X11 key autorepeat detection

[Descolada]

Hello,
I'm trying to differentiate artificial vs physical keystrokes in X11 (as well as possible, I don't think 100% accuracy is possible), and I'm running into issues with key autorepeat. If I generate a synthetic key press then it starts auto-repeating, but if the user presses down the same key then I have no way to determine whether it's physical or just auto-repeat. X11 generates automatic key down-up events if a key is held, where the up event can be disabled with XkbSetDetectableAutoRepeat, but this requires the Display connection which libuiohook apparently doesn't expose. Another way would be to peek the next XEvent and see whether it's a key-up event for the same key and with the same timestamp, but this also can't be done without a Display connection. Any ideas?

[TolikPylypchuk]

Hi! Thanks for posting this issue!

I think it makes sense to enable XkbSetDetectableAutoRepeat for SharpHook so that the auto-repeat behavior aligns with other OSes. But I don't have any other ideas about how to help with this situation - I'm not really an X11 expert.

Will this change be enough for your use-case?

[Descolada]

I agree that XkbSetDetectableAutoRepeat should be the default, though for backwards compatibility it should be left disabled until the next major version release. For now it could be exposed as a property or method in SharpHook.Native. I'm not yet sure whether this would be enough for my use-case though, would need some testing. But could you also expose the Display connection pointer? This is not only for the auto-repeat situation, but I'm using XTestEvent and some other raw X11 calls in my project, and using a different Display can cause synchronization issues.
For example I've had to mostly rewrite the functionality of SimulateTextEntry so I can remap a key, press it down and leave it down for a duration of time (SimulateTextEntry just does down-up). Also SimulateTextEntry was for some reason not working in my setup (Linux Mint) - not sending any keys at all. Don't remember the exact cause of failure though, I think it was related to how long the remap was kept active?

[TolikPylypchuk]

Regarding XkbSetDetectableAutoRepeat - yes, I agree, I can add a switch for that.

Regarding exposing the Display connection - I'm not sure. SharpHook provides an abstraction layer over the low-level stuff, and while not perfect, I'm hesitant to make it even more leaky. Even libuiohook itself doesn't expose the Display connection (it actually uses 2 Display connections and even I don't know why it does so). If you need this much customization, then I think you'd be better off writing these customizations in C and then creating your own wrapper around them - though I understand that it would entail much more work.

Regarding SimulateTextEntry not working well - yes, it really doesn't work well, but the solution that it uses is the best I've come up with. I just don't think there is a better way to do it on Linux and I'm not sure I will be able to help there either.