Re: [PATCH] fsckObjects tests: show how v2.17.1 can exploit downstream

From: Jeff King <peff@peff.net>
Date: May 29, 2018, 21:24 UTC
Message-ID: <20180529212458.GC7964@sigill.intra.peff.net>
In-Reply-To: <20180529211950.26896-1-avarab@gmail.com>

On Tue, May 29, 2018 at 09:19:50PM +0000, Ævar Arnfjörð Bjarmason wrote:

Show 11 quoted lines

> Something that's known but not explicitly discussed in the v2.17.1
> release notes, or tested for, is that v2.17.1 will still happily pass
> on evil .gitmodules objects by default to vulnerable downstream
> clients.
> 
> This could happen e.g. if an in-house git hosting site is mirroring a
> remote repository that doesn't have transfer.fsckObjects turned on.
> Someone can remotely push evil data to that remote hosting site
> knowing that it's mirrored downstream, and the in-house mirror without
> transfer.fsckObjects will happily pass those evil objects along, even
> though it's been updated to v2.17.1.

Sure, I agree with all of the above, but...

> It's worth testing for this explicitly. So let's amend the tests added
> in 73c3f0f704 ("index-pack: check .gitmodules files with --strict",
> 2018-05-04) to show how this can result in a v2.17.1 client passing
> along the evil objects.

I'm not sure what testing this buys us. If it stopped passing them along, would we consider that a regression? This seems more like a documentation issue than something that needs to be tested.

-Peff