Activity 3: Graft a repository

[selkins13]

Duration: 20 minutes

Grafting commands

• Step 1: Clean your repository
  • LFS
  • Unneeded or unwanted files
  • Add dependencies to dependency management
  • Remove independent components
  • Remove any binaries
  • Review automation in the repository
• Step 2: Create a new repository
• Step 3: Prepare for grafting
  • Communicate properly to the affected teams
  • Merge all in progress work
  • If something is not merged, it will not be moved
  • Call GitHub Professional Services if things go south
• Step 4: Delete your history

# Delete the git folder that contains git objects
rm -rf .git

# Initialize a new history
git init

# Set the new repository
git remote add origin git@github.com:githubuniverseworkshops/grafting-repo.git

• Step 5: Write a commit referencing to the previous repository as your first commit in the new repository

# Add all files to the stage
git add --all

# Add changes to history
git commit -m "Previous repo can be found on https://github.com/torvalds/linux"

# Submit your changes to upstream
git push --set-upstream origin main

Working with the new repository

To preserve the history while working with the new repository, follow the grafting command:

# Fetch the old history
git fetch git@github.com:torvalds/linux.git

# See you only have one commit in it
git log --oneline

# See the commits we are replacing
git rev-parse --short HEAD
git rev-parse --short FETCH_HEAD

# Perform the grafting operation replacing HEAD with FETCH_HEAD
git replace HEAD FETCH_HEAD

Check that all the changes that have happened to the repository are local and don't get pushed when new code goes to the repository:

# Check that the new commit goes to the right repo
# Modify a file
echo "Test" > test.txt
git add --all
git commit -m "Adding a test commit"

# Check that you can navigate the history
git log --oneline | head -n 10

# Push the change and see the number of commits is still 2
git push

Analysis after grafting

Re-run the command from Activity 2 to analyze the grafted repository.

Stats of repo size: git-sizer

1. Download the corresponding compiled version of git-sizer.

Optionally you can install git-sizer using Homebrew if you are on Mac.

2. Run the tool from the root of the repository to analyze:

/path/to/git-sizer --verbose

Find files that should be in LFS: git-find-lfs-extensions

1. Checkout the grafting-monorepos repository
2. Run the tool from the root of the repository to analyze:

/path/to/grafting-monorepos/scripts/git-find-lfs-extensions

Print directories with the number of files contained: git-find-dirs-many-files

1. Checkout the grafting-monorepos repository
2. Run the tool from the root of the repository to analyze:

/path/to/grafting-monorepos/scripts/git-find-dirs-many-files

Find dirs that should not be committed: git-find-dirs-unwanted

1. Checkout the grafting-monorepos repository
2. Run the tool from the root of the repository to analyze:

/path/to/grafting-monorepos/scripts/git-find-dirs-unwanted | head -n 15            

Analyze the repository: git-filter-repo --analyze

1. Clone the git-filter-repo tool
2. Execute the tool from the linux repository

/path/to/git-filter-repo/git-filter-repo --analyze

For examples and more information, please see README.md -> Activity 3.

[TestStefan]

It sounds like that there is the assumption that everything is merged into the main branch but what about existing bug-fix/release branches which still need to be maintained? 🚀

[larsxschneider]

@TestStefan Very good question! This approach only covers merged changes. Unmerged branches need a special treatment that is outside the scope of this workshop. But we can help with that in a Professional Services session.

[TestStefan]

@TestStefan Very good question! This approach only covers merged changes. Unmerged branches need a special treatment that is outside the scope of this workshop. But we can help with that in a Professional Services session.

I guess you need to cleanup all active branches and create in the new grafted repository these branches as well and link them back to the archive repository.

[neilwang0913]

After run the command "git push --set-upstream origin main" I got the fellowing error(I am using windows 10)
git@github.com: Permission denied (publickey).
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.

Any suggestion to fix it? Thanks

[neilwang0913]

Any clear instruction for step 1: Clean your repository: how to make it ?
LFS
Unneeded or unwanted files
Add dependencies to dependency management
Remove independent components
Remove any binaries
Review automation in the repository

If you could provide a video demo using a public example?

Many thanks.