Texture arrays example panics in v25.0.2.1

[jerrydzhang]

Attempting to run the texture arrays example causes a panic here is the output:

thread '<unnamed>' panicked at src/lib.rs:417:5:
wgpu uncaptured error:
Validation Error

Caused by:
  In wgpuDeviceCreateBindGroupLayout, label = 'bind group layout'
    Too many bindings of type BindingArrayElements in Stage ShaderStages(FRAGMENT), limit is 0, count was 6. Check the limit `max_binding_array_elements_per_shader_stage` passed to `Adapter::request_device`


stack backtrace:
   0: __rustc::rust_begin_unwind
   1: core::panicking::panic_fmt
   2: wgpu_native::default_uncaptured_error_handler
             at /home/user/wgpu-native/src/lib.rs:417:5
   3: wgpu_native::ErrorSinkRaw::handle_error
             at /home/user/wgpu-native/src/lib.rs:551:25
   4: wgpu_native::handle_error
             at /home/user/wgpu-native/src/lib.rs:632:5
   5: wgpuDeviceCreateBindGroupLayout
             at /home/user/wgpu-native/src/lib.rs:1865:9
   6: main
             at ./main.c:535:43
   7: __libc_start_call_main
   8: __libc_start_main_impl
   9: _start
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

thread '<unnamed>' panicked at library/core/src/panicking.rs:226:5:
panic in a function that cannot unwind
stack backtrace:
   0:          0x1166329 - <std::sys::backtrace::BacktraceLock::print::DisplayBacktrace as core::fmt::Display>::fmt::haad32dbdee93b0c8
   1:          0x1180493 - core::fmt::write::h6ad3ba0dedbcf173
   2:          0x112674f - std::io::Write::write_fmt::ha3ff7d1996e99b72
   3:          0x1166163 - std::sys::backtrace::BacktraceLock::print::hb6e0ad5a95b804a0
   4:          0x11548ce - std::panicking::default_hook::{{closure}}::hdde2e5ae82f46e0d
   5:          0x1154793 - std::panicking::default_hook::h24772f302d635910
   6:          0x1154dff - std::panicking::rust_panic_with_hook::h89de30867df48bb1
   7:          0x1166886 - std::panicking::begin_panic_handler::{{closure}}::h98d82b909a800684
   8:          0x1166559 - std::sys::backtrace::__rust_end_short_backtrace::h0b5e9482c92da20f
   9:          0x115496d - __rustc[b723cc1e8dfe1f7e]::rust_begin_unwind
  10:           0x407a9d - core::panicking::panic_nounwind_fmt::h5d6ed55febaa6a52
  11:           0x407b32 - core::panicking::panic_nounwind::hb707c1bf767e5855
  12:           0x407cf2 - core::panicking::panic_cannot_unwind::h19ccdc9b77ce87cb
  13:           0x42f6ab - wgpu_native::default_uncaptured_error_handler::h842fbc0e6c0c5ad0
                               at /home/user/wgpu-native/src/lib.rs:408:1
  14:           0x4301fd - wgpu_native::ErrorSinkRaw::handle_error::h3c4a3c041da3a9b6
                               at /home/user/wgpu-native/src/lib.rs:551:25
  15:          0x11bb29f - wgpu_native::handle_error::hf96917feaba998e8
                               at /home/user/wgpu-native/src/lib.rs:632:5
  16:           0x438bb6 - wgpuDeviceCreateBindGroupLayout
                               at /home/user/wgpu-native/src/lib.rs:1865:9
  17:           0x409849 - main
                               at /home/user/wgpu-native/examples/texture_arrays/main.c:535:43
  18:     0x7ffff7c2a4d8 - __libc_start_call_main
  19:     0x7ffff7c2a59b - __libc_start_main_impl
  20:           0x40a645 - _start
  21:                0x0 - <unknown>
thread caused non-unwinding panic. aborting.
make: *** [Makefile:175: run-example-texture_arrays] Aborted (core dumped)

It seems that wgpu-native does not provide any way to change max_binding_array_elements_per_shader_stage. If I downgrade to v24.0.3.1 the example runs without issue. As well running the example from the wgpu repo using v25 works as expect but there max_binding_array_elements_per_shader_stage is set explicitly.

[almarklein]

IIUC, it does not work with our latest release, but it does work with current master. Has it been fixed? I'm not sure what you mean by max_binding_array_elements_per_shader_stage being set explicitly.

[Vipitis]

My understanding is the following. For v25 there has been a change in wgpu-core which introduces this new limit. The v25 release of wgpu-native doesn't allow users to set this limit so it's always the default value of 0. There needs to be changes to make this example (and the functionality) work again

[almarklein]

running the example from the wgpu repo using v25 works as expect

Ah, I initially read this as the wgpu-native repo.

@Vipitis that makes sense. There's no such limit in webgpu.h yet. All limits ending in 'PerShaderStage':

  uint32_t maxSampledTexturesPerShaderStage;
  uint32_t maxSamplersPerShaderStage;
  uint32_t maxStorageBuffersPerShaderStage;
  uint32_t maxStorageTexturesPerShaderStage;
  uint32_t maxUniformBuffersPerShaderStage;