Right now, the tests work by comparing the pretty printed output of parsing the test script, and the pretty printed output of that first roundtrip parsed again. However, that may miss certain bugs. For example, using bash:
case 1 in
1) echo "1";&
2) echo "2";;
esac should print:
since the first case should fall through to the second. However, there used to be a bug where it was instead parsed as:
case 1 in
1) echo "1";;
2) echo "2";;
esac with no fall through. This passed the tests since the first and second round trips are the same, despite the different semantics. Adding additional tests to compare the execution would fix this. I don’t think this would cause an issue, as long as we filter out any tests that don’t terminate or are otherwise misbehaved.
Right now, the tests work by comparing the pretty printed output of parsing the test script, and the pretty printed output of that first roundtrip parsed again. However, that may miss certain bugs. For example, using bash:
should print:
since the first case should fall through to the second. However, there used to be a bug where it was instead parsed as:
with no fall through. This passed the tests since the first and second round trips are the same, despite the different semantics. Adding additional tests to compare the execution would fix this. I don’t think this would cause an issue, as long as we filter out any tests that don’t terminate or are otherwise misbehaved.