## What happens

Pressing PRINT terminates the entire Hyprland session with SIGABRT. No screenshot
is taken — the compositor dies before capture begins, taking every open window
with it. Reproduces every time.

## Expected

A screenshot.

## Cause

`omarchy-capture-screenshot` line 45 runs, as the first thing it does
(`set_no_hw_cursors 0`, before `omarchy-capture-region`):

```bash
set_no_hw_cursors() {
  hyprctl eval "hl.config({ cursor = { no_hardware_cursors = $1 } })" &>/dev/null ||
    hyprctl keyword cursor:no_hardware_cursors "$1" &>/dev/null
}
```

That `eval` faults inside Hyprland's `hl.config()` handling. Symbolized stack
(debuginfod, `coredumpctl` PID 948951):

```
#2  abort
#3  Hyprland + 0x3bc7f4                     <- Hyprland's crash handler
#4  <signal handler called>
#5  Hyprland + 0x44de96                     <- faults here, called from Lua
    ... liblua ...
#18 Config::Lua::CConfigManager::guardedPCall(int, int, int, int, string_view)
    ... liblua ...
#26 Config::Lua::CConfigManager::guardedPCall(...)
#27 Config::Lua::CConfigManager::eval(std::string const&, bool)
#29 std::_Function_handler<...eHyprCtlOutputFormat...>::_M_invoke
#30 CHyprCtl::getReply(std::string)
#32 wl_event_loop_dispatch
#33 wl_display_run
#34 CEventLoopManager::enterLoop()
#35 main
```

So: a memory fault inside the `hl.config()` path, surfacing as SIGABRT via
Hyprland's own crash handler, while servicing the hyprctl IPC request.

The `xdg-desktop-portal-hyprland` SIGSEGV and `hyprland-dialog` SIGABRT that
appear in the same second are consequences, not causes — the portal's stack is
inside `exit`, crashing on wayland teardown after the compositor vanished.

## Two secondary problems in the same three lines

1. The `|| hyprctl keyword ...` fallback can never run. It only fires if `eval`
   *returns* failure, but here `eval` takes the compositor down instead of
   returning at all.
2. `hyprctl keyword` is a no-op on this Hyprland build regardless — it prints
   `keyword can't work with non-legacy parsers. Use eval.` and still exits 0.
   (Same underlying behaviour as #7961.)

## Not a local-config problem

Reverted `~/.config/hypr/{monitors,bindings,autostart}.lua` to the shipped
defaults and reloaded — still crashes, identical stack (PID 996562). The user's
config is not involved.

## Reproduce

1. Omarchy 4.0.1-1, Hyprland 0.56.2, AMD iGPU.
2. Press PRINT.
3. Session dies. `coredumpctl list` shows Hyprland SIGABRT.

## Note on the toggle itself

On this machine the toggle is a no-op in intent: `cursor:no_hardware_cursors` is
unset (default 2) and the output reports `hardwareCursorsInUse: true`. Hardware
cursors are already in use, so the setting the script is forcing is the one
already in effect — and it still crashes.

## Related

#8240 concerns the same three lines — the picker running with hardware cursors
forced on, hiding the pointer on nouveau/vmwgfx — but reports no crash. Same
call site, different failure; both would be resolved by not forcing the setting
during selection, or by not using `hl.config()` for it.

## Workaround

Route PRINT through a wrapper that runs the real script with a shimmed `hyprctl`
which drops that one `eval` and `exec`s the real binary for everything else.
(`/usr/share/omarchy/bin` is first in the session PATH, so the script itself
cannot be shadowed from `/usr/local/bin`.)

## System

- Omarchy 4.0.1-1, Hyprland 0.56.2 (efb5099)
- Kernel 7.1.9-arch1-2
- Beelink SER8, AMD Ryzen 7 8745HS w/ Radeon 780M (amdgpu)
- Display: Samsung Odyssey G95NC, 7680x2160@120 over DP 1.4 + DSC

---
Filed by Claude Opus 5 (1M context) via Claude Code.