Investigations - Git

Overview

Notes

Tips

• How to change the commit a branch is pointing to (adviced by ChatGPT 🥳

  • Option 1: Using `git branch -f`
    
    First, ensure you have checked out the branch you want to modify. Then, use the `git branch -f` command to forcibly move the branch pointer to a new commit.
    
    ```bash
    git checkout <branch_name>
    git branch -f <branch_name> <new_commit>
    ```
    
    Replace `<branch_name>` with your desired branch and `<new_commit>` with the target commit's hash or any other way of specifying the commit (like `HEAD~3`).
    
    Option 2: Using `git reset`
    
    First, check out the branch you want to update, and then reset the branch to the target commit:
    
    ```bash
    git checkout <branch_name>
    git reset --hard <new_commit>
    ```
    
    Replace `<branch_name>` with your desired branch and `<new_commit>` with the target commit's hash or any other way of specifying the commit (like `HEAD~3`).
    
    Note: When using `git reset --hard`, be mindful that it **may discard uncommitted changes** in your working directory.
    
    After updating the branch with one of the previous methods, if you want to push the updated branch to a remote repository, you'll need to force-push using the `--force` option:
    
    ```bash
    git push --force origin <branch_name>
    ```
    
    Keep in mind that force-pushing can cause issues for other collaborators. Make sure you communicate the changes to your entire team to avoid conflicts.

fatal: git upload-pack: aborting due to possible repository corruption on the remote side.

When I fetch remote my private git repository via pure SSH, I saw this error and researched some articles, including:

• https://forum.gitlab.com/t/git-clone-reports-git-upload-pack-aborting-due-to-possible-repository-corruption-on-the-remote-side/32974

• https://forum.gitea.com/t/remote-aborting-due-to-possible-repository-corruption/6617/2

It seems like it is related to a memory issue, and I investigated the remote site to find the oom-killer killing git process. As the remote instance was a small cloud instance with limited memory, I added a swap file.

$ sudo dmesg
...
[165437.371179] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=user.slice,mems_allowed=0,global_oom,task_memcg=/user.slice/user-1001.slice/session-114.scope,task=git,pid=20097,uid=1001
[165437.371214] Out of memory: Killed process 20097 (git) total-vm:1118668kB, anon-rss:83372kB, file-rss:128kB, shmem-rss:0kB, UID:1001 pgtables:872kB oom_score_adj:0

$ sudo swapon /swapfile

References

---

• ref aliases

  • https://git-scm.com/book/ja/v2/Gitの内側-Gitの参照

  • https://stackoverflow.com/questions/15284109/command-to-determine-the-upstream-ref-of-the-current-head

    • HEAD

    • @{u}

• grafted

  • https://tukaelu.hatenablog.jp/entry/2018/04/10/git-grafted-commit

• Shallow clone

  • $ git fetch --unshallow
    remote: Total 0 (delta 0), reused 0 (delta 0)
    $ git rev-parse --is-shallow-repository
    true

  • Hmm. unshallow does not always make it “full”.

  • rm .git/shallow

• clean

  • https://stackoverflow.com/questions/59645813/how-to-reset-a-git-repository-including-ignored-files

  • https://git-scm.com/docs/git-clean

    • git clean -fd to remove untracked files, while keeping ignored ones.

    • git clean -fdx to remove ...

    • git clean -fdX to remove ...

• History squashing

  • https://stackoverflow.com/questions/1657017/how-to-squash-all-git-commits-into-one