Re: rogue tree objects, how to diagnose?

From: Jeff King <peff@peff.net>
Date: Mar 1, 2009, 03:20 UTC
Message-ID: <20090301032035.GB30384@coredump.intra.peff.net>
In-Reply-To: <20090228.191206.75012167.davem@davemloft.net>

On Sat, Feb 28, 2009 at 07:12:06PM -0800, David Miller wrote:

Show 9 quoted lines

> From git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next-2.6
>  * branch            master     -> FETCH_HEAD
> fatal: unable to read tree c9ee57c5c20c3b7a2d7784a4172aef8b34c3a844
> davem@sunset:~/src/GIT/net-next-2.6$ git show c9ee57c5c20c3b7a2d7784a4172aef8b34c3a844
> fatal: bad object c9ee57c5c20c3b7a2d7784a4172aef8b34c3a844
> davem@sunset:~/src/GIT/net-next-2.6$ 
> 
> Now, what's funny is that this tree object does exist in my
> tree on master.kernel.org:

"bad object" is about an object that did not parse correctly, as opposed to one that is missing. So it's possible that your repository is corrupted.

> I don't understand how this can happen, and I'd like to be able
> to diagnose this intelligently.

I'm not up on the latest techniques for verifying and fixing corruption. But I think with recent enough git, you can grab a good copy of the object from another repo and put it in

  .git/objects/c9/ee57c5c20c3b7a2d7784a4172aef8b34c3a844

that git would do the right thing.

> So, what is an easy way to figure out what commit references tree
> object X?

Try:

  git log --all --pretty=format:'%H %T' | grep c9ee57c5c | cut -d' ' -f1

keeping in mind that there may be multiple such commits.

-Peff