One issue I often have with `git bisect run` is that it’s prone to bark up the wrong tree when it trips over an unrelated error. In other words: one single unr

One issue I often have with `git bisect run` is that it’s prone to bark up the wrong tree when it trips over an unrelated error. In other words: one single unrelated error, even if fixed right away in the next commit, is enough to lead `git bisect run` down the wrong path. To get better results, I recommend using a shell script that has the following properties: - if the test fails and prints the same unique failure message as the bad commit prints, return an exit status of 1; - if the test fails without printing that unique message, return an exit status of 125 (instructing `git bisect run` to skip the commit); - otherwise, return an exit status of 0.