I’ve recently been pushing for use of pull requests in our workflow at LunarLincoln. I believe code review to be an important aspect of development and I feel that pull requests promote code review with every feature added. This kind of attention to detail isn’t just for open source projects on Github.
My Bad
There is an issue to opening an, ideally, small pull request for every new feature: inevitably, a new feature will rely on a feature that exists in an open pull request that has not been approved yet. What do we do in this situation? Normally, I would just create my new feature branch from the open pull request branch and work away. I did not do that this time. I had not thought that far ahead. After a few commits on my new feature branch, I realized I needed the work from an open pull request branch. So I did the logical thing and merged. I did not include the --no-commit --no-ff. I unintentionally merged code from an open pull request onto my new feature branch. To make matters worse, I edited the merge commit and royally screwed up my feature branch. So much so that later, I was testing my new feature branch after having merged in the approved pull request, the changes from the approved pull request was missing. Yikes.
My Experience
What was I to do? I had never run into this situation before. My first thought was to create a new branch and manually copy over my changes, then open a pull request for this copied branch. This would surely work, but there were a few commits worth of work (about 8) and that seemed tedious.
After pondering for a bit, I did create a new branch. I then used git’s cherry-pick command to bring over the commits from my fubar’d branch and save face. I had never used the cherry-pick function, but I was aware of its existence from the docs and from previous developer given talks at meet ups. While I didn’t have direct experience with it, I was able to use my indirect experience and solve my problem.