ls: Improve Error Handling

[kimono-koans]

I've been trying to improve the error handling of 'ls' and have made some progress in describing and printing the errors, but I'm stuck trying to understand how/when uucore decides to print an error message. All stderr messages print before stdout messages, which is undesirable for 'ls.'

What is the best way to get interleave stdout and stderr output like GNU ls?

I'm sure this is a rookie mistake, but @tertsdiepraam could you assist?

[tertsdiepraam]

Hi! I confirmed the behaviour that you described. I don't think the uucore error functions contribute to this problem. There is a loop like this (simplified a bit), which I think is the culprit of this:

for loc in &locs {
    let p = PathBuf::from(loc);
    let path_data = PathData::new(p, None, None, &config, true);

    if path_data.md().is_none() {
        show_error!(
            "cannot access {}: No such file or directory",
            path_data.p_buf.quote()
        );
        set_exit_code(1);
        continue;
    }
    files.push(path_data);
}
sort_entries()
display_entries()

(see here for the full loop)

After which the files get sorted and only afterwards get printed. show_error is simply a an eprintln which prepend the name of the util, so these are equivalent:

eprintln("ls: {}", "error");
show_error("{}", "error");

For more information on the error handling, you could take a look at the documentation or just post your questions here.

[tertsdiepraam]

understand how/when uucore decides to print an error message.

I guess I should expand on this a little bit. If an error is passed to show_error, it is printed immediately. The only "magic" thing that happens is that errors that are returned from uumain a printed at the end (this is done by the gen_uumain proc macro, the source of which is here)

[kimono-koans]

Thank you! The level of magic was exactly what I was wondering about. I think I tracked it down -- looks like we just need to do a buffer flush after an item is displayed!

[tertsdiepraam]

I think has been resolved by #2809