5.2How Bisect Works

You tell Git one "good" commit (where the test passed) and one "bad" commit (where the test fails). Git checks out the midpoint and asks you to evaluate it. Based on your answer, it narrows the range by half and repeats.

Commit history: A B C D E F G H I J K L M N O P
                good?                         bad

Step 1: Git checks out H (midpoint)
  → You test: PASS → good
  Range narrows to: I J K L M N O P

Step 2: Git checks out L (midpoint of remaining)
  → You test: FAIL → bad
  Range narrows to: I J K L

Step 3: Git checks out J (midpoint)
  → You test: PASS → good
  Range narrows to: K L

Step 4: Git checks out K
  → You test: FAIL → bad
  Result: K is the first bad commit

Four steps to find the problem among 16 commits.