From: Nicolas Ferre Date: Mon, 03 Nov 2008 16:18:53 GMT Subject: Re: error: packfile while git fsck Message-ID: <490F246D.1040307@atmel.com> In-Reply-To: Nicolas Pitre : > On Wed, 22 Oct 2008, Nicolas Ferre wrote: > >> Nicolas Pitre : >>> On Wed, 22 Oct 2008, Nicolas Ferre wrote: >>> >>>> Hi all, >>>> (please cc me on response) >>>> >>>> I am facing error during git status & git fsck on my tree. >>>> This tree is cloned from various linux kernel trees. >>>> >>>> Here are a sample of the error I see : >>>> >>>> $ git fsck >>>> error: packfile >>>> .git/objects/pack/pack-2ab31ad1f8cb69d091a56fe936634e4796606d49.pack does >>>> not >>>> match index >>>> error: packfile >>>> .git/objects/pack/pack-2ab31ad1f8cb69d091a56fe936634e4796606d49.pack >>>> cannot be >>>> accessed >>> [...] >>> >>> What git version? >> $ git --version >> git version 1.5.3.7 > > OK. Since this is not bleeding edge, it is pretty unlikely that the > corruption is due to git itself. Furthermore, the git packs are always > read only once they've been created, meaning that if they weren't > corrupted at some point then something outside of git caused the > corruption. You really should consider the possible causes for that > (dying disk, pilot error, etc). > > As to recovery... That really depends if you have personal work > committed to your repository. If not then the easiest solution is > simply to recreate it by refetching from upstream. If you have > personal work in there then you could try to fetch your work branch into > the newly created repository. The latest git version could help with > the extraction of non-corrupted objects out of a bad pack, but if the > objects you are interested in are themselves corrupted then your only > hope is to have a copy of those objects somewhere else. Nicolas, Thanks for your help. I will consider migrating to a brand new git tree with my work imported in (it seems that my local work is not affected). Just to know, as the object I try to access is not present, is there a way to rebuild the index to match this mater of fact ? Regards, -- Nicolas Ferre