Mailing-List: contact dev-help@subversion.apache.org; run by ezmlm
Precedence: bulk
Received-SPF: pass (athena.apache.org: domain of gstein@gmail.com designates
 209.85.212.43 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAJjMeYP7CCKDC2ZbtHEQz+ymp0FFmq4EZ069E7dTTKeQouJ=bg@mail.gmail.com>
References: <20120206174836.DB1AC238889B@eris.apache.org>
	<CABD8fLUg4uoBsM58sDcGK1EyKzCExuAN4bBNABv7E-rgFnQQfw@mail.gmail.com>
	<CAJjMeYP7CCKDC2ZbtHEQz+ymp0FFmq4EZ069E7dTTKeQouJ=bg@mail.gmail.com>
Date: Mon, 6 Feb 2012 15:03:00 -0500
Message-ID: 
 <CABD8fLXSYSq6hC9Q4KbNWot9WCO_rYY6-m1VKPr_kW2wzn5eoQ@mail.gmail.com>
Subject: Re: svn commit: r1241097 - in /subversion/trunk/subversion:
 libsvn_client/repos_diff.c libsvn_delta/compat.c libsvn_fs_base/tree.c
 libsvn_fs_fs/tree.c libsvn_wc/externals.c libsvn_wc/update_editor.c
From: Greg Stein <gstein@gmail.com>
To: Hyrum K Wright <hyrum.wright@wandisco.com>
Cc: dev@subversion.apache.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

On Mon, Feb 6, 2012 at 14:32, Hyrum K Wright <hyrum.wright@wandisco.com> wr=
ote:
> On Mon, Feb 6, 2012 at 1:28 PM, Greg Stein <gstein@gmail.com> wrote:
>> On Mon, Feb 6, 2012 at 12:48, =A0<hwright@apache.org> wrote:
>>> Author: hwright
>>> Date: Mon Feb =A06 17:48:36 2012
>>> New Revision: 1241097
>>>
>>> URL: http://svn.apache.org/viewvc?rev=3D1241097&view=3Drev
>>> Log:
>>> Ev2 shims: Truthfully report our base checksum as being that of the emp=
ty
>>> stream.
>>>
>>> Note: This breaks several assumptions in various delta-editor receivers=
 about
>>> the validity of this checksum. =A0These have been patched to ignore the=
 checksum
>>> if it is that of the empty stream. =A0This will not affect correctness,=
 as the
>>> final checksum, as supplied by close_file() is still used to detect cor=
ruption,
>>> and it hasn't changed.
>>
>> How does this even work? Sure, the checksum isn't checked, but the
>> delta-editor receiver is going to choose the wrong base contents to
>> apply the delta against.
>
> Huh? =A0The delta-editor receiver isn't using this to choose any base
> contents, it's simply using it as some kind of intermediate
> verification step.
>
> Before this change, we just passed a NULL checksum, which was *always*
> ignored, and everything Just Worked. =A0With this change, we started
> being honest, and various pieces of the system didn't appreciate that
> honesty.

If the shim sends a delta against the empty-stream, then how can the
Ev1 receiver properly apply that? Won't the receiver use some selected
base contents? ie. NOT the empty stream?

I'm thinking of the delta application to a base. How can that possibly
work, if the shim creates a delta against the empty stream?

Cheers,
-g