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

Re: Always reproducible base checksum mismatch error

From: David Glasser <glasser_at_davidglasser.net>
Date: 2007-11-22 20:49:49 CET

On Nov 22, 2007 2:13 PM, David Glasser <glasser@davidglasser.net> wrote:
> On Nov 22, 2007 2:10 PM, David Glasser <glasser@davidglasser.net> wrote:
> > On Nov 22, 2007 1:10 PM, Philipp Wollermann <philipp@igowo.de> wrote:
> > > Hi!
> > >
> > > David Glasser wrote on 2007-11-22:
> > > > Do you have the ability to test against trunk svn?
> > >
> > > I just tested against the trunk, it didn't change anything.
> >
> > Both client and server, right?
> >
> > Hmm. After running your script, can you give the output of
> >
> > svn info url://to/the/file -r123
> >
> > where 123 is replaced with each revision from 1 to the last revision committed?
> >
> > And also the error message again, and the contents of the .svn/entries
> > file again (the whole thing).
> >
> > Thanks!
>
> Actually, the info calls won't give me the checksums I want; can you instead:
>
> * run the script until it fails
> * send the output of the script (including error message)
> * send tarballs of the working copy and repository

Philipp sent me these tarballs (I encourage you to send them to the
list instead so everyone can help...)

r2 is broken. The rev file revs/2 contains the delta rep for the
change to 'test' in r2, but it is not referred to from anywhere: the
contents dump for the "/" directory claims that the node "test" is
"file 1.0.r1/52", that is, the unchanged file from r1. There is no
noderev for the r2 version.

Philipp, are you sure you were able to reproduce this with trunk
client *and server*? There definitely have been recent relevant
changes, such as r27256 (currently nominated for backport to 1.4.6 but
lacking enough votes).

If you still can reproduce it with trunk client and server, try
applying the attached patch (to the server, at trunk) and see if you
still can reproduce? This patch more or less disables the "directory
cache" in FSFS (though if this doesn't fix it I can give you a patch
that disables it more completely).

Index: subversion/libsvn_fs_fs/fs_fs.c
===================================================================
--- subversion/libsvn_fs_fs/fs_fs.c (revision 27972)
+++ subversion/libsvn_fs_fs/fs_fs.c (working copy)
@@ -2628,6 +2628,7 @@ svn_fs_fs__rep_contents_dir(apr_hash_t *

   /* If we have this directory cached, return it. */
   if (! svn_fs_fs__id_txn_id(noderev->id) &&
+ FALSE &&
       ffd->dir_cache_id[hid] && svn_fs_fs__id_eq(ffd->dir_cache_id[hid],
                                                  noderev->id))
     {

-- 
David Glasser | glasser_at_davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Nov 22 20:50:00 2007

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.