keyboardio / kaleidoscope-hidadaptor-keyboardiohid Goto Github PK

So, I broke rollover from a modifier-flagged key again with #17.

That PR looked good, but the test it needs to do is more specific than in can do (without additional changes to KeyboardioHID, or violation of its encapsulation). The problem is that it was allowing updates of modifier bits and keycode bits in the same report, which leads to a pseudo race condition on the host (does it process the modifier update first, or the keycode update?).

With stock firmware, this can be easily reproduced by using the FUNCTION layer, and rolling over from the } key to [ – you'll see }{ instead of }[. But if you roll over from { to [ instead, you'll get {[ as you should, because the same keycode that was just pressed was already held, so it triggers an extra report, clearing the shift before the [ is added again.

PR on the way shortly.