Discrepancy between documentation and implementation for WSL #218
Description
In the README section about WSL it is explicitly stated that headers in the /include/wsl folder are explicitly not following LLP64 (Windows) but LP64 (Unix-likes, eg. Linux, BSD, etc) standard when it comes explicitly LONG and ULONG.
On 64-bit Windows, a long is 4 bytes, but on Linux it is typically 8 bytes. The D3D12 ABI for WSL uses long and therefore these should be 8 bytes.
Yet in /include/wsl/stubs/basetsd.h it explicitly hard codes LONG and ULONG to LLP64.
// Note: using fixed-width here to match Windows widths
// Specifically this is different for 'long' vs 'LONG'
typedef uint32_t UINT32, UINT, ULONG, DWORD, BOOL, WINBOOL;
typedef int32_t INT32, INT, LONG;This is tangential related to #156 as currently DirectXShaderCompiler links to the four year old commit 980971e which predates the existence of basetsd.h.
The current version of DirectXShaderCompiler compiles just fine against a Linux libc but compiling against a newer version of DirectX-Headers causes errors due to incompatible redefinitions between this repositories header /include/wsl/stubs/basetsd.h and their header /include/dxc/Support/D3DReflection.h. With their version adhering to what this repositories README.md states (which by the way predates the additions of basetsd.h).
This isn't me saying "Please fix /include/wsl/stubs/basetsd.h" but rather me asking if basetsd.h is wrong or if a workaround for LP64 based systems should be written.
PS: I am aware of:
WSL support is not intended for general purpose application development.
However since DirectXShaderCompiler is also a Microsoft repository and explicitly mentions compilation under WSL/Linux I am asking for clarifications. There are other incompatibilities between these two headers that probably should be changed in the DXC repository. (Probably either reducing code duplication or fully porting/include/wsl/stubsfunctionality)