When you start a new project in GitHub or GitLab as part of a team, it’s easy to just push the first version of the code to your trunk/main branch and then be left with no easy way to create a Pull Request or Merge Request that is an accurate view of your changes. This makes it harder for colleagues to comment on specific parts of the code, and for you to test the CI/CD that runs on branches.

To avoid this, start by committing everything that you’ve worked on:

git add -A .
git commit -a -m "my lovely work, v0.1"

Rename your current trunk/main branch, assuming it’s called main:

git branch -m my-feature-branch

Then create a fresh new main with no commits:

git checkout --orphan main
git rm -rf .

Create an empty initial commit:

git commit --allow-empty -m "initial commit"

DANGER: Now we’re going to forcefully push up the new empty branch, overwriting what is already in our remote. Bookmark https://ohshitgit.com if you’re so inclined. Worse case scenario, it’s always faster to write the second time, right?!

git push -f --set-upstream origin main

Switch over to the feature branch you renamed above:

git reset --hard HEAD
git checkout my-feature-branch

Rebase it onto your new, empty, main branch:

git rebase main

And push ready for a Pull or Merge Request:

git push --set-upstream origin my-feature-branch

What if I want to avoid this pain next time?

Create a folder for your new project, initialise git and then create your first commit:

git init
git commit --allow-empty -m "initial commit"

Then start a feature branch:

git checkout -b my-feature-branch

Much easier!

References

• https://gist.github.com/ozh/4734410