git fsck not identifying corrupted packs

Hi,

I have a pack that contains a corrupted object. It is an old corrupted repo that I have conserved.

As expected, git gc cries out loud about it. It indicates an inflate error (data stream error with incorrect data check), and then the impossibility to read an object from a certain offset in the pack.

However, git fsck does not complain at all about the repo. I guess that for speed reasons, git fsck does not try to inflate the objects.

Is there a means to have fsck to a truly full check on the sanity of a repo?

This both on git 1.6.5.1 and 1.6.4.2.

Thanks

Sergio Callegari

Sergio Callegari schrieb:

git fsck --full

RTFM, please.

-- Hannes

Hi,

On Mon, 19 Oct 2009, Johannes Sixt wrote:

Now, now.

If you were to test a new filesystem, say, wonderfulfs, and wanted to check its integrity, would you not just run "fsck-wonderfulfs" if that exists, rather than reading the fantamagastic manual? Would you not expect that it Does The Right Thing? Would you not expect that it follows the Law Of Minimal Surprise?

So FWIW I can see where Sergio is coming from.

Ciao, Dscho

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Linus and other git developers from the early days trained their fingers to type the command, every once in a while even without thinking, to check the consistency of the repository back when the lower core part of the git was still being developed. Developers who wanted to make sure that git correctly dealt with packfiles could deliberately trigger their creation and checked them after they were created carefully, but loose objects are the ones that are written by various commands from random codepaths. It made some technical sense to have a mode that checked only loose objects from the debugging point of view for that reason.

    Side note.  I think the help description of --full option is wrong (or
    at least stale).  We always look at alternate object store these days
    since e15ef66 (fsck: check loose objects from alternate object stores
    by default, 2009-01-30).  It probably should read "check packed
    objects fully" or something.

The above paragraph is merely a historical background, and in this case the "history" refers to early-to-mid 2005. Even for git developers there no longer is any reason to type "git fsck" in fear of some newly created objects might be corrupt due to recent change to git these days.

The reason we did not make "--full" the default is probably we trust our filesystems a bit too much. At least, we trusted filesystems more than we trusted the lower core part of git that was under development ;-)

Once a packfile is created and we always use it read-only, there didn't seem to be much point in suspecting that the underlying filesystems or disks may corrupt them in such a way that is not caught by the SHA-1 checksum over the entire packfile and per object checksum. That trust in the filesystems might have been a good tradeoff between fsck performance and reliability on platforms git was initially developed on and for, but it might not be true anymore as we run on more platforms these days.

It probably makes sense to ship 1.7.0 with a version of "fsck" in which "--full" is the default; it would still accept "--full" but it would be a no-op. This would be a backward incompatible change, but the difference is primarily about performance ("it takes a lot longer than before!"), and not correctness, so we probably can live with it. As I already said, there is not much reason to run "fsck" every five minutes anymore to begin with (unless your filesystem is so unreliable that it might eat one file every five minutes, that is).

It probably is also a good idea to add a "--loose" option that does what "fsck" currently does without "--full". It is a good name because (1) to people who do not know the internal of git, it means "check only loosely", which would discourage them from running "fack" with that option to begin with, and (2) to others, it exactly tells what the option makes the command check.

(Not CCing everyone, since this is mostly curiosa in the "using git as it was never intended" section):

On Monday 19 October 2009 13:03:42 Junio C Hamano wrote:

Filesystems are mostly reliable, but only until your crazy users do strange and terrible things. I have a real, non-toy environment where I use this stack as a [horrible] workaround for some issues beyond my control:

git -> ext4 -> lvm -> dmcrypt -> loop -> sshfs -> cygwin sshd -> SMB share

Amazingly, this works pretty reliably with many gigabytes of data in a git repository, even with the occasional crash because of flakiness with the "sshfs -> cygwin sshd" piece of the puzzle. But a good "git fsck" sure doesn't hurt in this environment! =)

måndag 19 oktober 2009 21:27:48 skrev  Wesley J. Landaker:

The obvious follow up question here is: Why?

-- robin

On Tuesday 20 October 2009 09:41:50 Robin Rosenberg wrote:

[...]

My main point was to illustrate that having "git fsck" do a REALLY GOOD CHECK is still desirable, as we still haven't reached the days of file- system utopia where nothing ever gets corrupted (even with a smaller, simpler stack).

The actual application where I use this stack is because of odd requirements and circumstances like data must be physically stored on a particular Windows server on the network that uses a weird authentication method that samba doesn't support, and it has to go over the network encrypted anyway, there are lots of holes in the data, so I want ext4 for the extent support, file-size limitations on the target, etc.

It's a really an exotic love-hate mix between an off-by-one-please-no-never- again kind of situation coupled with a bit of "because I can".

If you are both nerdy and morbidly curious enough to care, send me a "but, no ... really, WHY?!" with the git list CC dropped and we can talk about details and/or other crazy stuff. (I don't want to get wildly off-topic on this list.)

Junio C Hamano <gitster@pobox.com> writes:

Thanks for the historical background.

+1

+1 too.

Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:

Actually, I changed my mind. I do not think this so big that we need to wait for a major version bump. Why not shoot for 1.6.6?

