Re: Recovering from repository corruption

From: Denis Bueno <dbueno@gmail.com>
Date: Jun 10, 2008, 19:38 UTC
Message-ID: <6dbd4d000806101238v2bb975abqd39916e45d4bf866@mail.gmail.com>
In-Reply-To: <m3abhtp42o.fsf@localhost.localdomain>

On Tue, Jun 10, 2008 at 13:55, Jakub Narebski <jnareb@gmail.com> wrote:

Show 41 quoted lines

> Assume that history looks like this
>
>    ...---.---a---*---b---.---...
>
> where by '*' is marked corruped commit (commit shich tree contains
> corrupted blobs).
>
> First, you can check the commit message for '*' using git-cat-file or
> git-show, you can get the difference between 'a' and 'b' using
> "git diff a b".  When you know how repaired commit 'X' should look
> like, do something like:
>
>  $ git checkout -b <temp-branch> 'a'
>  $ <edit edit edit>
>  $ git commit
>
> Then history would look like this
>
>    ...---.---a---*---b---.---...
>               \
>                \-X
>
> Now with grafts make 'b' be a child of 'X', i.e. modify parent of 'b'
> for history to look like below:
>
>    ...---.---a---*   b---.---...
>               \     /
>                \-X-/
>
> Examine history using git-log, git-show, check tree with git-ls-tree
> and examining files, use graphical history browser like gitk.
>
> Then if possible use git-filter-branch to make history recorded in
> grafts file permanent...
>
> HTH
> --
> Jakub Narebski
> Poland
> ShadeHawk on #git
>

Thanks for the help.

My situation was:

    ...---a---*---b---c---d---*---e---...

Following your example, I believe I got this to:

    ...---a---*   b---c---d---*   e---...
           \     /         \     /
            \-X-/           \---/

That is, I replaced the first problematic commit and deleted the second, since I forgot how I changed 'd' to get that commit. I put the following in .git/info/grafts:

    'b' X
    'e' 'd'

(which I gathered from here:
http://thread.gmane.org/gmane.comp.version-control.git/66398/focus=66402.
 I've never use grafts before.  A bit about them should be put in the
manual, if it's not there already. =])

Then I ran:

    git-filter-branch HEAD ^X ^'d'

Now "git log --raw --all" doesn't show any of the problematic SHA-1 hashes anymore!

However:

identity.fb[173] > git fsck --full
    error: 320bd6e82267b71dd2ca7043ea3f61dbbca16109: object corrupt or missing
    error: 4d0be2816d5eea5ae2b40990235e2225c1715927: object corrupt or missing
    missing blob 320bd6e82267b71dd2ca7043ea3f61dbbca16109
    missing blob 4d0be2816d5eea5ae2b40990235e2225c1715927

Shouldn't these be unreferenced now that I've run filter-branch?

-- 
                              Denis