Stack overflow with ambiguity detection and `.run_if(in_state(...))`

[SorteKanin]

Bevy version

bevy = "0.11.2"

[Optional] Relevant system information

> cargo --version -v
cargo 1.71.1 (7f1d04c00 2023-07-29)
release: 1.71.1
commit-hash: 7f1d04c0053083b98fa50b69b6f56e339b0556a8
commit-date: 2023-07-29
host: x86_64-pc-windows-msvc
libgit2: 1.6.4 (sys:0.17.1 vendored)
libcurl: 8.0.1-DEV (sys:0.4.61+curl-8.0.1 vendored ssl:Schannel)
os: Windows 10.0.19045 (Windows 10 Home) [64-bit]

What you did

Minimal example:

use bevy::{
    ecs::schedule::{LogLevel, ScheduleBuildSettings},
    prelude::*,
};

#[derive(States, Clone, Copy, Default, Debug, PartialEq, Eq, Hash)]
enum DummyState {
    #[default]
    Foo,
    Bar,
}

fn main() {
    App::new()
        .edit_schedule(Update, |schedule| {
            schedule.set_build_settings(ScheduleBuildSettings {
                ambiguity_detection: LogLevel::Error,
                ..default()
            });
        })
        .add_state::<DummyState>()
        .add_plugins(DefaultPlugins)
        .add_systems(
            Update,
            (dummy_system, dummy_system).run_if(in_state(DummyState::Foo)),
        )
        .run()
}

fn dummy_system() {}

What went wrong

thread 'main' has overflowed its stack upon running.

Additional information

Note that using distributive_run_if seems to fix the problem. Also adding just one system (i.e. dummy_system.run_if(in_state(DummyState::Foo)) does not exhibit the problem. Problem happens even if different systems are used, the fact that it's the same system in the tuple is unrelated.

Only confirmed on Windows. Would be great if someone with a Linux/Mac machine, could confirm/deny if this happens on other platforms.

[EmiOnGit]

Can confirm the same problem on linux

[hymm]

I can't reproduce on main branch, so it might have been fixed already. Can someone else test main to confirm?

[EmiOnGit]

I can't reproduce on main branch, so it might have been fixed already. Can someone else test main to confirm?

Same for me; main works without problems
@SorteKanin can you check?

[SorteKanin]

Yes it does seem to be fixed on the main branch (should've checked before probably...). Closing.

[hymm]

I bisected this back to #9260. If I revert that change and run the above code it still stack overflows. Since that change was just moving systems around, the problem likely still exists on main too. Will try and see if I can come up with a better minimal reproduction.

[hymm]

use bevy::prelude::*;

fn main() {
    let mut schedule = Schedule::default();

    fn exclusive(_world: &mut World) {}
    
    fn dummy() {}
    
    schedule.set_build_settings(bevy::ecs::schedule::ScheduleBuildSettings {
        ambiguity_detection: bevy::ecs::schedule::LogLevel::Error,
        ..Default::default()
    });
    
    schedule.add_systems((exclusive, (dummy, dummy).run_if(|| true)));
    
    schedule.initialize(&mut World::new()).unwrap();
}

This stack overflows. Probably has something to do with anonymous system sets and exclusive systems.

[hymm]

This turned out to be a problem with getting an anonymous system sets name when the report_sets settings was true. See above PR for more details.