Command Execution

ApiV1::new_command builds a Linux command to run in a WSL session. The program path and arguments must be valid for the target Linux environment:

#![allow(unused)]
fn main() {
extern crate wslplugins_rs;
use std::ffi::OsString;
use wslplugins_rs::prelude::*;
use wslplugins_rs::windows_core::HRESULT;

const E_FAIL: HRESULT = HRESULT(0x80004005u32 as i32);

fn read_kernel_version(
    context: &WSLContext,
    session: &WSLSessionInformation,
) -> PluginResult<String> {
    let mut stream = context
        .api
        .new_command(session, "/bin/cat")
        .with_arg("/proc/version")
        .execute()?;

    let mut kernel_version = String::new();
    std::io::Read::read_to_string(&mut stream, &mut kernel_version).map_err(|error| {
        PluginError::with_message(E_FAIL, OsString::from(error.to_string()))
    })?;

    Ok(kernel_version)
}
}

The returned stream is connected to stdin and stdout. Drop it when no more interaction is needed.

By default, WSLCommand targets the system/root namespace through the WSL API ExecuteBinary. This namespace is not the same as a user distribution. Microsoft documents it as a minimal Mariner-based root filesystem backed by writable tmpfs, so changes made there disappear when the WSL2 VM shuts down.

Use a user-distribution target when the command must run inside a distribution rather than the root namespace. That path uses ExecuteBinaryInDistribution, which is represented by WSLVersionCapability::ExecuteBinaryInDistribution (2.1.2). If that command path is central to the plugin, declare the capability in #[wsl_plugin_v1(...)]. If it is optional, handle the error returned by execute().

The returned TcpStream is connected to the command’s stdin and stdout. It does not carry stderr; WSL sends stderr output to Linux dmesg.