mirror of
https://github.com/bevyengine/bevy
synced 2024-11-22 04:33:37 +00:00
424e563184
# Objective - Make the example deterministic when run with CI, so that the [screenshot comparison](https://thebevyflock.github.io/bevy-example-runner/) is stable - Preserve the "truly random on each run" behavior so that every page load in the example showcase shows a different contributor first ## Solution - Fall back to the static default contributor list in CI - Store contributors in a `Vec` so that we can show repeats of the fallback contributor list, giving the appearance of lots of overlapping contributors in CI - Use a shared seeded RNG throughout the app - Give contributor birds a `z` value so that their depth is stable - Remove the shuffle, which was redundant because contributors are first collected into a hashmap - `chain` the systems so that the physics is deterministic from run to run ## Testing ```bash echo '(setup: (fixed_frame_time: Some(0.05)), events: [(100, Screenshot), (500, AppExit)])' > config.ron CI_TESTING_CONFIG=config.ron cargo run --example contributors --features=bevy_ci_testing mv screenshot-100.png screenshot-100-a.png CI_TESTING_CONFIG=config.ron cargo run --example contributors --features=bevy_ci_testing diff screenshot-100.png screenshot-100-a.png ``` ## Alternatives I'd also be fine with removing this example from the list of examples that gets screenshot-tested in CI. Coverage from other 2d examples is probably adequate. --------- Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com> |
||
|---|---|---|
| .. | ||
| alien_cake_addict.rs | ||
| breakout.rs | ||
| contributors.rs | ||
| desk_toy.rs | ||
| game_menu.rs | ||
| loading_screen.rs | ||
| stepping.rs | ||