Git unstage without losing merge?

Question

I have a branch with hundreds of new and different files and I just want to merge a dozen or so of these into my release branch. I tried running a normal git merge with no commit and that automatically staged hundreds of files I don't want (in addition to finding dozens of conflicts that require manual merging.) I then tried to do a git revert to unstage all of the automerged files and just add the ones I want back to the index, but that also aborted the merge and leaves all the conflict markers in the files. I thought I could just run the mergetool but it now doesn't recognize the conflict markers...

Am I going about this the totally wrong way...or is there something I'm missing?

Oversell

Merge and revert talk about commits, staging applies to the index. What exactly are you referring to where? It’s not clear what you are doing. – Lole

I tried to clarify a bit - but I already like Manzur's answer # 1 below. – Oversell

Answer 1

It's as simple as using git reset. I was using git stash pop and had a merge conflict and it also staged some files with changes by me and merge. I resolved the conflict and unstaged all the files with the following command:

git reset HEAD <file_1> <file_2> ... <file_n>

Twinkle

Actually I found you don't need to specific the file(s), just do git reset HEAD. – Twinkle

Answer 2

The question is a bit ambiguous. I'll try to rephrase it and answer accordingly. Say you want to merge only the files(file2, file3) from the branchA:

1) Get the tree sha1 of the branch commit:

git cat-file -p branchA

2) Checkout the changes under that tree for file2 and file3:

git reset <sha1> -- file2 file3

3) Commit this changes:

git commit -m 'Merge of branchA'

If you've already message with the staging, do(in case all the needed changes are committed ahead of time): git reset --hard

Puttee

Manzur- thanks, that's perfect. I'll still need to check for conflicts in these files but I can do that when merging back to the release branch. At least this simplifies the merge to just the files I care about. – Oversell

Experimenting with this now... it seems git checkout works better than git reset in your step 2 manzur because then I don't have the original versions showing up as unstaged changes. Also I can just use the branch name instead of the <sha1>. Make sense? – Oversell

Sure, you can use sha1 instead of branch name. They're interchangeable for almost all cases(look at the .git/refs/heads/<branchname> it contains sha1). – Puttee

regarding checkout vs reset it depends. reset is better when you know what happens on the index and filesystem. checkout is more simple, you can checkout a file to other directory and copy-paste the changes you want. – Puttee

Answer 3

Similar to Chef Pharaoh's answer, but shorter and more ergonomic:

git reset "*"

(The quotes are there to avoid shell expansion.)

Ferromanganese

Answer 4

I have a branch with hundreds of new and different files and I just want to merge a dozen or so of these into my release branch.

Then make a commit with just those dozen or so files changed, and merge that.

git checkout -b frotzybitsoffeature $(git merge-base feature release)
git checkout feature -- thisfile thatfile theseothers/
git commit -m 'selecting the frotzy subset of feature'

git checkout release
git merge frotzybitsoffeature

git checkout feature            # now tell git the frozybits work is all still in feature
git merge frotzybitsoffeature
git branch -d $_

Illimani

Answer 5

To revert an exact merge commit, and making all the changes as staged, below command can be used.

git revert -m 1 <commit-hash>

Basidiospore