Help with splitting SVN database into smaller bits

I'm starting to convert from CVS to Subversion. I have a single CVS
repository that holds code for a number of entirely separate projects.

I ran the database through cvs2svn, and it imported cleanly. However, now
I'm starting to understand the way SVN and cvs2svn organises branches and
tags, I would like to separate the projects out. The data is currently
organised as...

trunk/project1
trunk/project2
branches
branches/branch1/project1
branches/branch2/project1
branches/branch1/project2
etc.

There are about 900 revisions.

I would like to reorganise it into separate repositories

project1
 trunk
 branches

project2
 trunk
 branches

I read the book (well, at least the first half), and figured I needed to do
svnadmin dump followed by some svndumpfilters. But I am having serious
trouble with svndumpfilter.

I assume (from experiment) that I can't do something like

svndumpfilter '*/project1'

So I made myself a list of all the Node-path: lines in the dump file, then
pruned the list to just include the top-level nodes I wanted - e.g.
branches/branch1/project1
branches/branch2/project1
trunk/project1

The file contained about 100 wanted nodes (lots of branches and tags I
wanted to keep).

I then ran
svndumpfilter --drop-empty-revs --renumber-revs include `cat list`
on the dump file, but the resultant file contained almost no nodes.

So I approached it from the other direction, and made a list of all the
nodes I wanted to delete, and ran
svndumpfilter --drop-empty-revs --renumber-revs exclude `cat list`
Again, the resultant file did not contain all the nodes I wanted.

Have I found a bug, or do I just misunderstand how it is supposed to work?

-- 
Nikki Locke, Trumphurst Ltd.      PC & Unix consultancy & programming
http://www.trumphurst.com/
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org