Verify a desktop installation
For contributors and package maintainers: use this checklist to verify a release on a real desktop. It covers login, input and visible behaviour that headless tests cannot prove. Ordinary installation ends with the steps in the install guide.
1. Log in
Use the package guide or complete source session registration. Choose Gnoblin at the login screen.
Record the build you installed, distribution, GPU, displays and scale factors.
2. Check your shell
Verify the bar, dock, launcher and notifications appear and respond to input. Open and close menus near each screen edge. Test every display.
If the shell fails, confirm desktop right-click opens a terminal and the recovery panel appears. Check the shell's log before restarting it.
3. Check windows and keys
Test focus, typing, minimise/restore, maximise, fullscreen, drag and resize. Check app switching, launcher shortcuts and any custom bindings.
For SSD windows, check all buttons, titlebar drag and edge resize. A correct screenshot alone does not prove input works.
4. Reload
Change one visible config setting, then:
gnoblinctl config reload
gnoblinctl reloadConfirm the setting changed and existing windows survived. The second command also reloads theme and user scripts.
5. Lock and unlock
Confirm the lock screen receives input, desktop controls are unavailable while locked, and the shell recovers after unlock.
6. Screen sharing and remote input
With the patched portal backend installed:
- Under the default policy, start a share and complete the source picker.
- Confirm the selected monitor is the one actually shared.
- Test an explicit
askrule, then adenyrule. - For an
allowrule, test both permitted and broader device requests. - Stop sharing and confirm the stream ends.
See portal permissions for current permission policy.
7. Return to GNOME
Log out and select GNOME. Confirm its desktop still works. For package-release testing, also remove Gnoblin using its documented uninstall path and check that the Gnoblin login entry disappears.
Record failures separately from skipped checks. Include the exact commands, logs and visible outcome; a successful build is not a successful desktop test.