搞砸了Git分支，如何修复

Question

I am having some trouble with git (github), because I am new to this, my mind is thinking in SVN, and I guess I am just not getting this... please help.

So I was working on some branch_1 doing some changes and commits, then doing git push origin branch_1 (and making a pull request on that branch) then again doing some changes and commits.

then I decided to switch to other feature, so I did git checkout -b branch_2

I changed some file, did commit of that alone file, then: git push origin branch_2 But when I tried to make a pull request on branch_2 it was containing commits both from branch_2 and some commits from branch_1

My question is basically, what did I do wrong ( how is this usually correctly handled? ) And most importantly how do I fix this now? I need to have branch_2 with just one commit of last changes I did on it, and to do a pull request for it with that one commit.

Basically what I need is to work on different branches switching back and forth, doing commits accordingly and doing pull requests for each branch with work done only on appropriate branch.

Thanks a lot!

Answer 1

So let's see what happened here with an example. The following is a replication of what I think you did:

$ git init && touch README && git add README && git commit -m 'Initial commit'
Initialized empty Git repository in /home/peter/ex/.git/
[master (root-commit) 56c1728] Initial commit
 0 files changed
 create mode 100644 README
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit       
* 56c1728 (HEAD, master) Initial commit
$ git checkout -b branch1 
Switched to a new branch 'branch1'
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* 56c1728 (HEAD, master, branch1) Initial commit

git checkout -b <new_branch> will create a branch wherever HEAD is. See how branch1 is now pointing to the same commit that HEAD was.

Now let's make some commits.

$ touch A
$ git add A
$ git commit -m 'Add A'
[branch1 298c3f9] Add A
 0 files changed
 create mode 100644 A
$ touch B
$ git add B
$ git commit -m 'Add B'
[branch1 24ffff3] Add B
 0 files changed
 create mode 100644 B
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* 24ffff3 (HEAD, branch1) Add B
* 298c3f9 Add A
* 56c1728 (master) Initial commit

So now, if we create a branch at HEAD , this is what happens.

$ git checkout -b branch2
Switched to a new branch 'branch2'
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* 24ffff3 (HEAD, branch2, branch1) Add B
* 298c3f9 Add A
* 56c1728 (master) Initial commit

That's not what you intended to do, but you continued to work on branch2 .

$ touch C
$ git add C
$ git commit -m 'Add C'
[branch2 2cdb51b] Add C
 0 files changed
 create mode 100644 C
$ touch D
$ git add D
$ git commit -m 'Add D'
[branch2 db7fa2b] Add D
 0 files changed
 create mode 100644 D
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* db7fa2b (HEAD, branch2) Add D
* 2cdb51b Add C
* 24ffff3 (branch1) Add B
* 298c3f9 Add A
* 56c1728 (master) Initial commit

So now branch2 is ahead of master by 4 commits, but you really only want branch2 to be 2 commits ahead of master ('Add C' and 'Add D'). We can fix that with git rebase .

$ git rebase --onto master branch1 branch2
First, rewinding head to replay your work on top of it...
Applying: Add C
Applying: Add D
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* c8a299f (HEAD, branch2) Add D
* b9325dc Add C
| * 24ffff3 (branch1) Add B
| * 298c3f9 Add A
|/  
* 56c1728 (master) Initial commit

Next time you create a branch, you can use the git checkout -b <new_branch> <start_point> form.

$ git checkout -b branch3 master
Switched to a new branch 'branch3'
$ git log --decorate --graph --all --pretty=oneline --abbrev-commit
* c8a299f (branch2) Add D
* b9325dc Add C
| * 24ffff3 (branch1) Add B
| * 298c3f9 Add A
|/  
* 56c1728 (HEAD, master, branch3) Initial commit