Keyboard layout switching not working on Ubuntu 24.04

[kpushkaryov]

Describe the bug
I have US and RU keyboard layouts set up on client and server side, switched by Caps Lock shortcut. Switching has stopped working after I've upgraded xpra from 3.1.5+dfsg1-0.2ubuntu1 (Ubuntu package) to 6.2.2-r0-1 (xpra package).

Both client and server are on Ubuntu 24.04 with Xorg.

The root cause seems to be a bug of inverted condition at https://github.com/Xpra-org/xpra/blame/master/xpra/x11/server/core.py#L249 introduced in 84a5a81, which causes skipping of the X11Keyboard.set_layout_group(grp) call.

I've fixed the condition on my server and that solved the issue.

To Reproduce
Steps to reproduce the behavior:

1. xpra start :10 --daemon=no -d keyboard
2. xpra attach ssh://.../10 --start=xterm
3. Press layout switching shortcut.
4. Type Cyrillic letters.

System Information (please complete the following information):

• Server OS: Ubuntu 24.04
• Client OS: Ubuntu 24.04
• Xpra Server Version 6.2.2-r0
• Xpra Client Version 6.2.2-r0

Additional context
I've enabled keyboard debug logging on the server. I see that Cyrillic letters are actually received by the server:

2025-01-08 12:49:41,604 handle_key((1, True, 'Cyrillic_shorti', 1738, 24, ['mod2'], False, True))
2025-01-08 12:49:41,605 handle keycode pressing    24: key 'Cyrillic_shorti'
2025-01-08 12:49:41,605 fake_key(24, True)
2025-01-08 12:49:41,770 process_key_action(('key-action', 1, 'Cyrillic_shorti', False, ('mod2',), 1738, 'й', 24, 1)) server keycode=24, group=1
2025-01-08 12:49:41,770 set_keyboard_layout_group(1) ignored, value unchanged
2025-01-08 12:49:41,771 filtered_modifiers_set(['mod2'])={'mod2'}
2025-01-08 12:49:41,771 filtered_modifiers_set(['mod2'])={'mod2'}
2025-01-08 12:49:41,771 is_modifier(24) not found
2025-01-08 12:49:41,772 handle_key((1, False, 'Cyrillic_shorti', 1738, 24, ['mod2'], False, True))
2025-01-08 12:49:41,772 handle keycode unpressing  24: key 'Cyrillic_shorti'
2025-01-08 12:49:41,772 fake_key(24, False)

But Latin letters are printed anyways.

My problem is similar to #4445, #3857 (comment).

[totaam]

Thanks!

I never use this feature and I have no idea how I would test it.

This fix is so obviously correct that I have already backported it to older branches, and partially pushed.