Events<GamepadEventRaw> appears to be getting updated twice per world tick (gilrs)

[CAD97]

Bevy version

git master

What you did

Looked at the code (I'm playing around with building a custom mini engine using bevy ECS, for learning purposes).

What you expected to happen

Events::update says it should be called once per tick, to manage the double buffer correctly.

What actually happened

update appears to be getting called twice per tick:

• bevy_input calls add_event::<GamepadEventRaw> https://github.com/bevyengine/bevy/blob/99199338ad5e3d89d767f639ef4cd7b6da52ed83/crates/bevy_input/src/lib.rs#L65 and this registers the Event::update_system https://github.com/bevyengine/bevy/blob/99199338ad5e3d89d767f639ef4cd7b6da52ed83/crates/bevy_app/src/app.rs#L583-L589
• bevy_gilrs's gilrs_event_system calls event.update on the Events<GamepadEventRaw> https://github.com/bevyengine/bevy/blob/99199338ad5e3d89d767f639ef4cd7b6da52ed83/crates/bevy_gilrs/src/gilrs_system.rs#L19-L23

Additional information

It's possible I'm just misunderstanding what's going on here, but this seems wrong at first glance. At the very least, the update call in bevy_gilrs probably deserves a comment calling out that it's a second update call in the simulation tick explaining why we want to discard the double buffer in this case.

[alice-i-cecile]

Good find. These events should not be cleared twice. Removing the update from gilrs is the correct way to fix this.

Incidentally, that system looks like it no longer need to be exclusive: we should fix that in the same PR.

[CAD97]

AIUI systems that access non-send resources (like Gilrs is) still need to be exclusive. Or if they don't, I didn't find the correct way to request access.

[alice-i-cecile]

Nope, they'll work fine in ordinary systems :) Just use NonSend as a system parameter.