Lemme try it. If I don't ping back in 24 hours, feel free to assign it to someone else c:

Okay, I have a candidate fix. Let me test it and then make a PR

I might need some advice. Here is what I have for now.

Summary

I have fixes for the 3 build errors that appeared in CI. The three of them, I think, can be dealt with without much fuss.

Further testing does reveal another build error, which I'll describe in the second section.

Fixes for the 3 errors seen in CI

These are the errors, reproduced on my end:

The first one is due to the new MemoryHints field that is fed to DeviceDescriptor. I am currently asking for a MemoryUsage profile, on the grounds that it should help with compatibility (I think, I might be wrong on this).

This error is fixed by just adding a memory usage profile.

The second and third errors are due to a new field, cache: Option<PipelineCache> that is now a part of RenderPipelineDescriptor. I have set it to None for now, and can be changed later if more performance is needed.

Those fixes make cargo build, test and clippy all pass without a problem.

The new issue I'm looking at (advice is much appreciated)

I ran just ci locally after applying the fixes above, and this is the error I get:

The RUSTFLAGS that were used are the following:

Running RUSTFLAGS="-D warnings" "cargo" "build" "--all-features" "--target" "wasm32-unknown-unknown" "-p" "fj-viewer"

I'm not sure what to do about this error.

Reading the release notes, it looks like enumerate_adapters is now native-only:

I'm guessing that means we can't use it on wasm32-unknown-unknown anymore, and hence a different approach to obtaining the adapters is needed. request_adapters, perhaps? I'll take a look at it tomorrow, and see if something clicks about it. I'm pretty sure something will.

Thank you for looking into this, @felix91gr!

My reading matches yours: We need to avoid enumerate_adapters with wasm32-unknown-unknown, and need request_adapter instead. This requires a surface, as part of RequestAdapterOptions, but that's not a problem, since we already have one where we need it.

Looking over the code, I think this shouldn't be too hard to adapt. Let's look at what we're doing:

1. Use enumerate_adapters to log all adapters before trying to do anything. Should be fine to just slap a #[cfg(not(target = "wasm32-unknown-unknown"))] on there. (Not sure about the specific syntax, but something along those lines should work.
2. Call Device::from_preferred_adapter, which uses request_adapter internally. This is the main code path and should require no change.
3. In case of failure, call Device::try_from_all_adapters as a fallback. Since this is just fallback behavior, it should be fine to just slap another #[cfg(not(...))] on there.

So to summarize, the main code path is already alright, and just debug output and fallback in case of errors use the problematic enumerate_adapters. Since, according the the changelog (as highlighted in the screenshot you posted), enumerate_adapters returned an empty Vec all along on WASM, just configuring them out as I suggested should not change any behavior.

Addressed in #2427.