Retrospectively creating a branch in GIT

Posted March 20, 2012 by Tim Sutton & filed under GIT.

I often do this – I work away on my master branch and suddenly realise ‘damn I should have branched before I started this’. Usually this happens because what I think will be a trivial change snowballs into something much more sweeping. I thought I would record here how to deal with situations like this by retroactively creating a branch and rolling master back to a previous state. Note: backup your data first and this is only ideal if you haven’t pushed changes to a remote repo yet.

So here was the situation. My last ‘known good’ state was 8399227db and I wanted to move all subsequent work out of master and into a branch. So first lets create that branch starting from the last known good state:

git branch refactor_branch 8399227db

Next I check out the branch and merge all the changes from master into it:

git checkout refactor_branch
git merge master

Now I switch back to master and chuck away all changes made since the branch point (did I mention you should backup your work first?):

git checkout master
git reset --hard 8399227db

So I rebased the master branch to an older commit. Now I push my branch up to the remote repository and breath a sigh of relief that my work is backed up in the cloud and master remains usable for others working on it:

git push origin refactor_branch

As a last couple of steps, I remove my local branch and checkout the remote branch as a tracking branch:

git branch -D riab_to_inasafe
git branch --track riab_to_inasafe origin/riab_to_inasafe
git checkout riab_to_inasafe

Git being git, there are probably a thousand other ways to accomplish the same thing, but the above process works for me!

2 Responses to “Retrospectively creating a branch in GIT”

Nathan Woodrow

Are you sure git rebase SHA1 works? It just tell me the current branch is up to date.

I find using:

git reset –hard {SHA1}

works if you need to move the HEAD commit back.
March 20th, 2012
- Tim Sutton
  
  Hi
  
  Sorry you are absolutely right it was probably a cut & paste error. Here is a minimal test showing reset –hard doing its job:
  
  timlinux@ultrabook:/tmp$ mkdir test timlinux@ultrabook:/tmp$ cd test timlinux@ultrabook:/tmp/test$ git init Initialized empty Git repository in /tmp/test/.git/ timlinux@ultrabook:/tmp/test$ echo "Hello" > readme timlinux@ultrabook:/tmp/test$ git add readme timlinux@ultrabook:/tmp/test$ git commit -m "Version 1" [master (root-commit) 6ef3224] Version 1 1 files changed, 1 insertions(+), 0 deletions(-) create mode 100644 readme timlinux@ultrabook:/tmp/test$ echo "World" > readme timlinux@ultrabook:/tmp/test$ git commit -m "Version 2" -a [master 9162717] Version 2 1 files changed, 1 insertions(+), 1 deletions(-) timlinux@ultrabook:/tmp/test$ git branch testbranch timlinux@ultrabook:/tmp/test$ git log commit 91627170c4ff126dc8eda55f83ec961d904b31f6 Author: Tim Sutton Date: Thu Mar 22 12:14:08 2012 +0200
  Version 2 commit 6ef3224f61adb6424d9a6dacf5febd29324f53bb Author: Tim Sutton Date: Thu Mar 22 12:13:42 2012 +0200
  Version 1 timlinux@ultrabook:/tmp/test$ git reset --hard 6ef3224f61adb6424d9a6dacf5febd29324f53bb HEAD is now at 6ef3224 Version 1 timlinux@ultrabook:/tmp/test$ cat readme Hello timlinux@ultrabook:/tmp/test$ git checkout testbranch Switched to branch 'testbranch' timlinux@ultrabook:/tmp/test$ cat readme World timlinux@ultrabook:/tmp/test$ git merge master Already up-to-date. timlinux@ultrabook:/tmp/test$ cat readme World timlinux@ultrabook:/tmp/test$ git checkout master Switched to branch 'master' timlinux@ultrabook:/tmp/test$ cat readme Hello
  
  I will adjust the article accordingly.
  March 22nd, 2012