AWS Global Infrastructure

DevOps

Topics Covered
  • DevOps (92 Blogs)
  • Docker (9 Blogs)
  • DevOps Engineer Masters Program (18 Blogs)
  • Kubernetes (11 Blogs)
SEE MORE

What are the common Git mistakes and how to fix them?

Last updated on Sep 14,2024 13.1K Views

Divya Bhushan
A passionate Content Developer with key skills – Git, Dockers, Databases, Linux... A passionate Content Developer with key skills – Git, Dockers, Databases, Linux and Unix.
6 / 10 Blog from Git & GitHub

With the boom of DevOps technology, it becomes inevitable for any IT person to work on multiple data simultaneously and your data is ever-evolving with time. It is also essential to track every change in the data and be prepared to undo or revert any undesired change when required.

I must confess, versioning my data in Git allow me to be more experimental in my project development. If I mess up, I know git always has a way to undo and/or revert that version of my project to the way it was before I screwed up. Each Git workflow layer is designed to allow the data changes to be reviewed and modified and/or corrected before moving the data in the next stage. So the following are the mistakes that are covered in this blog:

    Un-stage files/directories from Index

    While adding and/or modifying files you often tend to use the default behavior of ‘git add’ command, which is to add all files and directories to the Index. Many a time you feel the need to un-stage certain files or modify them one last time before committing them.

    Syntax: git reset <filename/dirname>


    remove files from Index - common git mistakes -Edureka

    Un-staging files from the Index area gives you another chance to re-work on your data before committing to a local repo.

    Edit the last committed message

    Command: git commit --amend
    You can edit the latest commit message without creating a new one. To list the commit logs, I have set an alias ‘hist’:
    Command: git config --global alias.hist 'log --pretty=format:"%C(yellow)%h%Creset %ad | %C(green)%s%Creset%C(red)%d%Creset %C(blue)[%an]" --graph --decorate --date=short'x

    edit the last commit object- common git mistakes -Edureka

    Do not amend the commit message which is already pushed on to a remote repository and shared with others, as that would make the earlier commit history invalid and thus any work based on that may be affected.

    Forgot some changes in the last commit

    Let’s say you forgot to make some modifications and already committed your snapshot, also you do not want to make another commit to highlight your mistake.
    Command: git commit --amend


    Add missing data to the last commit- common git mistakes -Edureka

    I have highlighted how the recent commit object’s sha-1 id has been recreated and changed. I pretended to have made a single commit blending both the changes into one.

    Discard local changes

    So, here is a case where I modified the ‘README’ file and staged it. Next, I modified the same file a second time but realized that I didn’t want the second change.
    same file both staged and modified(un-staged)- common git mistakes -Edureka
    Now, let me not undo the entire change manually, I can simply pull the staged version of the file.
    Syntax:
    git checkout -- <filename> –local changes in a file
    git checkout -- <dirname> –local changes in all the files in the directory­­

    Command: git checkout -- README
    local change discarded, staged flie applied- common git mistakes -Edureka

    So, I discarded my last changes to the file and accepted the staged version of the file. In the next commit, just the staged version of the file goes in the local repository.

    Committed personal data to the local repository

    I want to remove certain data from the local repository but keep the files in the working directory.
    Syntax:
    git reset --mixed HEAD~
    git reset --mixed <commit-id>

    Command: git reset --mixed HEAD~1
    HEAD~1 indicates a commit just before the recent commit pointed by the current branch HEAD.

    remove latest commit to undo the latest snapshot- common git mistakes -Edureka

    Files in the current snapshot removed from both the local repository and the staging area. Add the following patterns in the global .gitignore file to exclude them from being tracked by git.
    vim ~/.gitignore_global
    # password files #
    *.pass
    *.key
    *.passwd

    With this, the commit that had the snapshot of password files is removed, and you get a clean staging area. My files are still present on my working directory but no longer present in the local repository, also will not be pushed on a remote repository.

    Caution: If you lose them git cannot recover them for you as it does not know about it.

    Replace the latest commit with a new commit

    Syntax: git reset --soft [ <commit-id>/HEAD~n>]

    The ‘–soft’ option just remove the committed files from the local repository while they are still staged in the Index and you can re-commit them after a review. <commit-id> is the sha-1 of the snapshot that you want to remove from the local repo. <HEAD~n> where n is the number of commits before the HEAD commit

    Command: git reset --soft HEAD~1


    reset --soft latest commit only from local repo- common git mistakes -Edureka

    Modify files and stage them again
    reviewed and staged- common git mistakes -Edureka

    Command: git commit -m 'Adding index.html and style.css'
    Your commit history now turns out to be:

    latest commit replaced- common git mistakes -Edureka

     

    Committed the wrong data

    Syntax:
    git reset --hard HEAD~n –reset the project to ‘n’ commits before the latest committed snapshot
    git reset --hard <commit-id> –reset the project to given commit id snapshot

    latest commit that has wrong data- common git mistakes -Edureka

    Command: git reset --hard HEAD~1

    removed the wrong commit and data- common git mistakes -Edureka
    The latest commit and the corrupt files are removed from the local repository, staging area as well as the working directory.

