Check status in Bats

A Bats test needs to check that a command fails with a particular status, and a failing command normally stops the test.

cat > status.bats <<'EOT'
bats_require_minimum_version 1.5.0

@test "capture the status and output" {
  run bash -c 'echo out; echo err >&2; exit 3'
  [ "$status" -eq 3 ]
  [ "$output" = $'out\nerr' ]
}

@test "assert the status on the run line" {
  run -3 bash -c 'exit 3'
}

@test "run alone checks nothing" {
  run false
}

@test "a wrong status fails" {
  run false
  [ "$status" -eq 0 ]
}
EOT
bats status.bats
echo "bats exit: $?"

1..4
ok 1 capture the status and output
ok 2 assert the status on the run line
ok 3 run alone checks nothing
not ok 4 a wrong status fails
# (in test file status.bats, line 19)
#   `[ "$status" -eq 0 ]' failed
bats exit: 1

run executes the command, stores its exit code in $status and its combined standard output and error in $output, and always succeeds itself, so the test goes on to your own checks. The third test passes while testing nothing, because false is run and nobody looks at the result. With Bats 1.5 or later run -3 also asserts that the status is 3, and run ! asserts any non-zero status. Both are in the Bats documentation.

run starts the command in a subshell, so a variable it sets is lost and cd does not stick. A pipe is not part of the command: run a | b runs b on the output of run, so use run bash -c 'a | b'. Without bats_require_minimum_version 1.5.0 the -N form prints a warning. The dollar-quoted string in the first test needs Bash.