[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

RE: [PATCH v2] Add svnrdump

From: Bert Huijben <rhuijben_at_collab.net>
Date: Wed, 14 Jul 2010 21:25:39 +0200

> -----Original Message-----
> From: Ramkumar Ramachandra [mailto:artagnon_at_gmail.com]
> Sent: woensdag 14 juli 2010 18:02
> To: Stefan Sperling
> Cc: dev_at_subversion.apache.org; Bert Huijben; Daniel Shahaf; Will Palmer;
> David Michael Barr; Jonathan Nieder; Sverre Rabbelier; Git Mailing List
> Subject: Re: [PATCH v2] Add svnrdump
> Hi Stefan,
> Stefan Sperling writes:
> > Playing with svnrdump and comparing its output to the output of
> > svnadmin dump --deltas, I noticed that:
> Thanks for testing!
> > - svnrdump doesn't dump revision 0.
> > It should dump revision 0, because that revision can contain
> > revprops such as metadata for svnsync (svn:sync-last-merge-rev etc.)
> Yeah, I forgot to ask about this: passing 0 as an argument to the
> replay API doesn't seem to work. Why? How do I dump revision 0 then?
> > - You're missing a couple of fields:
> > The UUID of the repository.
> > Text-content-sha1
> > Text-delta-base-md5
> > Text-delta-base-sha1
> Yes, I'm aware. Since these fields aren't strictly necessary, I
> decided not to take the extra effort to print them out: you'll notice
> that I'm printing the md5 sum that the server gives me instead of
> calculating anything. SHA1 sum would require /some/ calculation. UUID
> and text-delta-base-md5 aren't a big deal though: I'll fix these
> later.

We added the sha1 field to the format in 1.6, as we have it available in the
fs layer anyway and we might want to prefer it over md5 in later version
(for future proofing the dump format). It's not used by our tools yet and we
don't send the value over the ra layer yet. Without revving the editor layer
it will be pretty hard to calculate it remotely even from the most recent

I assume you don't get the SHA1 values when you use a recent svnadmin dump
on an older repository. (Untested statement)

> > - I've seen a "Prop-delta: true" line which svnadmin dump does not
> Correct. `svnadmin dump` has a logic for determining when the prop is
> really a delta (as opposed to a delta against /dev/null). Since
> there's no harm printing extra Prop-delta headers, I decided not to
> implement this logic.

Do you know if this is this something as simple as: 'Is this a new node?' or
if this is some advanced scheme?
> > - You're missing some newlines that svnadmin dump prints (cosmetic,
> > but it would be nice if both produced matching output).
> This isn't in the dump-load-format spec document (atleast afaik), and
> it's very hard to get this right (yes, I tried). Moreover, it's very
> ungratifying to have a few extra newlines (reverse engineered from
> `svnadmin dump`) printed at the end of 10+ hrs of work; yes, that's
> what I estimate it'll take to fix this.
> > How to reproduce what I'm seeing:
> > Use svnsync to get a copy of the numptyphysics repository at
> > https://vcs.maemo.org/svn/numptyphysics (I had a dump of that lying
> > around... other repositories might do the job just as well, of
> > Dump the repository using svnadmin dump --deltas.
> > Dump the repository using svnrdump.
> > Compare output with diff -u.
> Right. My validation script (validate.sh in the original repository)
> runs the following filter on the diff and validates if nothing seeps
> through. In other words, I know that these differences exist, and have
> determined that they're safe.
> gawk '$0 !~ "Prop-delta: true|Text-delta-base-|sha1|Text-copy-source-|^-
> $" && $0 ~ "^+|^-" { print; }'

Your mail explains Prop-delta, sha1, but what about these Text-delta-base
and Text-copy-source lines?

Received on 2010-07-14 21:26:37 CEST

This is an archived mail posted to the Subversion Dev mailing list.