I find myself in the following scenario with Git every so often:

1. From the main branch, I create and checkout branch A
2. On branch A, I do some work, create some commits, etc., then push my changes
3. I need to do some unrelated work, so I create and checkout branch B
4. I create some more commits on branch B and push
5. I go to create a PR and... oops. I Created branch B from branch A, so suddenly I have all these irrelevant commits from branch A in my PR. I needed to create branch B from the main branch instead.

Just wondering what the best workflow is here. Sure, I can delete branch B, create it again and replicate my commits, but that seems tedious. A rebase is not an option, as I have already pushed to a remote repo. Maybe I can revert all the commits that were carried over from branch A to B? Maybe I should just create an alias that will checkout main before creating the branch?

CodePudding user response：

Simplest solution: rebase onto

git rebase has the onto mode to do exactly what you want:

git rebase --onto main A B

will take B off of A and rebase it onto main.

If you have B checked out when you run this, you don't need to specify it on that rebase command line.

Cherry Pick option You can also do it with a cherry pick operation.

Create a new branch off of the right place, cherry pick all the commits from B onto it, and then reset B to it afterwards.

git checkout main  # or where the branch should start from
git checkout -b B-new  # pick a new name, you'll need access to both B and B-new
git cherry-pick A..B  # cherry pick all commits on B not on A

Now B-new is the branch you wanted to create in the first place. You can just use the new name, or if you want to keep the name B, reset it to B-new:

git checkout B
git reset --hard B-new

And clean up, since you no longer need B-new now:

git branch -d B-new

But all that is doing exactly what the rebase onto did...