Device::as_hal crashes after ID removal in #6134

[ErichDonGubler]

Description

@MarijnS95 noted in this comment from #6134 that Device::as_hal appears to have regressed to crashing when called in some (?) cases.

Repro steps

TBD by @MarijnS95.

Expected vs observed behavior

We should keep previous behavior, that is, Device::as_hal should not crash.

Platform

Windows on DX12.

[MarijnS95]

Platform is Windows. I was running a modified version of the hello_triangle sample while playing with DirectComposition (needed access to the D3D12Device for that).

Will push a repro branch soon.

[MarijnS95]

Try https://github.com/MarijnS95/wgpu/compare/repro-6169 with:

$ WGPU_BACKEND=dx12 cargo r -p wgpu-examples hello_triangle
...
thread 'main' panicked at wgpu\src\context.rs:737:5:
assertion failed: data.is::<T>()
stack backtrace:
...

Note that in this caseit's already failing for surface, before even getting to adapter or device which also both fail. Only instance.as_hal() is successful.

[ErichDonGubler]

Confirmed locally on my M1 Mac that the suggested repro crashes, and that latest trunk does not reproduce the crash. I did this on Mac, instead of Windows, by replacing dx12 with metal in the tip commit of the suggested repro, i.e.:

diff --git a/examples/src/hello_triangle/mod.rs b/examples/src/hello_triangle/mod.rs
index 7c82d49cf0..0100b79024 100644
--- a/examples/src/hello_triangle/mod.rs
+++ b/examples/src/hello_triangle/mod.rs
@@ -12,7 +12,10 @@
 
     let instance = wgpu::Instance::default();
 
-    let surface = instance.create_surface(&window).unwrap();
+    unsafe { instance.as_hal::<wgpu::hal::metal::Api>() };
+
+    let mut surface = instance.create_surface(&window).unwrap();
+    unsafe { surface.as_hal::<wgpu::hal::metal::Api, _, _>(|s| ()) };
     let adapter = instance
         .request_adapter(&wgpu::RequestAdapterOptions {
             power_preference: wgpu::PowerPreference::default(),
@@ -22,6 +25,7 @@
         })
         .await
         .expect("Failed to find an appropriate adapter");
+    unsafe { adapter.as_hal::<wgpu::hal::metal::Api, _, _>(|a| ()) };
 
     // Create the logical device and command queue
     let (device, queue) = adapter
@@ -38,6 +42,7 @@
         )
         .await
         .expect("Failed to create device");
+    unsafe { device.as_hal::<wgpu::hal::metal::Api, _, _>(|d| ()) };
 
     // Load the shaders from disk
     let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {

[MarijnS95]

Yep, the repro branch no longer crashes after rebasing, thanks!