This appears to only happen on FreeBSD (tested on 11.3)
This does NOT appear on CentOS Linux 8.0

https://github.com/ogham/exa/blob/master/src/fs/dir.rs#L38

This is the exact location of the error. I don't really have any Rust skills to debug what is going on, or how to properly trap the error. If my very basic debugging skills are correct, the read_dir works, .map works (both of them), and it bails on something involving the .collect part.

Can you give the output of strace exa? Using the strace CLI tool.

as stated above, this is freebsd, not linux, so there is no strace utility.

We have ktrace, dtrace, and truss, though.

Relevant bit:

 12974 exa      CALL  getdirentries(0x3,0x8016ad000,0x1000,0x8016172e8)
 12974 exa      RET   getdirentries 840/0x348
 12974 exa      CALL  getdirentries(0x3,0x8016ad000,0x1000,0x8016172e8)
 12974 exa      RET   getdirentries -1 errno 22 Invalid argument

GNU ls works, but produces a warning from this error. BSD ls appears to swallow it somewhere.

If I filter out that error, exa works fine in a .zfs:

        let contents = fs::read_dir(&path)?
            .map(|result| result.map(|entry|entry.path()))
            .filter(|result| result.is_ok() || result.as_ref().unwrap_err().kind() != std::io::ErrorKind::InvalidInput)
            .collect::<Result<_,_>>()?;

Maybe more generally this code should return the entries it does manage to get and there should be a mechanism to warn about encountered errors, rather than discarding the entire thing.