The GNOME backend has shipped since v0.1.6 (libei input half) and v0.3.x (Shell extension for window ops and pointer position), but I've never run it on a real GNOME session because I'm on Hyprland and there's no GNOME Shell on this machine to dogfood against. Until someone does, "I think it works" is the strongest claim I can make about it.
The verification doc is here: docs/verification/gnome-shell.md.
It's about 30-45 minutes if everything passes. The setup adds one step beyond the KDE matrix: copy packaging/gnome-extension/wdotool@wdotool.github.io/ into ~/.local/share/gnome-shell/extensions/, log out and back in, then gnome-extensions enable wdotool@wdotool.github.io. Without the extension the backend silently falls back to bare libei (input only), which is a useful failure mode to confirm but not what I'm hoping you test.
Then it's thirteen operations across six conditions: default, fractional 125%, fractional 175%, mixed-scale dual-monitor, Wayland session restart, IBus active. Plus three special tests for portal token revoke + recovery, extension disable + re-enable mid-session, and cross-workspace activation (raising a window on a different workspace).
To help: fork the repo, edit the doc with PASS / FAIL / N/A in each cell, write a one-line repro under the table for any FAIL, open a PR referencing this issue.
The bug classes I'm most curious about: whether the RemoteDesktop portal prompt fires once and persists or every invocation (kills scriptability if the latter), whether the extension's D-Bus surface comes back cleanly after a disable/enable cycle without a Shell restart, whether MetaWindow.get_stable_sequence() IDs survive rapid window churn, and whether windowactivate correctly switches workspaces when the target is on a different one.
If the extension installation itself is the part that breaks, that's the most useful failure I could hear about.
For background: input is libei via the GNOME RemoteDesktop portal; window ops go through a small Shell extension exposing org.wdotool.GnomeShellBridge on the session bus. Extension targets Shell 45-48. Source: wdotool-core/src/backend/gnome.rs and packaging/gnome-extension/.
The GNOME backend has shipped since v0.1.6 (libei input half) and v0.3.x (Shell extension for window ops and pointer position), but I've never run it on a real GNOME session because I'm on Hyprland and there's no GNOME Shell on this machine to dogfood against. Until someone does, "I think it works" is the strongest claim I can make about it.
The verification doc is here: docs/verification/gnome-shell.md.
It's about 30-45 minutes if everything passes. The setup adds one step beyond the KDE matrix: copy
packaging/gnome-extension/wdotool@wdotool.github.io/into~/.local/share/gnome-shell/extensions/, log out and back in, thengnome-extensions enable wdotool@wdotool.github.io. Without the extension the backend silently falls back to bare libei (input only), which is a useful failure mode to confirm but not what I'm hoping you test.Then it's thirteen operations across six conditions: default, fractional 125%, fractional 175%, mixed-scale dual-monitor, Wayland session restart, IBus active. Plus three special tests for portal token revoke + recovery, extension disable + re-enable mid-session, and cross-workspace activation (raising a window on a different workspace).
To help: fork the repo, edit the doc with PASS / FAIL / N/A in each cell, write a one-line repro under the table for any FAIL, open a PR referencing this issue.
The bug classes I'm most curious about: whether the RemoteDesktop portal prompt fires once and persists or every invocation (kills scriptability if the latter), whether the extension's D-Bus surface comes back cleanly after a disable/enable cycle without a Shell restart, whether
MetaWindow.get_stable_sequence()IDs survive rapid window churn, and whetherwindowactivatecorrectly switches workspaces when the target is on a different one.If the extension installation itself is the part that breaks, that's the most useful failure I could hear about.
For background: input is libei via the GNOME RemoteDesktop portal; window ops go through a small Shell extension exposing
org.wdotool.GnomeShellBridgeon the session bus. Extension targets Shell 45-48. Source: wdotool-core/src/backend/gnome.rs and packaging/gnome-extension/.