ext: Replace terminal_size with comfy-table

[cgwalters]

I was looking at our vendoring set and while it's not actually relevant I found myself wondering why we had three versions of windows-sys. Having that many crate versions is often a signal that there's an unmaintained dependency.

And indeed, terminal_size is no longer cool. The "in" crowd has moved on to newer, hipper things. Life moves fast, we need to keep up.

(OK but yes also this drops some manual column printing code
we had which is also a win)

[omertuc]

Trying to test this PR, I built it into a bootc image, bootc switch'd into it, then I was wondering if I was getting the ostree-rs-ext version from this repo or some other one:

[omer@hal9000 ~]$ ls -lah /usr/libexec/libostree/ext/ostree-container
lrwxrwxrwx. 1 root root 23 Jan  1  1970 /usr/libexec/libostree/ext/ostree-container -> ../../../bin/rpm-ostree

Why is this pointing at /usr/bin/rpm-ostree? I would've expected it to point at a binary related to ostree-ext?

[cgwalters]

Why is this pointing at /usr/bin/rpm-ostree? I would've expected it to point at a binary related to ostree-ext?

That's exactly it - right now rpm-ostree is vendoring the ostree-ext code in Fedora. The recent merge here will let us switch that over to bootc (and we'll make rpm-ostree hard Requires: bootc).

If you want to test this you'll need to explicitly take over the symlink, via make install-ostree-hooks at build time or just on a live system bootc usroverlay && ln -sfr /usr/bin/bootc /usr/libexec/libostree/ext/ostree-container

[cgwalters]

Hmm though actually in our own CI jobs we should start taking over the hooks now. I'll do a PR for that

[omertuc]

Not sure which is worse 😆

But seriously, without | head the older version doesn't have this weird newline issue

[cgwalters]

The | head thing is an old, old problem... rust-lang/rust#46016

[cgwalters]

I'm not getting the extra newlines here...I did sanity check this too before pushing

[omertuc]

Seems like it only happens if the terminal is smaller than the largest line

size.mp4

EDIT: jinx

[cgwalters]

Ahh wait I do get the newlines if one of the output lines is long enough to wrap past the terminal size. I think this may be fixable with https://docs.rs/comfy-table/latest/comfy_table/enum.ContentArrangement.html ?

[omertuc]

Ahh wait I do get the newlines if one of the output lines is long enough to wrap past the terminal size. I think this may be fixable with docs.rs/comfy-table/latest/comfy_table/enum.ContentArrangement.html ?

Yeah

use comfy_table::{presets::NOTHING, Table};

fn main() {
    let mut table = Table::new();

    table
        .load_preset(NOTHING)
        .set_header(vec!["REPOSITORY", "TYPE"])
        .set_content_arrangement(comfy_table::ContentArrangement::Dynamic);

    table.add_row(vec!["foo", "bar"]);
    table.add_row(vec![
        "fooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo",
        "baaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaar",
    ]);

    println!("{table}");
}

Seems well behaved

[omertuc]

use std::env::args;

use comfy_table::{presets::NOTHING, Table};

fn main() {
    let mut table = Table::new();

    table
        .load_preset(NOTHING)
        .set_header(vec!["REPOSITORY", "TYPE"])
        .set_content_arrangement(match args().nth(1).as_deref() {
            Some("d") => comfy_table::ContentArrangement::Disabled,
            Some("dyn") => comfy_table::ContentArrangement::Dynamic,
            _ => comfy_table::ContentArrangement::DynamicFullWidth,
        });

    table.add_row(vec!["foo", "bar"]);
    table.add_row(vec![
        "fooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo",
        "baaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaar",
    ]);

    println!("{table}");
}

Not sure I understand the difference between Dynamic and DynamicFull, didn't make a diff in my toy demo