Errors setup teardown still not caught correctly

[Chemaclass]

📚 Description

Closes #512

Fixed lifecycle hooks not catching failing commands (non-zero exit codes) in set_up, tear_down, set_up_before_script, and tear_down_after_script.

Problem: Failing commands like git checkout non-existing-ref, ls /non/existing/path, or false were silently ignored in lifecycle hooks, allowing tests to run even when setup failed.

Root cause: Using if ! command; then status=$? - the $? inside the if block returns 0 (success of the conditional) instead of the command's actual exit code.

Solution: Changed to command || exit $? to capture the exit code immediately.

🔖 Changes

• Fixed src/runner.sh:349 - Changed if ! runner::run_set_up to runner::run_set_up || exit $?
• Added test coverage for all lifecycle hooks with failing commands
• Updated CHANGELOG.md

The Real Fix

Both set_up and tear_down were broken for the same reason - they both call runner::execute_test_hook which was working correctly. The actual bug that affected BOTH was that runner::execute_test_hook wasn't catching failing commands.

But once runner::execute_test_hook returns the correct error code:

• set_up needed the fix at line 349 to properly exit on error
• tear_down was already correctly using || teardown_status=$? in the trap

So both are fixed now, but through different mechanisms:

1. The common fix: runner::execute_test_hook now correctly detects all failing commands
2. The set_up fix: || exit $? ensures we exit when setup fails
3. The tear_down "fix": It was already correctly written with || teardown_status=$?

✅ To-do list

• I updated the CHANGELOG.md to reflect the new feature or fix
• I updated the documentation to reflect the changes

[carlfriedrich]

@Chemaclass Thanks a lot for the super quick fix!

I can confirm that bashunit now correctly displays errors and does not execute the test function when the set_up function returns an error code. This is a big win! 👏

However, we still have to handle errors explicitly. The following set_up function will still silently pass:

function set_up() {
  false
  true
}

In order to detect the false statement, we have to:

function set_up() {
  false || return -1
  true
}

IMO this is not ideal, because it bears the risk of potential errors being overlooked and set_up functions becoming overloaded with error handling (which is usually not what you want in a test environment).

However, I am open for a discussion here, since at least it is possible now to detect errors in these functions. I think that we should detect any failing statement, though, in order to force robust and clean test implementations.

[Chemaclass]

@carlfriedrich, thanks for the feedback, I implemented the fix here #515 | release 0.27