git-bigfiles - Git for big files

git-bigfiles is our fork of ​Git. It has two goals:

• make life bearable for people using Git on projects hosting very large files (hundreds of megabytes)
• merge back as many changes as possible into upstream Git once they’re of acceptable quality

Features

git-bigfiles already features the following fixes:

• git-config: add core.bigFileThreshold option to treat files larger than this value as big files
• git-p4: fix poor string building performance when importing big files (git-p4 is now only marginally faster on Linux but 4 to 10 times faster on Win32)
• git-fast-import: avoid computing deltas on big files and deflate them on-the-fly (fast-import is now twice as fast and uses 3.7 times less memory with big files)

Results

Git fast-import memory usage and running time when importing a repository with binary files up to 150 MiB:

Development

git-bigfiles development is centralised in a repo.or.cz Git repository.

Clone the repository:

git clone git://repo.or.cz/git/git-bigfiles.git

If you already have a working copy of upstream git, you may save a lot of bandwidth by doing:

git clone git://repo.or.cz/git/git-bigfiles.git --reference /path/to/git/

The main Git repository is constantly merged into git-bigfiles. See the ​git-bigfiles repository on ​repo.or.cz.

To do

The following commands are real memory hogs with big files:

• git diff-tree --stat --summary -M <sha1> HEAD
• git gc (as called by git merge) fails with warning: suboptimal pack - out of memory

Readings

• 2006/02/08: Linus on ​Handling large files with GIT
• 2006/07/10: ​Revisiting large binary files issue
• 2008/09/09: ​malloc fails when dealing with huge files