LLVM missing error when running VSCode-extension

[styrix560]

I regularly get an error, that LLVM is missing when trying to run the VSCode extension. After it crashes 3 times in 5 minutes, the extension cannot be restarted due to the following compile error:

error: No suitable version of LLVM was found system-wide or pointed
              to by LLVM_SYS_150_PREFIX.
       
              Consider using `llvmenv` to compile an appropriate copy of LLVM, and
              refer to the llvm-sys documentation for more information.
       
              llvm-sys: https://crates.io/crates/llvm-sys
              llvmenv: https://crates.io/crates/llvmenv
   --> [user_dir]\.cargo\registry\src\index.crates.io-6f17d22bba15001f\llvm-sys-150.1.2\src\lib.rs:489:1
    |
489 | / std::compile_error!(concat!(
490 | |     "No suitable version of LLVM was found system-wide or pointed
491 | |        to by LLVM_SYS_",
492 | |     env!("CARGO_PKG_VERSION_MAJOR"),
...   |
499 | |        llvmenv: https://crates.io/crates/llvmenv"
500 | | ));
    | |__^

Meanwhile clearing the cargo-cache (with the cargo-cache extension and cargo-cache -a) enables the extension to run again until the next crash happens.

My cargo version is cargo 1.76.0-nightly (623b78849 2023-12-02) and my OS is Windows 10.

The error also occurs on my second windows-machine.

[ctiedt]

I don't have a Windows machine to test this on right now, but the inkwell backend is not supported on Windows yet. I'll see if I can reproduce the issue the next time I'm on a Windows machine. My guess is that for some reason the inkwell feature is active (it should be deactivated by default because of the limited platform support), so the language server fails to compile leading the extension to throw errors.

[styrix560]

Well that's the odd thing. It does compile the first time (or at least it doesn't throw an error), but after that the compile error is thrown. Maybe I could manually disable the inkwell feature? Alternatively, I also have WSL installed if that helps.

[skyz1]

I was facing the same issue when the inkwell backend was introduced. To solve it I compiled LLVM myself and set the environment variable.

[styrix560]

Yes, I will probably do that. However, nobody can realistically expect users to perform a full LLVM-installation.

[JonasWanke]

@ctiedt just started working on a new native backend based on Cranelift, which is written in pure Rust and is smaller than LLVM. (I created #886 to track this.) We're not yet sure about what we're going to do with the Inkwell/LLVM-based backend:

• keep it (and make it feature-complete) for, e.g., optimized release builds since the code it produces is generally faster than that of Cranelift but compile times are longer
• delete it