panic when piping into a bad process?

[PlexSheep]

$ netpulse -d | return 1                 
thread 'main' panicked at std/src/io/stdio.rs:1117:9:
failed printing to stdout: Broken pipe (os error 32)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[PlexSheep]

I've noticed this is a problem with more of my programs (like hedu). It is not the case in ripgrep for example. The panic is in writing of the std lib (since writing fails in that case of course, even something like print!

[PlexSheep]

This seems to be due to a design difference between rust and c programs. When a rust program receives a SIGPIPE, it does nothing, a c program would get terminated however.

more to read

[PlexSheep]

Based on this: I seem to have two ways: register a signal handler for SIGPIPE, or just ignore this. Seeing that I already have a signal_hook function that I use in the daemon, I might just do the same for all things.

[PlexSheep]

Tried this approach, but it does not work.

fn signal_hook() {
    unsafe {
        signal::signal(Signal::SIGTERM, SigHandler::Handler(handle_signal))
            .expect("failed to set up signal handler");
        signal::signal(Signal::SIGPIPE, SigHandler::Handler(handle_signal))
            .expect("failed to set up signal handler");
        signal::signal(Signal::SIGINT, SigHandler::Handler(handle_signal))
            .expect("failed to set up signal handler");
    }
}

extern "C" fn handle_signal(signal: i32) {
    let _signal: nix::sys::signal::Signal =
        nix::sys::signal::Signal::try_from(signal).expect("got an undefined SIGNAL");
    {
        // the default behavior is terminating
        std::process::exit(1);
    }
}

The logic would work out, but you can't exit the program in a signal handler or panic there either without causing deeper errors. the handler is a function that cannot be unwinded, so panicing does not work (not that it would actually do anything), and just using exit() results in another panic somewhere in the std lib, so both unsuitable.

[PlexSheep]

I don't really know what to do with this to make the unfriendly error go away.

more conversation on the topic

[PlexSheep]

This seems to be the official issue for it in rust. I'm definitely not rewriting every print or checking some global variable from the signal handler for this.