Kirill Chibisov

It also means that once you re-enable it you adjust them, so if you have different caps based on the field you type in in your application, you will have...

>There is a complication in that the protocol must receive some value in the first call. I assumed that if a client advertises a capability, then it will always use...

>The initial state contains the supported capabilities. The enable event is set immediately. If set_state() is called with a capability that wasn't advertised, the user gets a runtime error. I...

I was thinking by the time of setting to get a runtime error/assert if user supplied something that they haven't enabled, but IME won't get enabled, since it's more of...

I mean, winit should just cache the entire state and send it in that case when it can. Like there's nothing wrong with caching the entire thing in the backend,...

Could you look into the changes I've sent? I've squashed + rebased your PR before applying my changes afterwards separately. Note: I'll update docs afterwards. I'm in favor of e.g....

>I think warnings should be there unconditionally - the way I see this design leading to annoying bugs is when the application developer mainly develops on Windows or whatever, and...

>Now I realize that perhaps making the cursor position optional was a mistake, everyone should support it. If you don't want popup window visible as a user it could be...

I'd assume that if you have `warn!` logging in place, that should probably be enough for most cases, especially if we recommend backends to not perform operation when it was...

Could you post `--print-events` while doing so? I'd assume it does switch, but you just have garbage written?