Re: [PATCH v2 8/8] tests: mark tests broken under GIT_TEST_PROTOCOL_VERSION=2
- From
Jeff King <peff@peff.net>
- Date
- Dec 18, 2018, 12:41 UTC
- Message-ID
- <20181218124127.GB30471@sigill.intra.peff.net>
- In-Reply-To
- <8736qv18w4.fsf@evledraar.gmail.com>
On Tue, Dec 18, 2018 at 10:28:27AM +0100, Ævar Arnfjörð Bjarmason wrote:
Show 7 quoted lines
> I.e. have a repo with "master" and a "root-password" branch, where the > "root-password" branch has content that's irresistible to "git repack" > for delta purposes, do we end up sending the "root-password" content > over on a fetch even when that branch isn't advertised & forbidden? > > Or, if that fails are there ways to make it work? E.g. using hidden/* in > combination with delta islands?
Delta islands wouldn't generally help here. They only tell us not to store on-disk deltas that fetching clients aren't likely to be able to reuse (i.e, they want X but will generally not have Y, so don't make a delta there).
In the attacks I mentioned in my previous email, the deltas would usually be computed on the fly for each fetch. So the lack of on-disk deltas across "security boundaries" wouldn't help.
You could disable on-the-fly delta compression, but the resulting packs are much larger (and you'd think it would save some server CPU, but experiments I've done show that it helps a lot less than you'd think, since we often end up zlib-deflating more bytes).
-Peff