behave

I built behave to check that a rewrite of a command-line tool behaves exactly like the original. You record cases (arguments, stdin, files) and it saves the original's exact stdout, stderr and exit code as YAML specs. Then you replay the same specs against the new implementation and read the diff. Running it against uutils coreutils (the Rust rewrite of GNU coreutils) has so far led to 26 upstream bug reports and 33 fix PRs, 17 of them merged. Free and MIT licensed, single Go binary.