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

Re: Strange merge behaviour with rc9

From: Kari Grano <kari.grano_at_gmail.com>
Date: Fri, 20 Jun 2008 20:23:01 +0300

On Thu, 19 Jun 2008 09:12:22 +0100 Julian Foad wrote:
> Thanks for the report.
> It will help very much if you can show us a transcript, meaning copy
> and paste the actual commands and the output that Subversion prints.

Hi Julian,

Thanks for replying. I do have an actual case where the error manifests
itself. Unfortunately, the repository is not public, and I tried to
generalize the issue in my previous mail. Since then, I've managed to
narrow the problem down: it seems that Subversion has problems with repeated
merging of conflicting properties.

I've attached a script that reproduces the problem on Windows platform.
Notes: 1) you need to replace the repository creation command with 'svnadmin
create'; I don't have that tool installed since I'm using the Collabnet
client SVN distribution. 2) Save the script as .bat file instead of .txt.

When run, the script fails with

svn: In directory 'dir'
svn: Error processing command 'append'
svn: Missing 'dest' attribute in 'dir'

I'm using rc9, but I suspect the problem is also present in the 1.5.0


> Hi Subversion devs,
> I've recently stumbled into a problem trying to perform a largish
> merge using TortoiseSVN 1.5.0-RC3 (see
> http://svn.haxx.se/tsvn/archive-2008-06/0192.shtml). The problem
> appeared to be twofold: first of all, TSVN was merging very slowly
> compared to the SVN command-line client. This issue is being dealt
> with by Stefan Kng. The other problem is that SVN merge is acting in a
way I don't understand.
> Specifically, it seems that the speed and outcome of
> svn merge --accept postpone -r A:B URL
> and
> svn merge --accept postpont -r A:A1 -r A2:A3 ... -r An:B URL
> differ quite a lot. In my testcase the range A:B is non-contiguous
> with about 60 subranges (A:A1, A2:A3, ..., An:B). The first command
> runs very fast (about 40 seconds) and correctly to the end. The second
> case runs slowly (about 5 minutes) and ends in an error.
> I can somewhat understand the speed problem, but shouldn't the
> outcomes of the two commands be the same? The subranges in question
> represent all the relevant revisions on 'URL', ie. there is no reverse
> cherry picking or anything like that going on (in fact, the ranges
> have been obtained by using the TSVN log output for 'URL').
> The error I'm getting in the second case is
> Error: In directory 'd:\va88\uit\libs'
> Error: Error processing command 'append' in 'd:\va88\uit\libs'
> Error: Missing 'dest' attribute in 'd:\va88\uit\libs'
> Error: Error reading spooled REPORT request response
> This happens when about 27 submerges have been performed. I have no
> idea what 'append' or 'dest' mean here.
> I ran both tests using the 1.5.0-rc9 command line client for Windows.
> The server and repositories are using SVN 1.4.3 on Apache 2.0.59. Serf
> is not in use. The access protocol is http.
> Thanks,
> Kari.

To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org

Received on 2008-06-20 19:23:19 CEST

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.