GlobalTransform component not synchronized in bevy 0.10 for entities spawned after TransformSystem::TransformPropagate?

[antilankin]

Hi all,

I'm working on a tile-based map editor for a game my brother is building, and I'm in the process of migrating it to bevy 0.10. While, on the whole, this works really well thanks to the excellent migration guide, I ran into a problem:

I'm running, among others, three systems:

In bevy 0.9, I ran them like this:

app
    .add_system_to_stage(
        CoreStage::PostUpdate,
        mouse_system
            .after(TransformSystem::TransformPropagate)
            .label("Selection")
    )
    .add_system_set_to_stage(
        CoreStage::PostUpdate,
        SystemSet::new()
            .after("Selection")
            .with_system(spawn_entity)
            // with other systems
    )
    .add_system_to_stage(
        CoreStage::Last,
        clear_selection_system
    );

In bevy 0.10, the systems look like this:

app
    .add_system(mouse_system.in_base_set(CoreSet::PostUpdate).after(TransformSystem::TransformPropagate))
    .add_system(spawn_entitiy.in_base_set(CoreSet::PostUpdate).after(mouse_system))
    // add other systems
    .add_system(clear_spawn_marker.in_base_set(CoreSet::Last));

The mouse_system adds a resource containing tile coordinates when the left button is clicked. I'm using Camera::viewport_to_world, that's why I scheduled the system after TransformSystem::TransformPropagate. The spawn_entity system looks for the resource and spawns an entity at the specified position. The clear_spawn_marker system removes the resource again.

While this worked in bevy 0.9, my spawned entities appear at the origin since the migration to 0.10. Some debugging revealed that while the Transform component shows the correct position, the GlobalTransform component of the spawned entities is (0,0,0).

I was expecting that the GlobalTransform would be updated on the next tick, but this isn't the case. While my issue is easily solved, for instance by re-arranging the systems, I'm curious where the difference between 0.9 and 0.10 is coming from.

[tim-blackbird]

It's likely because of the bug that's being fixed here: #8029

[antilankin]

Looks like that's it, thanks for the quick answer!