Question

The .git/objects in my rails project directory is still massive, after deleting hundreds of Megabytes of accidentally generated garbage.

I have tried git add -A, as well as other commands to update the index and remove nonexistent files. I gather, perhaps incorrectly, that the files with two character names in the directory are blobs. I have tried rolling back to previous commits, but no luck.

What can I do to clean this directory?

Answer 1

• If you added the files and then removed them, the blobs still exist but are dangling. git fsck will list unreachable blobs, and git prune will delete them.

• If you added the files, committed them, and then rolled back with git reset --hard HEAD^, they’re stuck a little deeper. git fsck will not list any dangling commits or blobs, because your branch’s reflog is holding onto them. Here’s one way to ensure that only objects which are in your history proper will remain:

  git reflog expire --expire=now --all
  git repack -ad  # Remove dangling objects from packfiles
  git prune       # Remove dangling loose objects
  

• Another way is also to clone the repository, as that will only carry the objects which are reachable. However, if the dangling objects got packed (and if you performed many operations, git may well have packed automatically), then a local clone will carry the entire packfile:

  git clone foo bar                 # bad
  git clone --no-hardlinks foo bar  # also bad
  

  You must specify a protocol to force git to compute a new pack:

  git clone file://foo bar  # good