On Tue, Oct 20, 2009 at 08:45, Junio C Hamano <gitster@pobox.com> wrote:

--no-full works

Hi,

On Tue, 20 Oct 2009, Alex Riesen wrote:

It works. Technically. For human users, though, --loose-objects-only (with a shortcut "--loose") would be better.

Disclaimer: this email was written by a bot and is valid without signature

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

OTOH, the advantage of "--no-full" is that it's compatible with existing Git versions. If I learn Git 1.6.6 with --no-full, and use it in a script, then my stript works also with older Gits.

But anyway, I think very few people are actually interested in "git --no-full" (or call it whatever you like), so I don't think this is very important.

Linus and other git developers from the early days trained their fingers to type the command, every once in a while even without thinking, to check the consistency of the repository back when the lower core part of the git was still being developed. Developers who wanted to make sure that git correctly dealt with packfiles could deliberately trigger their creation and checked them after they were created carefully, but loose objects are the ones that are written by various commands from random codepaths. It made some technical sense to have a mode that checked only loose objects from the debugging point of view for that reason.

Even for git developers, there no longer is any reason to type "git fsck" every five minutes these days, worried that some newly created objects might be corrupt due to recent change to git.

The reason we did not make "--full" the default is probably we trust our filesystems a bit too much. At least, we trusted filesystems more than we trusted the lower core part of git that was under development.

Once a packfile is created and we always use it read-only, there didn't seem to be much point in suspecting that the underlying filesystems or disks may corrupt them in such a way that is not caught by the SHA-1 checksum over the entire packfile and per object checksum. That trust in the filesystems might have been a good tradeoff between fsck performance and reliability on platforms git was initially developed on and for, but it may not be true anymore as we run on many more platforms these days.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
---

Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> writes:

For human users, I think --full vs --no-full is quite a nice suggestion, given that we already have advertised --full and people know the option.

Also people know that splicing "no-" after the double dash is often the way to negate a boolean-looking option.

The actual patch to do this is tiny, but that is just a bonus ;-)

 Documentation/RelNotes-1.6.6.txt |   10 ++++++++++
 Documentation/git-fsck.txt       |    5 +++--
 builtin-fsck.c                   |    2 +-
 3 files changed, 14 insertions(+), 3 deletions(-)

On Tue, 20 Oct 2009, Junio C Hamano wrote:

             ^^

Might be worth mentioning explicitly that, because of that change, plain fsck is now going to take much longer to complete.

Nicolas

Nicolas Pitre <nico@fluxnic.net> writes:

Sounds fair; thanks.

On Mon, 19 Oct 2009, Junio C Hamano wrote:

Agreed. With a prominent note in the release notes to point people at it when they don't read release notes and complain that fsck suddenly became very slow after they upgraded.

Nicolas

On Tue, Oct 20, 2009 at 20:39, Nicolas Pitre <nico@fluxnic.net> wrote:

I have a feeling that it either wont be noticed at all (i.e., I have run fsck with something other than --full only one time in my whole bash history, and never shown it otherwise to anyone else), or people will immediately like it ("Oh, finallly! Now that feels like it is doing something!" :)

Johannes Sixt wrote:

Right... sorry for the noise, I mismatched --strict for --full in a script.

BTW, the short help for fsck at --full only says "consider objects in alternate repositories".

My apologize.

Sergio

On Monday 19 October 2009 04:56:07 Sergio Callegari wrote:

Until I read this thread, I didn't realize you needed --full to check objects in packs.

Since just every git repository I ever use has 99%+ of it's objects in packs, this means every time I've run "git fsck" it's essentially been a no-op and I didn't know it. I imagine this is a common confusion.

Also, having --full mean both "check alternate object pools", and "check objects in packs" seems to be rolling up two orthogonal issues.

But anyway, here is a patch that at least fixes the short option help to match the manual and the current behavior:

--- 8< ---
From 8fc3cd68d496bf00faad4f0a7b6ae4fee9437e68 Mon Sep 17 00:00:00 2001
From: Wesley J. Landaker <wjl@icecavern.net>
Date: Mon, 19 Oct 2009 12:48:07 -0600
Subject: [PATCH] Update git fsck --full short description to mention packs

The '--full' option to git fsck does two things:

  1) Check objects in packs
  2) Check alternate objects

This is documented in the git fsck manual; this patch reflects that in the short git fsck option help message as well.

Signed-off-by: Wesley J. Landaker <wjl@icecavern.net>
---
 builtin-fsck.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

"Wesley J. Landaker" <wjl@icecavern.net> writes:

Same here.

IMHO, something like this should be applied to "maint", this is kind of a serious bug indeed. Just check the "alternate" thing, according to Junio:

Junio C Hamano <gitster@pobox.com> writes:

On Mon, Oct 19, 2009 at 11:11:33AM +0200, Johannes Sixt wrote:

That still does not catch everything. About a week ago I wanted to check out a branch in a local repo and I got an error that it was corrupt. But "git fsck --full" only complained about some dangling objects, it did not notice the corruption. I used git version 1.6.4.3 (Debian unstable at that time). It would be nice to have a "git fsck --i-really-want-to-check-everything".

Gabor