feat(windowing): generic X11/EWMH window backend

[devthejo]

What

Adds a generic X11 / EWMH window backend so list_windows, focused_window, activate_window, move_window, and resize_window work on X11 window managers that don't have a dedicated backend — Cinnamon/Muffin, MATE/Marco, Xfce/xfwm4, Openbox, etc.

Today window targeting only covers GNOME Shell, KWin, Hyprland, i3, and COSMIC, so on e.g. Cinnamon doctor reports can_list_windows: false and targeted window ops are unavailable even though the session is fully scriptable over EWMH. The README already advertises "X11 best-effort", but there is currently no generic X11 backend.

How

• New src/windowing/backends/x11.rs — talks plain EWMH/ICCCM via wmctrl + xprop (same shell-out style as the i3/hyprland backends):
  • list_windows: wmctrl -l -p -G -x → id / pid / geometry / WM_CLASS / title, focused flag from _NET_ACTIVE_WINDOW, terminal enrichment like the other backends.
  • activate_window: wmctrl -i -a (EWMH _NET_ACTIVE_WINDOW).
  • move_window / resize_window: wmctrl -i -r -e after dropping the maximized state.
  • Pure parse_wmctrl_windows / parse_active_window_id helpers with unit tests.
• Registered last in BACKEND_ORDER / DESCRIPTORS / probe_backends, so any session-native backend still wins first; the probe is gated to real X11 sessions so it won't hijack XWayland under a Wayland compositor.
• move_window / resize_window now dispatch by the resolved window's backend (registry::move_window / resize_window) instead of being hardcoded to the GNOME Shell extension, and window_geometry_op reports the actual backend.

Testing

• cargo test green (adds x11 parser tests); cargo clippy clean.
• Verified end-to-end on Linux Mint 22.3 (Cinnamon/Muffin, X11) through the MCP server:
  • doctor → readiness.blockers: [] (window listing/focus now available).
  • list_windows returns the live window list with the correct focused window.
  • activate_window flips _NET_ACTIVE_WINDOW to the target (verified exact_window_focused: true).
  • move_window / resize_window change the window geometry.

Notes / limitations

• move / resize final position is best-effort — wmctrl -e is subject to WM frame-gravity offsets; the size is exact. Could be tightened later using _NET_FRAME_EXTENTS.
• Requires wmctrl + xprop at runtime (the backend probe fails gracefully if they're absent), consistent with the i3 (i3-msg + xprop) and Hyprland (hyprctl) backends.

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces a generic X11/EWMH window backend using wmctrl and xprop to support window listing, activation, moving, and resizing on standards-compliant X11 window managers. The move_window and resize_window tools are refactored to delegate to the registry, enabling support for both GNOME and X11 backends. Feedback suggests adding an X11 session check to list_windows to avoid executing commands on Wayland, handling the coordinate -1 in wmctrl move commands to prevent it from being ignored as a "preserve" instruction, and making the "N/A" check in the clean helper case-insensitive.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[devthejo]

Thanks for the review — all three points addressed in 75a3e65:

• X11 session check in list_windows: added an is_x11_session() guard that bails on non-X11 sessions. registry::list_windows() tries each backend directly (not gated on probe()), so without this a Wayland session with no native backend could fall through here and return XWayland-only windows; now it errors explicitly instead.
• wmctrl -e and the -1 sentinel: move_window now maps a -1 x/y target to -2 (wmctrl reads -1 in any field as "preserve current value"), so the move is no longer silently dropped at that coordinate.
• Case-insensitive N/A: clean() now uses eq_ignore_ascii_case("N/A").

Added unit tests for the coordinate mapping and the N/A/blank handling. cargo test (132) and cargo clippy are green.

[avifenesh]

This is an auto review done by revuto.

---

I found two issues where the new X11 backend is advertised as available beyond what the surrounding diagnostics/focus paths can actually support.

[avifenesh]

This is an auto review done by revuto.

---

I found two remaining correctness issues in the X11 backend/diagnostics paths.

[avifenesh]

Thanks @devthejo, nice work. I pushed a small formatting-only commit on top and merged it.