Back

1. https://github.com/bevyengine/bevy/pull/12720

A 2D camera example! This is the sort of "practical gamedev task" example that we're sorely missing. Lots of "how do each of the tools and primitives that Bevy provides" works, but very little advice that assembles those tools into user-facing workflows. You really need both.

Good back-and-forth with reviewers and the author: this is much stronger for it. I'm happy with this as learning material: merging.

2. https://github.com/bevyengine/bevy/pull/13248

Ah, Francois's screenshot PR still doesn't have a secret added. Well, that gets the Ready-For-Final-Review label swapped out for Waiting-On-Author so it stops showing up in this search :P

3. https://github.com/bevyengine/bevy/pull/13677

And now for something much more exciting: the first work for upstreaming bevy_mod_picking! We're doing this in chunks: copy-pasting portions of the library with minimal adaptation to start. Once that's all in place, we'll make sure it works, add examples and do some tech debt cleanup.

I'm feeling good about getting this in place for 0.15 :D Sorely needed, and it'll be nice to share the load in its maintenance.

4. https://github.com/bevyengine/bevy/pull/13714

A "usability" fix to our scene methods to make them more flexible. This PR makes me happy, because the author was able to swap from a +150 line feature to a +6 function signature change to meet the same needs. Engineering!

The author was hoping this would make 0.14, but that train has left. A shame, but this is a breaking change and it's not vital enough / impacting enough users to warrant bending the rules. Anyways, this is a nice consistency improvement. Merging!

5. https://github.com/bevyengine/bevy/pull/13719

Another feature PR, but this one is non-breaking. This adds the ability to automatically segment up extruded 2D shapes. Really fun to see procedural geometry in action. Thanks math wizs!

I'm fine with this getting cherrypicked: it's non-breaking, low risk and any damage will be very well-isolated to a brand-new, relatively niche feature. Merging and adding to the 0.14 milestone.

6. https://github.com/bevyengine/bevy/pull/13741

By contrast, *this* math PR by the same author is not getting added to the milestone. Inspired by a @acegikmo video about how lerp smoothing is flawed, it provides a trait to cover the correct mathematical requirements for smooth interpolation. Too impactful for a last minute addition, even without breaking changes! QC matters too.

Video: https://www.youtube.com/watch?v=LSNQuFEDOyQ

Animation's concerns have been allayed, quality is exceptional, there's ASCII art! Merging :)

@acegikmo 7. https://github.com/bevyengine/bevy/pull/13745

And *here's* a breaking change that I want to backport to 0.14 (is it really backporting before it's been released??). Jojojet previously made a PR to add functionality to our asset server to load assets in a custom async way (due to needs discovered at work), but discovered that their error handling strategy wasn't helpful.

This change allows a custom error type to be supplied. Very niche, and a dramatic improvement to a new API. Merging :)

@acegikmo 8. https://github.com/bevyengine/bevy/pull/13752

A simple cleanup to our shader code, removing unused variables. People talk a lot about how shaders are hard to learn, painful to work with, complex and so on. But most of that is just terrible tooling (and language limitations). No dead code analysis, very basic language server, no fancy linters, bad error messages... Writing graphics code should be easier!

Anyways, change is fine. Merging :P

@acegikmo 9. https://github.com/bevyengine/bevy/pull/13759

Improvements to the new color API (color to u8 array), which are getting cherrypicked to 0.14. This sort of thing is a huge victory for the release candidate: Bevy should be ready to use and delightful right on release day.

This author got sniped into being a contributor after complaining about a painful migration :D Really good back-and-forth with the rest of the working group on how to best solve these problems. I love onboarding new devs. Merging!

10. https://github.com/bevyengine/bevy/pull/13761

More docs and a rename for meshlets. Jasmine has really been leading this effort, and these changes look solid. Merging!

Go read their very detailed blog post about the technical details here: https://jms55.github.io/posts/2024-06-09-virtual-geometry-bevy-0-14/

11. https://github.com/bevyengine/bevy/pull/13762

And this is a *nasty* regression, caught during the release candidate process. We managed to break event (and RemovedComponent) double-buffering, by adding an extra flush to subapps. Turns out this was intentional to make sure that the render app can use RemovedComponents correctly.

Ahhh tech debt. I expect that pretty much all of the concepts here will be gone in a year lol, in part due to nonsense interactions like this. Good tests, good fix, merge conflicts

12. https://github.com/bevyengine/bevy/pull/13763

Solid work by the I-can't-believe-it's-not-a-working-group states gang. The system ordering of the transitions for our new computed states weren't ordered correctly, leading to subtle and hard-to-reproduce bugs.

Very pleased with this fix: this is an important bug. A previous version relied on a new `StatesRegistry` concept, which I pushed back on due to introducing new concepts for a bug fix. Now it's just system set ordering! Great stuff! Merging :D

13. https://github.com/bevyengine/bevy/pull/13766

A simple error message improvement. The error describes how a shader is "not loaded yet", but in most cases, the shader was *never* going to be loaded, likely due to a bad path. Both error messages are true, but the new one is a lot less misleading. This stuff adds up!

Great simple, easy to review fix from a new contributor. Merging.

14. https://github.com/bevyengine/bevy/pull/13780

Users on older hardware sometimes complain that Bevy is mysteriously slow. Most of the time, this is due to their GPU drivers not actually working! wgpu simply rejects broken, terrible graphics drivers, while many other programs try and work around them. There's a CPU fallback but it's very slow.

This PR adds a warning explaining what's wrong and how to fix it. Rerequesting a review from Francois and waiting; he had concerns about a proposed fix.

15. https://github.com/bevyengine/bevy/pull/13772

Another release-candidate-inspired PR. Ida, the author of the indomitable Bevy Cheatbook, was wondering about how to migrate some of her states code, and noticed that there was still a reference to the now-removed `apply_state_transition` system.

This PR cleans that up, and explains how to do the "manually update states" pattern that she was previously using. Great! Merging.

16. https://github.com/bevyengine/bevy/pull/13783

Whether or not fog was enabled, the shader was branching on it. If you know anything about GPU programming: branches are very slow! Swap this to something closer to a compile-time decision and bam, zero-cost rendering feature. Compile-time here is "compiled at the start of the game on the user's machine", but same deal!

Reasonable, well-reviewed. Merging.

17. https://github.com/bevyengine/bevy/pull/13784

A nice rename and consistency fix PR for our run conditions. Run conditions can be combined in 0.13 using `and_then` and `or_else`. But there are two problems here: first, why are we using these weird monadic names? Second, why not have the rest of the binary operations?

I spent a while socializing this change: bikeshedding can get intense but everyone thought it was a good idea.

Merging, but not cherrypicking: this is a new feature.

I really like that a new contributor was able to get a relatively contentious but straightforward change in, especially in a reasonable period of time. Says really nice things about our culture and processes I think!

We do still struggle with new abstractions and big overhauls, but the more explicit status labelling, an intermediate category for controversy and especially working groups have done a great job building consensus.

Anyways, that's all the PRs for this week! I'm very pleased about the release candidate on several levels: less stress, more polish, fewer critical bugs, more time to update crates and docs...

Really delightful. 10/10 would recommend all #opensource projects do this.

Now if you'll excuse me, I'm going to go immerse myself in some crunchy coding (LWIM refactors!): these past few weeks have been all tech writing and PM stuff and I really need an excuse to do anything else.