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

Re: Yet another checksum error

From: <gward_at_mems-exchange.org>
Date: 2003-01-23 15:50:02 CET

On 22 January 2003, Karl Fogel said:
 :-) Don't worry, we'll get it solved. So far, we've had three reports
 of checksums problems: yours, nas's, and Justin's. [...] Yours and
 nas's symptoms are very similar, too. Whatever the problem is, we'll
 get to the bottom of it, with a little digging.

No surprise there: Neil (NAS) and I are co-workers, and we're using the
same repository. Said repository has a fairly short history: it was
created last week with cvs2svn, went immediately into production use
(svn 0.16 and http everywhere), and had one ugly hiccup when I Ctrl-C'd
a running svnlook. Ran svnadmin recover on it, after which things
seemed to be fine (even though it took nearly two hours!). We're now in
the process of upgrading to svn 0.17 (0.17.1 in my case) and switching
from http to ra_svn. The move to 0.17 on the server has caused no end
of trouble.

 Greg Ward gward@mems-exchange.org writes:
  $ svn up
  U memsnet/docroot/links/fabs.ht
  A instrument
  A instrument/test
  svn: A checksum mismatch occurred
  svn: (charset conversion failed)
 What platform are you running the client on?

Debian 'unstable', with the client built locally from
subversion-r4503.tar.gz -- ie. 0.17.1). I'm pretty sure I uninstalled
all svn-related Debian packages --

  dpkg -l '*subversion*' '*svn*' '*neon*' '*apache2*' '*apr*'

lists nothing. Wanted to make sure I was using only code from the
0.17.1 tarball.

[me, reporting a problem yesterday]
 svn: A checksum mismatch occurred
 svn: apply_window: checksum mismatch after applying text delta
    expected checksum: 32669b7770e5019e9bd5e8760e4b7f0a
    actual checksum: 2e9ef1ab231b053336d93d5c19151abb

Good thing Neil saved a copy of the repository that was giving us
problems before he dumped/reloaded -- using the broken repos, I can
reproduce this error today. Here's what I did:

  $ svn co http://.../broken/trunk broken-trunk
  $ cd broken-trunk
  $ rm -rf instrument
  $ svn up

It failed exactly the same as yesterday -- good! So let's dig in...

 How do the contents of
 look to you?

No such file. .svn/text-base contains the two files that were
successfully fetched in the update, and .svn/tmp/text-base is empty.

 What about the one really in the repository? (Can you
 retrieve the repos one by other means, so you can compare them?)

On the client, I did
  svn cat http://.../broken/trunk/instrument/grabber/default.css | md5sum
which returned 2e9ef1ab231b053336d93d5c19151abb -- ie. the actual
checksum from the above error message.

On the server:
  svn cat file:///data/subversion/repository-checksum-errors/trunk/instrument/grabber/default.css | md5sum
returned exactly the same checksum.

So where the heck is that expected checksum coming from? (I checked
.svn/entries, and there's no entry for default.css there.)

 What are the checksums of the real files (use the `md5sum' utility to
 find out), and do they match the checksums given above?

Not sure what you mean -- there's no default.css in my working copy, nor
in .svn/text-base, nor in .svn/tmp/text-base.


Greg Ward - software developer                gward@mems-exchange.org
MEMS Exchange                            http://www.mems-exchange.org
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 14 02:09:24 2006

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.