Guidance for dealing with callbacks

[sagacity]

Suggestion

The windows::Windows::Win32::Graphics::Direct3D::Fxc::D3DCompile API call supports providing an implementation of the ID3DInclude interface, so that application developers can provide custom callbacks to handle #include file handling.

In C++ this works by creating a class and deriving it from ID3DInclude, like so:

class MyCustomInclude : public ID3DInclude
{
    HRESULT Open(D3D_INCLUDE_TYPE IncludeType, LPCSTR pFileName, LPCVOID pParentData, LPCVOID* ppData, UINT* pBytes) override
    {
        // ...get some bytes based on input parameters...
    }

    HRESULT Close(LPCVOID pData) override
    {
        // ...free buffers, etc...
    }
};

It's unclear to me how this needs to be handled in windows-rs, though. ID3DInclude is not modelled as a trait I can implement, so my next thought was to manually fill the ID3DInclude_Vtbl with Open and Close function pointers. However, it's unclear to me how to take that vtable, construct a new ID3DInclude from that and provide it to D3DCompile.

It would be nice if there would be a snippet of documentation or a sample showing how to deal with these interface implementations, but I'm not sure how often that comes up in other areas of the Windows APIs.

[kennykerr]

You may be missing the "implement" feature in the latest release. This requirement has since been removed (#3333) which should make this easier to discover in future.

[dependencies.windows]
version = "0.58"
features = [
    "implement",
    "Win32_Graphics_Direct3D",
]

use std::ffi::c_void;
use windows::{core::*, Win32::Graphics::Direct3D::*};

struct Include;

impl ID3DInclude_Impl for Include {
    fn Open(
        &self,
        _: D3D_INCLUDE_TYPE,
        _: &PCSTR,
        _: *const c_void,
        _: *mut *mut c_void,
        _: *mut u32,
    ) -> Result<()> {
        todo!()
    }
    fn Close(&self, _: *const c_void) -> Result<()> {
        todo!()
    }
}

fn main() {
    let _include = ID3DInclude::new(&Include);
}

[sagacity]

I did indeed miss that, thanks!

[sagacity]

Sorry to reopen this, but there is currently a caveat: the new method of ID3DInclude is behind a std feature gate, which doesn't seem needed afaict. It does create a Box so it does need extern crate alloc, as well. Can this feature gate be removed? It does seem to run fine without this gate.

That also explains why I missed it, since I'm building with no-std.

[kennykerr]

Yes, I'm planning to fix that: #3335

[sagacity]

Wow, thanks for all the hard work, it's very much appreciated.