fcddb54ce5
# Objective - Add a test case for #14300 Fixes #14300 ## Solution `SceneEntityMapper` relies on operations on `Entities` that require flushing in advance, such as `alloc` and `free`. Previously, it wasn't calling `world.flush_entities()` itself and relied on its caller having flushed beforehand. This wasn't an issue before observers and hooks were released, since entity reservation was happening at expected times. Now that hooks and observers are a thing, they can introduce a need to flush. We have a few options: * Flush after each observer/hook run * Flush between each paired observer/hook and operation that requires a flush * Flush before operations requiring it The first option for this case seemed trickier to reason about than I wanted, since it involved the `BundleInserter` and its `UnsafeWorldCell`, and the second is generally harder to track down. The third seemed the most straightforward and conventional, since we can see a flush occurring at the start of a number of `World` methods. Therefore, we're letting `SceneEntityMapper` be in charge of upholding its own invariants and calling `flush_entities` when it's created. ## Testing Added a new test case modeled after #14300 |
||
|---|---|---|
| .. | ||
| src | ||
| Cargo.toml | ||
| README.md | ||
Bevy Scene
