I'm trying to add some C++ code to the generated header coming form the MIDL compiler, using the cpp_quote attribute. But nothing gets added to the generated output, and I have no idea why.
I use VS 2019 with the C++/WinRT extension, the project is a C++/WinRT project with the default config.
Input IDL file:
cpp_quote("#define FOO")
namespace FooWinRT
{
runtimeclass Class
{
Class();
Int32 MyProperty;
}
}
Generated output:
// WARNING: Please don't edit this file. It was generated by C++/WinRT v2.0.200224.2
#pragma once
#include "winrt/FooWinRT.h"
namespace winrt::FooWinRT::implementation
{
template <typename D, typename... I>
struct __declspec(empty_bases) Class_base : implements<D, FooWinRT::Class, I...>
{
using base_type = Class_base;
using class_type = FooWinRT::Class;
using implements_type = typename Class_base::implements_type;
using implements_type::implements_type;
hstring GetRuntimeClassName() const
{
return L"FooWinRT.Class";
}
};
}
namespace winrt::FooWinRT::factory_implementation
{
template <typename D, typename T, typename... I>
struct __declspec(empty_bases) ClassT : implements<D, Windows::Foundation::IActivationFactory, I...>
{
using instance_type = FooWinRT::Class;
hstring GetRuntimeClassName() const
{
return L"FooWinRT.Class";
}
auto ActivateInstance() const
{
return make<T>();
}
};
}
#if defined(WINRT_FORCE_INCLUDE_CLASS_XAML_G_H) || __has_include("Class.xaml.g.h")
#include "Class.xaml.g.h"
#else
namespace winrt::FooWinRT::implementation
{
template <typename D, typename... I>
using ClassT = Class_base<D, I...>;
}
#endif
(Of course the actual production code is more complex, this is just a minimal example of what I'm trying to do.)
What am I missing, why is it not doing what it's supposed to do?
The header "was generated by C++/WinRT v2.0.200224.2", i.e. cppwinrt.exe, not midl[rt].exe. The latter produces a .winmd file that's used by the former to emit the implementation header. I'm not aware of a way to instruct the tools to make cpp_quote
s show up in the implementation headers.