How can i use neovim and coc.nvim for develop windows c++ apps on linux

Question

I develop c++ apps on linux and i use neovim with coc.nvim and coc-clangd plugins. I want to develop an app for windows but i comfort with linux and neovim so i want to use them for it. But i get some include errors with some windows headers (etc. "windows.h").

I use linux only for writing the code and i'll compile the program on windows. How can i prevent this errors and use windows headers with coc.nvim?

Answer 1

i'll compile the program on windows

You can cross-compile it from Linux. It's only marginally more difficult than getting the code completion to work.

1. Get the standard library headers (and libraries, if you want to cross-compile) from MinGW.

   Your package manager might have those, or you can get them from https://winlibs.com/ .

   I prefer getting those from MSYS2, and made scripts to automate this (since MSYS2 is otherwise Windows-only):

    git clone https://github.com/holyblackcat/quasi-msys2 cd quasi-msys2/ make install _gcc

2. Figure out the Clang flags needed to cross-compile.

   Unlike GCC, which for every target platform requires a separate compiler distribution, Clang is inherently a cross-compiler. You only need a single Clang distribution to compile for any supported platform.

   Download Clang from your package manager, and point it to the freshly downloaded headers and libraries.

   Following flags work for me: clang++-14 1.cpp --target=x86_64-w64-mingw32 --sysroot=/path/to/quasi-msys2/root/mingw64 -fuse-ld=lld-14 -pthread -stdlib=libstdc++ -femulated-tls -rtlib=libgcc .

   --target and --sysroot are crucial. The latter needs to point to the files you've downloaded. The remaining flags are less important.

   Running this should produce a.exe , runnable with wine a.exe .

3. Feed the same flags to Clangd.

   There are several ways to set compiler flags for Clangd.

   The easiest one is to create a file named compile_flags.txt in your project directory, and put the flags into it, one per line:

    --target=x86_64-w64-mingw32 --sysroot=/path/to/quasi-msys2/root/mingw64 -fuse-ld=lld-14 -pthread -stdlib=libstdc++ -femulated-tls -rtlib=libgcc

   Then Clangd should do the right thing for any source files in this directory.

---

Apparently, my Quasi-MSYS2 can somewhat automate this.

After running the commands above ( make install _gcc and others), run make env/shell.sh , and run your editor from this shell.

Replace compiler_flags.txt with compiler_commands.json with following contents:

[
    {
        "directory": "/your/sources",
        "file": "/your/sources/1.cpp",
        "command": "win-clang++ 1.cpp"
    }
]

Where win-clang++ is a Clang wrapper I ship, which automatically adds the flags I listed above.

Configure your editor to add following flag to Clangd: --query-driver=/path/to/win-clang++ (use which win-clang++ from quasi-msys2 shell to get the full path).

This makes Clangd obtain the right flags automatically from this wrapper.

Answer 2

You can't use windows.h while you're compiling a Linux native application. If want to make your application platform ready and you're using some kind of OS native cals, then you have to probably use defines like #if _WIN32 / __linux__ and so on. At the end, you can cross-compile your application to Windows while you're running on Linux as well.