Is there a way to rebase a single commit from a branch onto another branch?
I have this branch structure:
-- -- -- -- -- (Master) \ -- -- -- -- -- XX (Feature-branch) All I want to do is to rebase the last commit of Feature-branch onto master and rollback Feature-branch one commit.
-- -- -- -- -- XX (Master) \ -- -- -- -- -- (Feature-branch) How do I do that?
git rebase --onto master branch~1 branch This says "rebase the range of commits between last-before-branch and branch (that is, XX commit) on the tip of master branch"
After this operation branch tip is moved on commit XX, so you want to set it back with
git checkout branch git reset --hard branch@{1}^ Which says "reset the branch tip to the commit before its previous state"
So a cherry pick is a simpler solution...
--onto before so I may be doing something wrong. BTW the OP said rebase but it seems like he wants to do a cherry-pick.
git rebase --onto master X branchYou can cherry-pick XX to master.
git checkout master git cherry-pick <commit ID of XX> And remove the last commit from the feature branch with git reset.
git checkout Feature-branch git reset --hard HEAD^ 
cherry-pick made them appear as if they were deleted from the old location and created at the new location. I suppose rebase would have taken care of that, but by now I've pushed upstream so I can't test that. In any case, beware if you have a similar situation.
Feature-branch to origin you'll need to git push -f origin Feature-branch as your Feature-branch is now considered to be 1 commit behind the origin/Feature-branch.
It's pretty simple to do actually. The solution is to do an interactive rebase and "drop" all of the commits you don't want to include in the rebase.
git rebase -i <target_branch> where target_branch is the branch you want to rebase on to
Then you will edit the file that is opened and pick the commits you do want and drop (or d for short) all the commits you don't want to bring along.
git rebase -i is my favourite Git command. It is so flexible and in the same time so simple and clear exactly what is happening. You don't have to memorize any other rebase command if you learn git rebase -i.
@Charles response is correct. Anyway I ended up using this so many times, most of all to rebase specific config on a project
* a8f9182 (HEAD -> production) production configuration | * daa18b7 (pre) preproduction configuration |/ | * d365f5f (local) local configuration |/ * 27d2835 (dev) amazing new feature that will save the world * | 56d2467 (master) boring state of the art for project |/
that I create a new command for it:
$ cat ~/bin/git-rebaseshot COMMIT=$1 DEST=${2:-HEAD} git rebase ${COMMIT}^ ${COMMIT} --onto $DEST normally you want to autocomplete branch names for that command, so add it sourcing this function (adding to .bashrc or .profile):
_git_rebaseshot () { __gitcomp_nl "$(__git_refs)" } git autocomplete will search for it
you can use this command like this:
# rebase config on prepro on actual HEAD $ git rebaseshot prepro # rebase config on local onto dev $ git rebaseshot local dev # rebase production config on master $ git rebaseshot pro master When you divide features correctly, possibities are endless.
* a8f9182 (HEAD -> postgres) BBDD config * a8f9182 (local) local config * a8f9182 (debug) log level config * a8f9182 (dev) new feature |
I guess this is what quilt people like to do.
this command will work anyway with whatever sha/ref you provide:
$ git rebaseshot <Feature branch> master $ git rebaseshot <commit of XX> master 
