You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Issues that can be raised with wgpu (marked items are patched in a local copy of wgpu):
InstanceInterface::create_surface returns a wgpu::CreateSurfaceError, but these can't be constructed other than by a conversion from wgpu_core::CreateSurfaceError.
custom RequestAdapterError
RenderPassDescriptor generic over label, with map_label
RenderPassInterface::end is never called. core implements its own Drop that calls it. But shouldn't this be called by the wrapper?
TextureViewDescriptor::map_label
TextureViewInterface::create_view has no knowledge of the texture's descriptor. It would need to store it, but the wrapping type does that already.
InstanceInterface::wgsl_language_features is behind a feature flag. This means when I implement the interface without the flag, but another crate in the dependency graph enables it, it breaks compilation of my crate.
InstanceInterface::new doesn't serve a purpose?
naga really needs something that returns TypeLayout for a TypeInner. type resolution sometimes can't give you a handle, so Layouter won't always do.
naga: make naga::ir::BindingCopy. This might be useful, but we don't need it anymore.
naga: bools have size and alignment 1. See naga_interpreter::tests::naga_bool_width_is_32bit.
naga: way to convert Alignment to NonZeroU32 (or anything usable). cranelift wants to know this, just using round_up won't always do. Alternative: just offer a getter for the log2 of the alignment
map_label for various descriptor structs (e.g. SamplerDescriptor)
Issues that can be raised with
wgpu(marked items are patched in a local copy of wgpu):InstanceInterface::create_surfacereturns awgpu::CreateSurfaceError, but these can't be constructed other than by a conversion fromwgpu_core::CreateSurfaceError.RequestAdapterErrorRenderPassDescriptorgeneric over label, withmap_labelRenderPassInterface::endis never called. core implements its own Drop that calls it. But shouldn't this be called by the wrapper?TextureViewDescriptor::map_labelTextureViewInterface::create_viewhas no knowledge of the texture's descriptor. It would need to store it, but the wrapping type does that already.InstanceInterface::wgsl_language_featuresis behind a feature flag. This means when I implement the interface without the flag, but another crate in the dependency graph enables it, it breaks compilation of my crate.InstanceInterface::newdoesn't serve a purpose?TypeLayoutfor aTypeInner. type resolution sometimes can't give you a handle, soLayouterwon't always do.naga::ir::BindingCopy. This might be useful, but we don't need it anymore.naga_interpreter::tests::naga_bool_width_is_32bit.AlignmenttoNonZeroU32(or anything usable). cranelift wants to know this, just usinground_upwon't always do. Alternative: just offer a getter for the log2 of the alignmentmap_labelfor various descriptor structs (e.g.SamplerDescriptor)