Git log: fatal object [sha1] is corrupted

Question

Is there any way I can repair my repository with commit history in tact.

 # git log
fatal: object 01aeb2bf2e93b238f0e0422816b3e55518321ae7 is corrupted

From reading the link below it looks like I'll have zap it and start over.

http://www.miek.nl/s/7e76eadefe/

I should add that the root cause of this was disk corruption on my virtual machine, which wasn't shutting down correctly. — Benbob, Apr 26 '11 at 21:12

score 11 · Accepted Answer · edited Oct 18 '11 at 13:04

11

Do you have clones of this repository elsewhere? You might want to read this post by Linus Torvalds to restore that corrupted object, assuming the corrupted object is a blob (file contents).

edited Oct 18 '11 at 13:04

Andrew

7,286
3
28
38

answered May 26 '10 at 00:09

Bram Schoenmakers

1,599
14
19

No clones. I just set this up yesterday. So I only had 10 commits. I ended up starting fresh. But will definitely push it somewhere else at the end of every day. Lesson learned. Still glad to be off svn. Git is fast! – Benbob May 26 '10 at 00:36
5

Would be helpful to include the gist of the post in your answer, just incase, say, kernal.org ever gets hacked and is offline :( – SpoonMeiser Sep 12 '11 at 09:08

score 8 · Answer 2 · answered Aug 06 '10 at 15:49

I wound up in the same situation, probably due to an improper shutdown of the virtual machine I was working in. There were approximately 10 objects in .git/objects that had zero length. As far as I can tell, the actual source code files were fine, just the repository was hosed.

$ git status
fatal: object fbcf234634ee04f8406cfd250ce5ab8012f92b08 is corrupted

Per some suggestions I saw elsewhere (including Linus's post referenced above), I tried temporarily moving the corrupted objects git was complaining about from .git/objects elsewhere. When had moved all of them, I got:

$ git status
fatal: bad object HEAD

After about an hour of Googling and trying various solutions, I gave up and started a new working copy using 'git clone' to pull from the origin (which was about 2 hours behind my working copy). I then used rsync -rC (-C excludes SCM files) to copy the changed files from the messed-up working copy to my new working copy.

Same thing here VM corruption. rsync worked perfectly - nothing lost, just the log which was easy to recreate. Thanks! — RidingTheRails, Aug 22 '12 at 10:23

score 2 · Answer 3 · answered Mar 30 '12 at 13:27

2

You could also try to restore these objects by merely copying them from other repositories.

My virtual machine crashed while recording a pushed commit, so the objects were safely stored on a local computer. I scp'ed them to virtual machine and voila — git fsck outputs no errors.

answered Mar 30 '12 at 13:27

Denis Gorbachev

507
4
12

it worked for me (git clone and copy/paste all my source code from broken folder to cloned folder). – kingsjester Nov 07 '22 at 13:33

score 1 · Answer 4 · answered Jun 29 '12 at 07:21

1

Simply delete the corrupt object that git is complaining about. I was able to resolve the same issue just now this way.

fatal: object 985a4870e7d890b314d2794377045a8b007c7925 is corrupted

For the above error, I was able to find corresponding object at:

project_directory/.git/objects/98/5a4870e7d890b314d2794377045a8b007c7925

Where you can see the file is 0 bytes and deleting it allowed the fetch to start working.

Presumably the previous fetch was interrupted, leaving the corrupt object with size = 0 bytes.

answered Jun 29 '12 at 07:21

Anson Kao

5,256
4
28
37

bizarre that this actually works, kind of like kicking the TV to clear up the static! – peter karasev Nov 20 '12 at 19:28
1

Didn't work for me and it sounds like a bad idea considering Torvalds' several page instruction on how to fix the corrupt object (listed above in the answer). – Ain Tohvri Dec 21 '12 at 01:32
Didn't work for me neither. My object is nowhere to be found – guillaume guerin Mar 29 '16 at 16:37

score 1 · Answer 5 · edited May 19 '14 at 14:37

Had the same problem, whichever git command I ran, It ended up with the message:

fatal: object <hash> is corrupted

I didn't have a backup and didn't want to lose my commits, so I decided to try Jase's solution and removed the 0 length file I had : .git/objects/00/<hash> Then got the same:

$ git status
fatal: bad object HEAD

Then, I tried to know what was wrong and looked into .git/refs/heads/masterwhere I had the hash.

I looked into .git/logs/refs/head/masterand found lines like this one:

<old commit> <new commit> <author> <timestamp> commit: <commit message>

I removed the last line (which had =) and pasted of this line into .git/refs/heads/master, erasing its content

I was then able to commit successfully.

score 1 · Answer 6 · answered Jul 16 '15 at 16:05

I had this same issue. I noticed that I wasn't logged in as root. When I logged in as root, I was able to check the log without the error sign.

To solidify this good status, I did this:

git add .
git commit -a -m "stabilize git"

I exited out of root and tried pulling from a client. It worked out for me afterwards.

When I did the add and commit, I knew that I was fine with what was in the directory. I had no changes visible through "git status".

Git log: fatal object [sha1] is corrupted

6 Answers6

Linked