Git Details

Some Interesting Questions

1. What happens during init/clone/add/commit/push?
2. Does amend + push -f really delete sensitive history?

Core Concepts

What is Git: Git is a content-addressable file system, and on top of that, it provides a user interface for a version control system.

Git directly records snapshots rather than performing diff comparisons.

If a file is not modified, Git does not re-store the file, but it re-stores the pointer to that file.

If a file is modified, Git re-stores the new file, but it compresses the overall size by packing (the latest version saves the complete content, historical versions only save differences).

Directory Structure

.git
├── COMMIT_EDITMSG
├── HEAD # Current pointer
├── config
├── description
├── hooks
├── index # Staging area information
├── info
├── logs
├── objects # All data
├── packed-refs
└── refs # Stores reference files

Data Types

top-down: ref, commit, tree, blob

Low-level Commands

Storing Files

git hash-object -w filename

Outputs a 40-character hexadecimal checksum: the first 2 characters as the directory name, the last 38 as the file name.

Implementation details: The data to be stored (content) plus header information are SHA-1 digested to get this checksum, then the file is compressed with zlib and stored.

Viewing Files

git cat-file -p checksum

Outputs file content.

git cat-file -t checksum

Outputs the type of data: blob, tree, commit.

Staging Area

git update-index --add --cacheinfo 100644 checksum filename

Adds a file to the staging area.

git write-tree

Writes the content of the staging area to a tree object.

Committing

git commit-tree tree_checksum

Gets a commit.

References

git update-ref reference_filename checksum

Gets a reference (branch/tag).

Answering Questions

Q: What happens during init/clone/add/commit/push?

A: During init, Git creates the .git directory, which contains the HEAD file, objects directory, and refs directory, etc.

During clone (smart protocol), Git calls the server's upload-pack process, then uses the fetch-pack program to retrieve data.

During add, the file is added to the staging area (index file).

During commit, the content of the staging area is written to a tree object and the tree object is committed, while also updating refs and HEAD.

During push (smart protocol), Git calls the server's receive-pack process, and uses the send-pack program to compare and transfer data not present on the server.

Q: Does amend + push -f really delete sensitive history?

A: After testing, the local objects directory still contains data after amend, but if you re-pull remotely after push -f, these objects are gone. So, it can be considered that sensitive records are deleted.