Nicolas Silva

Since I spend more time n code that use wgpu-core than wgpu I hadn't noticed that wgpu was only exposing a sort of placeholder error type (BufferAsyncError), and that it...

I removed the commit that replaces BufferAsyncError in favor of BufferAccessError for now. I'll revisit when I better understand the error handlign story.

I rebased the PR and applied crowlCats's suggestion. The PR still only affects wgpu-core (and not wgpu). If I remember correctly the main decision to make for wgpu is how...

> So, let me know what y'all think :) This would be great! For inspiration, see also [glium's book/tutorial](https://github.com/glium/glium/tree/master/book) which I think is quite good. > If we focus on...

Looks like this generalizes to constant expressions. I'm new to cbindgen's code but it looks like a fix for this would have to go somewhere around here: https://github.com/mozilla/cbindgen/blob/0b43f0bc6ce8aad77a5d1417657daddd32905853/src/bindgen/ir/constant.rs#LL342C33-L342C33 Unfortunately since...

I don't know off hands but it indeed looks like the exact same thing so I'll close this. Thanks @emilio !

While submitting this I realized that my fix was too simplistic. That particular validation needs to be redone in a more invasive way I think.

Thanks! @bors-servo r+

> I think it might be a bit understated, but breaking things up (not necessarily into traits) by making some source type visible could make downcasts less error (or panic)...

One of the questions to consider is at which layer of wgpu's architecture should we have the dyn traits. - A) `wgpu-hal`'s `HalApi. - B) Something in `wgpu-core` implements dynamic...