Removing a Big File from Git

If you accidentally commit a big file:

git reset HEAD~1

And that’s it!

What if it wasn’t the last commit

Run rebase -i --root. This will open a file. Find the bad commit and replace the word pick with the word edit (or just the letter e). Save and quit the file. Fix your mistake (e.g. compress the file). Run git add <the file>. Run git rebase --continue.

What if you want to save space locally

If you want to remove the object from your local .git to save space:

git reflog expire --expire=now --all
git gc --prune=now

What if you want to save space locally but like your reflog

If you don’t want to expire your entire reflog, edit these two files:

.git/logs/refs/HEAD
.git/logs/refs/heads/main

In each of them, delete 2 lines:

the line containing the bad commit and
the line containing the reset (called reset: moving to HEAD~1) The should be the last two lines in the file. And then:

git gc --prune=now

If you’re not working on main, use the appropriate reflog instead. (You can also just grep from the hash in .git/logs.) If you re-based, delete all the children of the bad commit.

What if you’re a hoarder like me

Note that there may be other unreachable commits. If you care about this, you should list them before you edit the reflog:

git fsck --unreachable

Then, assign temporary refs to those commits and check to make sure the output is clean. After you remove the bad commit, you can also remove the temporary refs.