git-repair can repair various forms of damage to git repositories.
It is a complement to git fsck, which finds problems, but does not fix
them.
As well as avoiding the need to rm -rf a damaged repository and re-clone, using git-repair can help rescue commits you've made to the damaged repository and not yet pushed out.
download
git clone git://git-repair.branchable.com/ git-repair- from Hackage
install
This is a Haskell program, developed as a spinoff of git-annex.
To build it, you will need to install the Haskell Platform.
Then to install it:
cabal update; cabal install git-repair --bindir=$HOME/bin
how it works
git-repair starts by deleting all corrupt objects, and
retrieving all missing objects that it can from the remotes of the
repository.
If that is not sufficient to fully recover the repository, it can also
reset branches back to commits before the corruption happened, delete
branches that are no longer available due to the lost data, and remove any
missing files from the index. It will only do this if run with the
--force option, since that rewrites history and throws out missing data.
After running this command, you will probably want to run git fsck to
verify it fixed the repository.
Note that fsck may still complain about objects referenced by the reflog, or the stash, if they were unable to be recovered. This command does not try to clean up either the reflog or the stash.
Also note that the --force option never touches tags, even if they are no
longer usable due to missing data, so fsck may also find problems with
tags.
Since this command unpacks all packs in the repository, you may want to
run git gc afterwards.