    git checkout featurebranch &amp;&amp; git rebase master

As long as you haven&#39;t pushed yet, it is better to replay your changes on top of master.  

See:

- [git rebase vs. merge][1]
- [git workflow and rebase vs merge questions][2]


  [1]: https://stackoverflow.com/questions/804115/git-rebase-vs-git-merge/804178#804178
  [2]: https://stackoverflow.com/questions/457927/git-workflow-and-rebase-vs-merge-questions/457988#457988

``git checkout featurebranch &amp;&amp; git merge master``

You can do this as many times as you like; it won&#39;t affect master and you will be able to easily do it the other way around whenever you find out you are done with the feature branch.

A riff on [git: show all changed files between two commits][1]: I want a listing of all files that have been changed between two commits, even if they are now the same (ie, changed and then changed back).

  [1]: https://stackoverflow.com/questions/1552340/git-show-all-changed-files-between-two-commits &quot;git: show all changed files between two commits&quot;

git: show all files changed between two commits

I was looking at this [blog post][1] and had following questions:

 - Why do we need the `new` keyword, is it just to specify that a base class method is being hidden. I mean, why do we need it? If we don&#39;t use the `override` keyword, aren&#39;t we hiding the base class method?
 - Why is the default in C# to hide and not override? Why have the designers implemented it this way?

  [1]: http://geekswithblogs.net/BlackRabbitCoder/archive/2010/06/24/c-fundamentals-beware-of-implicit-hiding.aspx

Why do we need the new keyword and why is the default behavior to hide and not override?

I have a project which uses git and I&#39;d like to start a new branch to add a major new feature.

Under the `main` branch, I&#39;ll continue to add bug fixes and minor features. At regular intervals I&#39;d like to pull the changes from the main branch into my &quot;major new feature&quot; branch. What&#39;s the best way to do this? 

Eventually, I&#39;ll merge the &quot;major new feature&quot; branch into the main branch.   

Git: getting changes from another branch

I have a project which uses git and I'd like to start a new branch to add a major new feature.
Under the <code>main</code> branch, I'll continue to add bug fixes and minor features. At regular intervals I'd like to pull the changes from the main branch into my "major new feature" branch. What's the best way to do this?
Eventually, I'll merge the "major new feature" branch into the main branch.

I got this message from Git:

&gt; You asked to pull from the remote &#39;origin&#39;, but did not specify
a branch. Because this is not the default configured remote
for your current branch, you must specify a branch on the command line.

Can anyone explain it? and more important how to fix it?

Confusing error message from git

We have a Git repository with over 400 commits, the first couple dozen of which were a lot of trial-and-error. We want to clean up these commits by squashing many down into a single commit. Naturally, git-rebase seems the way to go. My problem is that it ends up with merge conflicts, and these conflicts are not easy to resolve. I don&#39;t understand why there should be any conflicts at all, since I&#39;m just squashing commits (not deleting or rearranging). Very likely, this demonstrates that I&#39;m not completely understanding how git-rebase does its squashes.

Here&#39;s a modified version of the scripts I&#39;m using:

-----
repo_squash.sh (this is the script that is actually run):

-----

    rm -rf repo_squash
    git clone repo repo_squash
    cd repo_squash/
    GIT_EDITOR=../repo_squash_helper.sh git rebase --strategy theirs -i bd6a09a484b8230d0810e6689cf08a24f26f287a

-----
repo_squash_helper.sh (this script is used only by repo_squash.sh):

-----

    if grep -q &quot;pick &quot; $1
    then
    #  cp $1 ../repo_squash_history.txt
    #  emacs -nw $1
      sed -f ../repo_squash_list.txt &lt; $1 &gt; $1.tmp
      mv $1.tmp $1
    else
      if grep -q &quot;initial import&quot; $1
      then
        cp ../repo_squash_new_message1.txt $1
      elif grep -q &quot;fixing bad import&quot; $1
      then
        cp ../repo_squash_new_message2.txt $1
      else
        emacs -nw $1
      fi
    fi

-----
repo_squash_list.txt: (this file is used only by repo_squash_helper.sh)

-----

    # Initial import
    s/pick $251a190$/squash \1/g
    # Leaving &quot;Needed subdir&quot; for now
    # Fixing bad import
    s/pick $46c41d1$/squash \1/g
    s/pick $5d7agf2$/squash \1/g
    s/pick $3da63ed$/squash \1/g

-----

I&#39;ll leave the &quot;new message&quot; contents to your imagination. Initially, I did this without the &quot;--strategy theirs&quot; option (i.e., using the default strategy, which if I understand the documentation correctly is recursive, but I&#39;m not sure which recursive strategy is used), and it also didn&#39;t work. Also, I should point out that, using the commented out code in repo_squash_helper.sh, I saved off the original file that the sed script works on and ran the sed script against it to make sure it was doing what I wanted it to do (it was). Again, I don&#39;t even know why there *would* be a conflict, so it wouldn&#39;t seem to matter so much which strategy is used. Any advice or insight would be helpful, but mostly I just want to get this squashing working.

