Panic when fetching grammars fail

[Frojdholm]

Summary

When fetching/building grammars it's done in a thread pool where the results are communicated over a channel. On the first failure the receiver is closed and the error returned. This causes any sends that aren't completed yet to panic, spamming the terminal with jumbled error messages.

The panicing function is this (annotations mine)

fn run_parallel<F>(grammars: Vec<GrammarConfiguration>, job: F, action: &'static str) -> Result<()>
where
    F: Fn(GrammarConfiguration) -> Result<()> + std::marker::Send + 'static + Copy,
{
    let pool = threadpool::Builder::new().build();
    let (tx, rx) = channel();

    for grammar in grammars {
        let tx = tx.clone();

        pool.execute(move || {
            tx.send(job(grammar)).unwrap(); // <- this panics since rx is closed
        });
    }

    drop(tx);

    // TODO: print all failures instead of the first one found.
    rx.iter()
        .find(|result| result.is_err())
        .map(|err| err.with_context(|| format!("Failed to {} some grammar(s)", action)))
        .unwrap_or(Ok(())) // <- on the first error this returns, closing rx and causing any subsequent sends to fail
}

One solution could be to simply ignore the Results produced by tx.send (#2641). Another could be to use something like https://docs.rs/beau_collector/latest/beau_collector/ to collect all the errors into a single error that can be returned.

Reproduction Steps

$ mkdir runtime_test
$ sudo chown root:root runtime_test
$ HELIX_RUNTIME=$PWD/runtime_test hx -g fetch

Helix log

No response

Platform

Linux

Terminal Emulator

kitty

Helix Version

22.05-14-gfc666db6