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

Strange FSFS rev file...

From: John Szakmeister <john_at_szakmeister.net>
Date: 2005-09-15 11:31:33 CEST

I responded to a user's request on the users@ list a while back, and he just
got back to me in the past few days. He's having trouble checking out from
his repository. Here's the original message from him:

I've been working with Ranjit to try and get to the bottom of this, and to
help him get his repository up and running again (I'd appreciate any help I
could get from the FSFS gurus). I only have the revision file indicated in
the error message (94) since he cannot give me the whole repository. For a
little more background, Ranjit is running Subversion on a linux box and the
repository is in a directory that is local to the machine (i.e., no NFS or
anything of that nature).

I put together some scripts to help parse the rev file and ran across
something interesting. The node rev for the only file committed in r94 has
an interesting text field. Here's the dump of the node rev:
  NodeRev Id: 58.0.r94/2596808
   type: file
   pred: 58.0.r93/1570026
   text: 94 1249299 1347483 10119168 1a03b509ed99a9875da890673e542365
   props: 89 1051993 53 0 113136892f2137aa0116093a524ade0b
   cpath: /Projects/stacys/M320-CBT/StoryBoard.doc
   copyroot: 0 /

There are two things that are interesting about the text field. First, this
is the only file in the commit, and the last one in the chain, so I would
expect the offset to the text rep to be 0. When I go and look at offset
1249299, I'm in the middle of an svndiff. Doesn't make sense.

So, I changed the offset in the text rep to be 0000000, and attempted to parse
the rep. I get to offset 1250990 in the file, and then things fall apart
(the parameters to the svndiff window don't make much sense). However, the
data there looked to be a repeat of data that I saw previously in the file.
So I started investigating a little more and found something very
interesting. It turns out that there is a great deal of repeat data in the
file. In fact, there is a block starting at offset 1691 that is an exact
repeat of the block at offset 1250990. Here's the kicker: the block is
*exactly* 1249299 bytes in length. Everything else about the file appears to
be correct. With the exception of the repeated block, the svndiff windows
seem to be reasonable, all the dir node revs appear are correct, all of the
prop reps are correct. The only oddity is this repeat block.

I'm not sure where to go from here. I'm sure Ranjit would like to be able to
use his repository again, so if anyone more experienced can help him (and me)
out, I'd appreciate it. I have r94, so if there is something simple that I
can do to help him move on with life while we take a look at this problem
further, I think that would be the best solution. Right now he can't check
out from his repository, and the file is also corrupted in the backup. I've
already gotten permission from Ranjit to be able to give the file to maxb or
ghudson. If there is someone else, let me know, and I'll talk to Ranjit
about getting permission to give r94 to them (jpieper?).

I also searched the user archives, and this doesn't appear the first occurence
of a similar problem. Uwe Kitzmann also emailed the list with the same

Uwe's issue went unresolved because he was also unable to provide the
repository. I have a sneaking suspicion that the problem is related though.
I'm going to ping him and see if he can't at least provide the bad revision
to me as well.


To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Sep 15 11:32:20 2005

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