Changes for bevy 0.11-dev

[eerii]

This is a draft version of the changes needed for this project to work with bevy-main.

The main change is the new add_systems, which now takes the schedule pattern as a first parameter and a tuple of systems as the second. add_system (singular) is now deprecated.

Another breaking change was flipping the y coordinate for consistency. This means we can remove the inverted mouse in the process_input_system.

Will add more to this draft if there are further breaking changes.

[imalsogreg]

Thanks for doing this! I've noticed when running the ui example in this fork, it panics with

thread 'main' panicked at 'called `Option::unwrap()` on a `None` value', src/egui_node.rs:47:73

which corresponds to this line

let render_device = render_world.get_resource::<RenderDevice>().unwrap();

bevy_egui expects to find a RenderDevice resource, but there is none. The problem is limited to 0.11.0-dev (running on upstream main, the ui example works as expected).

I haven't been able to figure out what's going on. (Maybe in bevy-0.11.0-dev, inserting the RenderDevice resource happens at a later scheduler phase?

[diinki]

I haven't been able to figure out what's going on. (Maybe in bevy-0.11.0-dev, inserting the RenderDevice resource happens at a later scheduler phase?

I believe this is correct, In the lib.rs I changed the

    // Line 587
    .init_resource::<egui_node::EguiPipeline>()
    .init_resource::<SpecializedRenderPipelines<EguiPipeline>>()
    .init_resource::<EguiTransforms>()

resources to be added by another system with the ExtractSchedule schedule, fixes the RenderDevice resource being None. but I do have another potentially unrelated bug where the fonts aren't rendering anymore, but maybe this insight helps a bit for now.

[imalsogreg]

@diinki brilliant! This worked for me. @josekoalas in case it's helpful, this commit is what gets ui working for me: imalsogreg@0393b7c

(I didn't think very carefully about the naming)

[seabassjh]

@josekoalas @imalsogreg I was playing around with both of your changes on bevy main, but with webgpu and wasm-- and everything works great... just one thing :)

There's an error with the color attachment setup in the render pipeline pass.
I authored a fix on top of @imalsogreg's work which got it all working:
seabassjh/bevy_egui@7ee68d3

Tested on Chrome Stable version 114.
Commands I used to test WASM/Webgpu:
export RUSTFLAGS=--cfg=web_sys_unstable_apis; cargo build --release --target wasm32-unknown-unknown --example simple && wasm-bindgen --out-dir target --out-name wasm_example --target web target/wasm32-unknown-unknown/release/examples/simple.wasm && basic-http-server .

Hopefully someone can reproduce this error and fix on their machine!

[diinki]

I'd like to mention that this commit changed the name of schedule::apply_system_buffers to schedule::apply_deferred.

This commit causes more necessary changes in lib.rs ( just change the name of apply_system_buffers everywhere to apply_deferred c: ).

I'll update my fork to work with the latest version of Bevy in a bit, if anyone wonders which commits would fix the bugs as well :p

[mvlabat]

I'll close it as #188 has been merged. But thank you for your work!