Updated with extra information from discussion with Jefromi:
---

Before working on our massive &quot;real&quot; repository, I used similar scripts on a test repository. It was a very simple repository and the test worked cleanly.

The message I get when it fails is:

    Finished one cherry-pick.
    # Not currently on any branch.
    nothing to commit (working directory clean)
    Could not apply 66c45e2... Needed subdir

This is the first pick after the first squash commit. Running `git status` yields a clean working directory. If I then do a `git rebase --continue`, I get a very similar message after a few more commits. If I then do it again, I get another very similar message after a couple dozen commits. If I do it yet again, this time it goes through about a hundred commits, and yields this message:

    Automatic cherry-pick failed.  After resolving the conflicts,
    mark the corrected paths with &#39;git add &lt;paths&gt;&#39;, and
    run &#39;git rebase --continue&#39;
    Could not apply f1de3bc... Incremental

If I then run `git status`, I get:

    # Not currently on any branch.
    # Changes to be committed:
    #   (use &quot;git reset HEAD &lt;file&gt;...&quot; to unstage)
    #
    # modified:   repo/file_A.cpp
    # modified:   repo/file_B.cpp
    #
    # Unmerged paths:
    #   (use &quot;git reset HEAD &lt;file&gt;...&quot; to unstage)
    #   (use &quot;git add/rm &lt;file&gt;...&quot; as appropriate to mark resolution)
    #
    # both modified:      repo/file_X.cpp
    #
    # Changed but not updated:
    #   (use &quot;git add/rm &lt;file&gt;...&quot; to update what will be committed)
    #   (use &quot;git checkout -- &lt;file&gt;...&quot; to discard changes in working directory)
    #
    # deleted:    repo/file_Z.imp

The &quot;both modified&quot; bit sounds weird to me, since this was just the result of a pick. It&#39;s also worth noting that if I look at the &quot;conflict&quot;, it boils down to a single line with one version beginning it with a [tab] character, and the other one with four spaces. This sounded like it might be an issue with how I&#39;ve set up my config file, but there&#39;s nothing of the sort in it. (I did note that core.ignorecase is set to true, but evidently git-clone did that automatically. I&#39;m not completely surprised by that considering that the original source was on a Windows machine.)

If I manually fix file_X.cpp, it then fails shortly afterward with another conflict, this time between a file (CMakeLists.txt) that one version thinks should exist and one version thinks shouldn&#39;t. If I fix this conflict by saying I do want this file (which I do), a few commits later I get another conflict (in this same file) where now there&#39;s some rather non-trivial changes. It&#39;s still only about 25% of the way through the conflicts.

I should also point out, since this might be very important, that this project started out in an svn repository. That initial history very likely was imported from that svn repository.

Update #2:
----

On a lark (influenced by Jefromi&#39;s comments), I decided to do the change my repo_squash.sh to be:

    rm -rf repo_squash
    git clone repo repo_squash
    cd repo_squash/
    git rebase --strategy theirs -i bd6a09a484b8230d0810e6689cf08a24f26f287a

And then, I just accepted the original entries, as is. I.e., the &quot;rebase&quot; shouldn&#39;t have changed a thing. It ended up with the same results describe previously.

Update #3:
---

Alternatively, if I omit the strategy and replace the last command with:

    git rebase -i bd6a09a484b8230d0810e6689cf08a24f26f287a

I no longer get the &quot;nothing to commit&quot; rebase problems, but I&#39;m still left with the other conflicts.

Update with toy repository that recreates problem:
---

test_squash.sh (this is the file you actually run):

    #========================================================
    # Initialize directories
    #========================================================
    rm -rf test_squash/ test_squash_clone/
    mkdir -p test_squash
    mkdir -p test_squash_clone
    #========================================================
    
    #========================================================
    # Create repository with history
    #========================================================
    cd test_squash/
    git init
    echo &quot;README&quot;&gt;README
    git add README
    git commit -m&quot;Initial commit: can&#39;t easily access for rebasing&quot;
    echo &quot;Line 1&quot;&gt;test_file.txt
    git add test_file.txt
    git commit -m&quot;Created single line file&quot;
    echo &quot;Line 2&quot;&gt;&gt;test_file.txt 
    git add test_file.txt 
    git commit -m&quot;Meant for it to be two lines&quot;
    git checkout -b dev
    echo Meaningful code&gt;new_file.txt
    git add new_file.txt 
    git commit -m&quot;Meaningful commit&quot;
    git checkout master
    echo Conflicting meaningful code&gt;new_file.txt
    git add new_file.txt 
    git commit -m&quot;Conflicting meaningful commit&quot;
    # This will conflict
    git merge dev
    # Fixes conflict
    echo Merged meaningful code&gt;new_file.txt
    git add new_file.txt
    git commit -m&quot;Merged dev with master&quot;
    cd ..
    
