we simply want to emit the debug information of eval() only if we are just debugging the program. Otherwise it’s just an unnecessary performance overhead.

This will not correctly handle the case when the user attaches a debugger after the eval has started, but before all execution artifacts (including callbacks/closures) are finished.