Old GPU unable to find layer VK_LAYER_KHRONOS_validation

[TheYahton]

Description When i try cargo run --bin wgpu-examples hello with env vars set RUST_LOG=trace and set WGPU_BACKEND=gl it gives an error:

C:\Users\oops\projects\rust\wgpu>cargo run --bin wgpu-examples hello
    Finished dev [unoptimized + debuginfo] target(s) in 0.96s
     Running `target\debug\wgpu-examples.exe hello`
[2024-01-31T09:50:29Z WARN  wgpu_hal::vulkan::instance] Unable to find extension: VK_EXT_debug_utils
[2024-01-31T09:50:29Z WARN  wgpu_hal::vulkan::instance] Unable to find extension: VK_EXT_swapchain_colorspace
[2024-01-31T09:50:29Z WARN  wgpu_hal::vulkan::instance] InstanceFlags::VALIDATION requested, but unable to find layer: VK_LAYER_KHRONOS_validation
error: process didn't exit successfully: `target\debug\wgpu-examples.exe hello` (exit code: 0xc0000005, STATUS_ACCESS_VIOLATION)

Repro steps

set RUST_LOG=trace
set WGPU_BACKEND=gl
cargo run --bin wgpu-examples hello

Expected vs observed behavior Vulkan shouldn't crash the program if I use opengl.

Extra materials

Platform Windows 11, NVIDIA GeForce GT 730 (GF108) with 391.35 drivers (latest available), no vulkan or directx 12 support. P.S. When i use Manjaro with nvidia drivers the same problem occurs. With open source driver (forgot the name :/) everything was ok.

[TheYahton]

I think it's related #4109

[cwfitzgerald]

Could you get a stack trace from a debugger?

[TheYahton]

Could you get a stack trace from a debugger?

0: std::backtrace_rs::backtrace::dbghelp::trace
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\..\..\backtrace\src\backtrace\dbghelp.rs:98
   1: std::backtrace_rs::backtrace::trace_unsynchronized
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\..\..\backtrace\src\backtrace\mod.rs:66
   2: std::backtrace::Backtrace::create
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\backtrace.rs:332
   3: std::backtrace::Backtrace::capture
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\backtrace.rs:298
   4: wgpu_examples::hello::main
             at .\examples\src\hello\mod.rs:23
   5: wgpu_examples::main
             at .\examples\src\main.rs:229
   6: core::ops::function::FnOnce::call_once<void (*)(),tuple$<> >
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26\library\core\src\ops\function.rs:250
   7: std::sys_common::backtrace::__rust_begin_short_backtrace<void (*)(),tuple$<> >
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26\library\std\src\sys_common\backtrace.rs:135
   8: std::sys_common::backtrace::__rust_begin_short_backtrace<void (*)(),tuple$<> >
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26\library\std\src\sys_common\backtrace.rs:135
   9: std::rt::lang_start::closure$0<tuple$<> >
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26\library\std\src\rt.rs:166
  10: std::rt::lang_start_internal::closure$2
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\rt.rs:148
  11: std::panicking::try::do_call
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\panicking.rs:500
  12: std::panicking::try
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\panicking.rs:464
  13: std::panic::catch_unwind
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\panic.rs:142
  14: std::rt::lang_start_internal
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26/library\std\src\rt.rs:148
  15: std::rt::lang_start<tuple$<> >
             at /rustc/eb26296b556cef10fb713a38f3d16b9886080f26\library\std\src\rt.rs:165
  16: main
  17: invoke_main
             at D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\exe_common.inl:78
  18: __scrt_common_main_seh
             at D:\a\_work\1\s\src\vctools\crt\vcstartup\src\startup\exe_common.inl:288
  19: BaseThreadInitThunk
  20: RtlUserThreadStart

[cwfitzgerald]

This is a backtrace of the segfault?

[tasogare3710]

If not vulkan compatible, InstanceError { message: "missing Vulkan entry points", .. } should be returned. Could it be a bug in the driver you are using?(Or related this?) No problem in my gpu.

---

set WGPU_ALLOW_UNDERLYING_NONCOMPLIANT_ADAPTER=0
set WGPU_BACKEND=gl
set RUST_LOG=trace

[TheYahton]

Anyway, I don't care about fixing bugs in wgpu or in my driver. I'm interested in stupid running wgpu-examples without checking vulkan. How can I do that?

[cwfitzgerald]

The hello and hello-triangle example do not honor WGPU_BACKEND as they were trying to avoid any complexity - this is largely confusing and should likely change.

Try the cube (or any other) example. We still shouldn't be segfaulting, but that should get Vulkan ignored.

[TheYahton]

The hello and hello-triangle example do not honor WGPU_BACKEND as they were trying to avoid any complexity - this is largely confusing and should likely change.

Try the cube (or any other) example. We still shouldn't be segfaulting, but that should get Vulkan ignored.

set WGPU_BACKEND=gl
cargo run --bin wgpu-examples cube

works perfect