So you're hunting down a regression (or just a bug) and want to use
git bisect to find out when it was introduced? Smart kid.
If you have a shell command ready to reveal if your current state is good or bad, you can have git do most of the work for you.
git bisect run <your command> you can tell git that your command will reveal the issue; git on the other hand will use the return value of that call to decide if the state is good or bad.
First, start bisecting
git bisect start
Then tell git which revisions are good and which are bad. You need to manually run your command for this, or just know which ones work and don't work.
git bisect bad git bisect good abcdef1234
Now tell git what to do
git bisect run bundle exec cucumber features/my.feature:42
Wait for the results to come in: Git will bisect until it knows which commit was the first bad one.
For more information, see the original "git bisect" card.
Mind that your command needs to work across all revisions that you will be bisecting. If you're referencing a test file that's under version control, make sure you always run the correct test; copying the file to a location outside of your repository might be a viable solution to keep the line number steady while you time-travel between commits.