Trait CapStdExtCommandExt

Source

pub trait CapStdExtCommandExt {
    // Required methods
    fn take_fd_n(&mut self, fd: Arc<OwnedFd>, target: i32) -> &mut Self;
    fn take_fds(&mut self, fds: CmdFds) -> &mut Self;
    fn cwd_dir(&mut self, dir: Dir) -> &mut Self;
    fn lifecycle_bind_to_parent_thread(&mut self) -> &mut Self;
}

Expand description

Extension trait for std::process::Command.

Required Methods§

Source

fn take_fd_n(&mut self, fd: Arc<OwnedFd>, target: i32) -> &mut Self

👎Deprecated:

Use CmdFds with take_fds() instead

Pass a file descriptor into the target process at a specific fd number.

§Deprecated

Use CmdFds with take_fds instead. This method registers an independent pre_exec hook per call, which means multiple take_fd_n calls (or mixing with take_fds) can clobber each other when a source fd’s raw number equals another mapping’s target. take_fds handles this correctly with atomic fd shuffling.

Source

fn take_fds(&mut self, fds: CmdFds) -> &mut Self

Apply a CmdFds to this command, passing all registered file descriptors and (if configured) setting up the systemd socket-activation environment.

§Important: Do not use `Command::env()` with systemd fds

When systemd socket-activation environment variables are configured (via CmdFds::new_systemd_fds), they are set using setenv(3) in a pre_exec hook. If Command::env() is also called, Rust will build an envp array that replaces the process environment, causing the LISTEN_* variables set by the hook to be lost. Command::envs() is equally problematic. If you need to set additional environment variables alongside systemd fds, set them via pre_exec + setenv as well.

Source

fn cwd_dir(&mut self, dir: Dir) -> &mut Self

Use the given directory as the current working directory for the process.

Source

fn lifecycle_bind_to_parent_thread(&mut self) -> &mut Self

On Linux, arrange for [SIGTERM] to be delivered to the child if the parent thread exits. This helps avoid leaking child processes if the parent crashes for example.

§IMPORTANT

Due to the semantics of https://man7.org/linux/man-pages/man2/prctl.2.html this will cause the child to exit when the parent thread (not process) exits. In particular this can become problematic when used with e.g. a threadpool such as Tokio’s https://kobzol.github.io/rust/2025/02/23/tokio-plus-prctl-equals-nasty-bug.html.