    #========================================================
    # Save off a clone of the repository prior to squashing
    #========================================================
    git clone test_squash test_squash_clone
    #========================================================
    
    #========================================================
    # Do the squash
    #========================================================
    cd test_squash
    GIT_EDITOR=../test_squash_helper.sh git rebase -i HEAD@{7}
    #========================================================
    
    #========================================================
    # Show the results
    #========================================================
    git log
    git gc
    git reflog
    #========================================================

test_squash_helper.sh (used by test_sqash.sh):

    # If the file has the phrase &quot;pick &quot; in it, assume it&#39;s the log file
    if grep -q &quot;pick &quot; $1
    then
      sed -e &quot;s/pick $.*$ $Meant for it to be two lines$/squash \1 \2/g&quot; &lt; $1 &gt; $1.tmp
      mv $1.tmp $1
    # Else, assume it&#39;s the commit message file
    else
    # Use our pre-canned message
      echo &quot;Created two line file&quot; &gt; $1
    fi


P.S.: Yes, I know some of you cringe when you see me using emacs as a fall-back editor.

P.P.S.: We do know we&#39;ll have to blow away all of our clones of the existing repository after the rebase. (Along the lines of &quot;thou shalt not rebase a repository after it&#39;s been published&quot;.)

P.P.P.S: Can anyone tell me how to add a bounty to this? I&#39;m not seeing the option anywhere on this screen whether I&#39;m in edit mode or view mode.




Why does git-rebase give me merge conflicts when all I&#39;m doing is squashing commits?

Let&#39;s say I have a repo that includes this directory structure:

    repo/
      blog/
        _posts/
          some-post.html
      another-file.txt

I want to move `_posts` to the top level of the repo, so the structure will look like this:

    repo/
      _posts/
        some-post.html
      another-file.txt

This is simple enough with `git mv`, but I want to make the history look as though `_posts` *always* existed at the root of the repo, and I want to be able to get the entire history of `some-post.html` via `git log -- _posts/some-post.html`. I imagine I can use some magic with `git filter-branch` to accomplish this, but I haven&#39;t figured out exactly how to do that. Any ideas?

How can I move a directory in a Git repo for all commits?

Every time I do a merge I need for a merge commit to be generated and I would like it to have more than just the summary of all the commits.

**My question is how can I format git-fmt-merge-msg or what determines this automated message** (I can do this manually after a commit by amending it and using git-log --pretty=format:&#39;...&#39;)

For example I would like to format it as such:

     Merge branch &#39;test&#39;  
        * test:  
          [BZ: #123] fifth commit subject  
          [BZ: #123] fourth commit subject  
          [BZ: #123] third commit subject  
          [BZ: #123] second commit subject  
          [BZ: #123] first commit subject  
    
     __________________________________________
     Merge details:  
         [BZ: #123] fifth commit subject  
                   at 2010-06-30 11:29:00 +0100  
           - fifth commit body  
     
         [BZ: #123] fourth commit subject  
                   at 2010-06-30 11:22:17 +0100  
           - fourth commit body  
     
         [BZ: #123] third commit subject  
                   at 2010-06-30 11:21:43 +0100  
           - third commit body  
     
         [BZ: #123] second commit subject  
                   at 2010-06-30 11:21:30 +0100  
           - second commit body  
     
         [BZ: #123] first commit subject  
                   at 2010-06-30 11:29:57 +0100  
           - first commit body

  [1]: https://stackoverflow.com/questions/3148863/how-can-i-customize-gits-merge-commit-message/3264341#3264341

Content Type	Original Author	Original Content on Stackoverflow
Question	FunLovinCoder	View Question on Stackoverflow
Solution 1 - Git	VonC	View Answer on Stackoverflow
Solution 2 - Git	David Winslow	View Answer on Stackoverflow

Git: getting changes from another branch

Git Problem Overview

Git Solutions

Solution 1 - Git

Solution 2 - Git

Why do we need the new keyword and why is the default behavior to hide and not override?

git: show all files changed between two commits

Attributions