Attempt to subtract with overflow when output exceeds terminal height

[smoelius]

I am observing attempt to subtract with overflow on this line: https://github.com/console-rs/indicatif/blob/40b40d29b6d06ae18c40b829f77e9c43bcebd7af/src/draw_target.rs#L534

self.lines.len() >= self.orphan_lines_count holds prior to entering this loop: https://github.com/console-rs/indicatif/blob/40b40d29b6d06ae18c40b829f77e9c43bcebd7af/src/draw_target.rs#L501

And real_len >= idx holds at the start of each iteration of that loop.

Those two facts causes me to squint at this if block: https://github.com/console-rs/indicatif/blob/40b40d29b6d06ae18c40b829f77e9c43bcebd7af/src/draw_target.rs#L517-L519

In particular, commenting out that if block causes my panics to go away.

I don't completely understand the purpose of that if block, but maybe it can go away?

(indicatif is a great tool, BTW.)

[smoelius]

Here are steps to reproduce the bug.

1. Create a debug build of Necessist, e.g.:

   git clone https://github.com/trailofbits/necessist
   cd necessist
   cargo build

2. Create a project with a test like this:

   #[test]
   fn many_lines() {
   let mut s = String::new();
   for _ in 0..100 {
       s.push('\n');
   }
   None::<()>.expect(&s);
   }

3. Within the project containing the above test, run Necessist:

   path-to-clone/target/debug/necessist

You should see something like this:

thread 'main' panicked at 'attempt to subtract with overflow', .../.cargo/registry/src/index.crates.io-6f17d22bba15001f/indicatif-0.17.6/src/draw_target.rs:534:28

[djc]

@smoelius thanks for the detailed issue and reproduction case. Unfortunately I don't have much time to dig into this in detail at the moment. The code you're referencing here was introduced in #563, maybe read up on that to figure out if/why the branch was necessary? Probably the easiest fix is to just slap a saturating_sub() on it, and I'm happy to merge that PR, but if you want to look into it a bit more to figure out the proper invariants (and add some comments to document them for future readers) that would of course be much appreciated!

[smoelius]

Thank you very much for providing the background. I will try to look into this more and propose a fix.