Turn a failure into a controlled experiment

First, build a mental model

An error is an observation, not yet an explanation. Reduce the uncertainty with a reproducible case, a specific hypothesis and a test that could prove that hypothesis wrong.

1. Observe the exact symptom and the conditions in which it appears; distinguish the first error from later consequences.
2. Reproduce with a small fixture so the same input reliably demonstrates the problem.
3. Trace actual values at the boundary where the behavior first diverges from the requirement.
4. Change one relevant cause, then rerun the reproduction and neighboring cases.
5. Add a regression test for the intended behavior. A suppressed exception is not evidence that the underlying operation became correct.

const lines = [{ quantity: 2 }, { quantity: 3 }];
let units = 0;
for (let index = 0; index <= lines.length; index++) {
  units += lines[index].quantity;
}
console.log(units);

1. Index 0

   Read quantity 2 and add it to units.

2. Index 1

   Read quantity 3 and reach units = 5.

3. Index 2

   The condition still permits 2, but lines[2] is undefined. Reading quantity fails before the final output.