    Caution: It’s a dangerous command as you end up losing files in the working directory. Not recommended on a remotely shared repository.

    Get back to my old project state

    You can ump to an older state of your project in the history of time. If you mess up in the latest version or need enhancements in older code, you may want to create another branch out of that old project snapshot to not hinder with your current work. Let’s see how:
    a. List out the project history and decide on the older commit id, command: git hist
    b. Create another branch out of the commit id: git checkout -b old-state e7aa9a5
    c. Continue working on the code and later merge/rebase with the ‘master’ branch.

    Get older version of your project state- common git mistakes -Edureka

    Recover a deleted local branch

    It is possible to regenerate the lost work on a reference branch. Say, I deleted the branch ‘old_code’ without merging with the main branch and lost the work. And no, I did not push the branch to a remote repository either, what then? Well git tracks and keep a journal entry of all the changes done on each reference, let’s see mine: git reflog

    reflog-recover deleted branch and its work- common git mistakes -Edureka

    So, HEAD@{2} is the pointer when I moved to ‘old_code’ branch, let’s recover that:

    Syntax: git checkout -b <branch-name> <commit-id>
    Command: git checkout -b old_code HEAD@{2}

    You must be now in the ‘old_code’ branch with your latest work at the time of its creation.Additionally, the ‘reflog’ pointer at HEAD@{1} was the recent commit made on the ‘old_code’ branch.To restore this unique commit just run the command as: git reset --hard HEAD@{1}.This also restores the modified files in the working directory.

    restore deleted work on the branch from reflog- common git mistakes -Edureka

    If you would want to know in detail how this command works and how can you manage the ‘reflog’ entries, you may as well read my earlier post on recovering the deleted branch from git reflog.

    Undo changes done in a commit

    git revert is used to record some new commits to reverse the effect of some earlier commits.
    Syntax: git revert <commit-id>
    From my commits logs, I would like to reverse the change done in the highlighted commit id:reverse the effect of an older commit- common git mistakes -Edureka

     

    Command: git revert 827bc0d

    new commit with reverse of an older one- common git mistakes -Edureka

    It is better, that you do not reset ‘–hard’ the shared commits, but instead ‘git revert’ them to preserve the history so that it becomes easier for all to track down the history logs to find out what was reverted, by whom and why?

    You can use the same logic of referring the commits concerning the HEAD pointer instead of giving the commit id, as in HEAD~3 or HEAD~4 and so on.

    Gave a wrong name to my branch

    You can rename a local branch name. It so happens many times that you may wish to rename your branch based on the issue you are working on without going through the pain of migrating all your work from one location to another. For instance, you could either be on the same branch or a different branch and still be able to rename the desired branch as shown below:
    Syntax: git branch -m <old_name> <new_name>
    Command: git branch -m old_code old_#4920

    renaming a branch- common git mistakes -Edureka

    As you may wonder does git keeps of a track of this rename? Yes, it does refer to your ‘reflog’ entries, here is mine:

    reflog entry when the branch was renamed- common git mistakes -Edureka

    Renaming a branch will not affect its remote-tracking branch. We shall see in the remote section how to replace a branch on the remote repository

    Re-arrange history logs before pushing to remote

    How I wish I would have made certain commits earlier than others and would have not made some commits at all. Interactively re-arrange and edit the old commits to effectively fix-up or enhance the code
    Syntax: git rebase -i <after-this-commit_id>
    Command: git rebase -i fb0a90e –start rebasing the commits that were made after the commit-id fb0a90e

    re-arrange and edit commit logs- common git mistakes -Edureka

    re-arrange and edit commit logs- common git mistakes -Edureka

    Re-visit the git rebase documentation to understand how is an ‘–interactive or -i’ rebase different from a regular rebase.

    Committed unrelated changes into a single commit

    In this case, you need split an old buried commit into multiple logical commits.
    Syntax: git rebase -i <after-this-commit_id>
    Command: git rebase -i fb0a90e
    In the rebase editor, you must choose e7aa9a5 commit id and change it to ‘edit’ instead of ‘pick’.

    unrelated changes - common git mistakes -Edureka

    You would now be in the project’s version of commit id-e7aa9a5. First, reset the commit history and staging area to the previous commit-Command: git reset HEAD~1
    Second, edit + stage + commit the files individually
    Commands:
    git add code && git commit -m 'Adding initial codes'
    git add newcode && git commit -m 'Adding new code'

    Third, continue the rebase and end.

    Command: git rebase --continue
    Fourth, view the history with additional commits.

