fix: client to server modifier press/release - #9056

Looks like the mac test had a random failure?

The key press tests are horrible and yeah it random fails on mac os sometimes.

This fix should be applied to the other screen types as well

I don't really know how this would affect the other types, so I left them no-op

@nbolton Thoughts on how we can use this on the other screen types ?

key press tests are horrible

Agreed.

@nbolton Thoughts on how we can use this on the other screen types ?

Edit: Could you elaborate?

key press tests are horrible

Agreed.

@nbolton Thoughts on how we can use this on the other screen types ?

Edit: Could you elaborate?

On other screen types i know we have some issues with stuck modifiers

screen types

You mean other OS or just Xorg?

Edit: It's been a long time, so that's paged out for me... I'd need to have a good dig to see if the same idea can be applied to other platforms. This rings a bell for me though so either I tried to solve it before or the fix is there somewhere but isn't working reliably.

all oses could have this issues ..since this is now part of the base class we could use this method if we need to

Pressing a modifier on the client and moving to the server and releasing it breaks the modifier on the client.

Can confirm this bug and that this PR fixes it.

Edit: Great PR, thanks.

Steps in other words:

1. Move mouse from server to client
2. Press and hold shift key (keep held)
3. Move mouse back to server
4. Release shift key
5. Move mouse back to client
6. Type 'a'

Expect: Lower case 'a'
Actual: Upper case 'A' (shift is stuck on)

all oses could have this issues ..since this is now part of the base class we could use this method if we need to

Edit: I'll have a poke and a think.

Edit: Hmm, well on macOS we use GetCurrentKeyModifiers (OS API) to get the active modifiers rather than maintaining a state like we do in our EI code. On Windows, I think we use KeyState::m_keys as the state, so maybe there's something there to look at?