Context dynamic dispatch

[i509VCB]

Checklist

• Run cargo clippy.
• Run RUSTFLAGS=--cfg=web_sys_unstable_apis cargo clippy --target wasm32-unknown-unknown if applicable.
• Add change to CHANGELOG.md. See simple instructions inside file.

Description
This pull request does most of the work required to allow wgpu backends to be set at runtime. In the future this would allow selection between webgl2 and webgpu on web and on native other backends (including my use case which crosses process boundaries).

What would need to be done after this is to solidify the internal Context trait and types so those can be publicly exported.

Testing
This pull request needs testing with crates across the ecosystem. The dynamic dispatch in the internals does pass the tests in wgpu, but best to be safe.

There hasn't been any testing for the web backend. There could be errors that may have been added during the dynamic dispatch internal mess. However the testing situation for the web has been grim for a while.

[expenses]

This, is exciting! I've been wanting a way to switch between the WebGL 2 and WebGPU backends at runtime. On linux I'm getting an error when trying to run the hello-triangle example natively:

thread 'main' panicked at 'IdSendSync was downcast to the wrong type', wgpu/src/context.rs:38:14
stack backtrace:
   0: rust_begin_unwind
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/std/src/panicking.rs:584:5
   1: core::panicking::panic_fmt
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/panicking.rs:142:14
   2: core::panicking::panic_display
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/panicking.rs:72:5
   3: core::panicking::panic_str
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/panicking.rs:56:5
   4: core::option::expect_failed
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/option.rs:1874:5
   5: core::option::Option<T>::expect
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/option.rs:738:21
   6: wgpu::context::IdSendSync::downcast_id
             at ./src/context.rs:36:9
   7: <wgpu::backend::direct::Context as wgpu::Context>::instance_request_adapter::{{closure}}
             at ./src/backend/direct.rs:869:37
   8: core::option::Option<T>::map
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/option.rs:929:29
   9: <wgpu::backend::direct::Context as wgpu::Context>::instance_request_adapter
             at ./src/backend/direct.rs:867:37
  10: <T as wgpu::context::DynContext>::instance_request_adapter
             at ./src/context.rs:654:22
  11: wgpu::Instance::request_adapter
             at ./src/lib.rs:1809:23
  12: hello_triangle::run::{{closure}}
             at ./examples/hello-triangle/main.rs:12:19
  13: <core::future::from_generator::GenFuture<T> as core::future::future::Future>::poll
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/future/mod.rs:91:19
  14: pollster::block_on
             at /home/ashley/.cargo/registry/src/github.aaakk.us.kg-1ecc6299db9ec823/pollster-0.2.5/src/lib.rs:125:15
  15: hello_triangle::main
             at ./examples/hello-triangle/main.rs:145:9
  16: core::ops::function::FnOnce::call_once
             at /rustc/a55dd71d5fb0ec5a6a3a9e8c27b2127ba491ce52/library/core/src/ops/function.rs:248:5
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

[i509VCB]

@expenses I have fixed the bug and all the tests and examples (except for mipmap due to features) run on my system (radv Vulkan).

I managed to forget that a Box<dyn Any + Send + Sync> is &(dyn Any + Send + Sync), but a reference to the box is a borrow over the box, not the inner value. Hence the as_ref calls.

[cwfitzgerald]

I need to take another look inside my IDE to see if I see anything weird, but here's some first pass comments.

[codecov-commenter]

Codecov Report

Merging #3051 (2b59812) into master (f0f700c) will decrease coverage by 1.36%.
The diff coverage is 66.66%.

❗ Current head 2b59812 differs from pull request most recent head 10b5b6f. Consider uploading reports for the commit 10b5b6f to get more accurate results

@@            Coverage Diff             @@
##           master    #3051      +/-   ##
==========================================
- Coverage   65.64%   64.27%   -1.37%     
==========================================
  Files          82       83       +1     
  Lines       39479    42260    +2781     
==========================================
+ Hits        25915    27164    +1249     
- Misses      13564    15096    +1532     

Impacted Files	Coverage Δ
wgpu-core/src/lib.rs	96.55% <ø> (ø)
wgpu-core/src/track/buffer.rs	89.08% <ø> (-3.07%)	⬇️
wgpu-core/src/track/metadata.rs	90.97% <ø> (+1.50%)	⬆️
wgpu-core/src/track/mod.rs	75.18% <ø> (ø)
wgpu-core/src/track/texture.rs	59.96% <ø> (-21.02%)	⬇️
wgpu-types/src/lib.rs	88.12% <ø> (ø)
wgpu/src/backend/direct.rs	57.14% <ø> (+0.70%)	⬆️
wgpu/src/context.rs	52.95% <ø> (ø)
wgpu/src/lib.rs	50.26% <ø> (-1.18%)	⬇️
wgpu/src/util/mod.rs	12.32% <0.00%> (-0.18%)	⬇️
... and 8 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[cwfitzgerald]

Alright, I think I can basically say that this is good enough to merge, minus the mentioned comments. I don't think we can catch all the issues with this massive refactor by just staring at it, so I want to merge and improve iteratively.

I do think there might be some macro magic that makes the amount of boilerplate we need less, but we can stew on that at some future time.

[cwfitzgerald]

Needs conflicts resolved, but I think it's time to just pull the trigger on this.

[94bryanr]

I am now getting an error when building a SurfaceConfiguration from a Web Worker (this use case worked before).

panicked at 'assertion failed: wasm.is_instance_of::<T>()', /home/b/.cargo/registry/src/github.aaakk.us.kg-1ecc6299db9ec823/wgpu-0.15.0/src/backend/web.rs:53:9

at <wgpu::backend::web::Identified<T> as core::convert::From<wgpu::context::ObjectId>>::from::h65779d5bfdededd6 (ccfe_bg.wasm:0x9a6d4f)

This happens when calling

adapter: &wgpu::Adapter ...

let caps = surface.get_capabilities(adapter);

The wgpu::Surface in this case was initialized via instance.create_surface_from_offscreen_canvas(...);

[i509VCB]

Yep the bug is known, it's being tracked in #3430