    Command: git hist

    splitting commit into multiple using rebase - common git mistakes - Edureka

    Change author-email in all commits on all branches

    I have been versioning and committing my project files in git since a long time now, but until now it never struck me that my email id was compromised in my commit history logs which are even published on remote repositories. Well, this may happen to anyone when you initially set up the configurations in the “.gitconfig” file.To my relief git can re-write the environment variables we provide when creating a commit object.

    First I get the list of email ids to decide the ones I want to change:
    Command: git log --all --pretty=format:"%an <%ae> %d" –This prints author-name<email-id> (refname/branch-name)

    Second, I run through every commit on each branch and re-write the commit object with the new email id
    Command:
    git filter-branch --env-filter '
    if [ "$GIT_AUTHOR_NAME" = "divya" ]
    then
    GIT_AUTHOR_EMAIL = "divya@github.com"
    fi
    ' -- --all

    email id renamed in all branches - common git mistakes -Edureka

    Lost and found files

    Suppose you have lost a certain file and you do not remember its name, but could recall certain words in the file. In this case, you can follow these steps-
    Step 1: List all the commits that ever contained the file snapshot with the searched pattern
    Command: git rev-list --all | xargs git grep -i 'timestamp'


    list of lost files that contain the pattern - common git mistakes -Edureka
    Step 2: Create a new branch ‘lost-found’ from this highlighted commit-id
    Syntax: git checkout -b lost-found d8c6a76a6dcb1fc6e8c3f6b097e1bd07e7cd328f

    missing file now on the new branch - common git mistakes -Edureka

    Forgot which branch has my commit-id

    At times, after you detect a buggy commit id you might as well want to know all the branches that have this commit on them so you could fix them all. Checking out each branch’s history is not very practical in a large multi-branch project.

    A bad commit made in my navigation building application once broke the code, that’s when I used the ‘git bisect’ command to detect the commit id that was bad followed by the command: git branch --contains <commit-id> to list the branches with that bad commit.

    Branches with bad commit id - common git mistakes -Edureka

    So, now I know all the branches that still have the bad commit, I could either revert or reset this changeset.

    Delete a commit from history

    Sometimes I feel the need to just wipe out a commit from the history and leave no trail of it. I would not recommend you to try this stunt on a shared branch but only on your local branch.
    Syntax: git rebase -i <commit-id>
    Command: git rebase -i 93859d8
    In the rebase editor-> replace ‘edit’ with ‘drop’ for the highlighted commit id: 69f4813

    Remove a commit from history - common git mistakes -Edureka

    In some of the cases, this re-writing may result in conflicts. You must resolve conflicts then proceed further.

    Warning: This is a dangerous command as this re-writes history and may lose data.Such a branch differs from its remote-counterpart and will have to be pushed with the --force or --force-with-lease option.

    Pushed a wrong branch to the remote

    Now, here is what I want to do- I want to delete a remote branch and also stop tracking it from my local branch.’git push‘ command when used with the --delete option deletes the remote branch So, this is how I get the local copy of the cloned project –

    git clone https://github.com/greets/myProj.git
    cd myProj


    Delete remote repo and un-track it - common git mistakes -Edureka

    Once, the remote branch is deleted others on the shared repo must refresh and update their remote references with the --prune option to delete the missing object references: git fetch --prune -v origin

    In this post, I have mentioned some of the common mistakes or changes that git can help you fix. Every code is unique and developed in its way, so there are also different ways of approaching and fixing a problem. You could always refer to the official git documentation to understand how various git commands safeguard your source code and how to utilize the commands the best way possible.

    Now that you have understood the common Git mistakes, check out this DevOps Certification Training by Edureka, a trusted online learning company with a network of more than 250,000 satisfied learners spread across the globe. The Edureka DevOps Certification Training course helps learners to understand what is DevOps and gain expertise in various DevOps processes and tools such as Puppet, Jenkins, Nagios, Ansible, Chef, Saltstack and GIT for automating multiple steps in SDLC.

    Got a question for us? Please mention it in the comments section of this “common Git mistakes” and we will get back to you

     

    Upcoming Batches For DevOps Certification Training Course
    Course NameDateDetails
    DevOps Certification Training Course

    Class Starts on 23rd November,2024

    23rd November

    SAT&SUN (Weekend Batch)
    View Details
    DevOps Certification Training Course

    Class Starts on 25th November,2024

    25th November

    MON-FRI (Weekday Batch)
    View Details
    DevOps Certification Training Course

    Class Starts on 21st December,2024

    21st December

    SAT&SUN (Weekend Batch)
    View Details
    Comments
    0 Comments

    Join the discussion

    Browse Categories

    webinar REGISTER FOR FREE WEBINAR
    REGISTER NOW
    webinar_success Thank you for registering Join Edureka Meetup community for 100+ Free Webinars each month JOIN MEETUP GROUP

    Subscribe to our Newsletter, and get personalized recommendations.

    image not found!
    image not found!

    What are the common Git mistakes and how to fix them?

    edureka.co