my  block

where uzairs roam

I posted a message to the Subversion dev mailing list last night asking if there’s a formal spec for the dump format. Fingers crossed, I’ll get some answers.

In the meanwhile, I’ve been playing around with dumpfiles generated from my test repository to see how the parser responds to variable whitespace etc. My Clearcase-to-SVN script is just about done in that the output looks right-ish, but trying to load them into SVN causes various errors. Some testing has revealed this about the format: * if the length of a property field is incorrect, svnadmin: Dumpstream data appears to be malformed is thrown * the order of properties is significant — I tried moving svn:date above svn:author and got svnadmin: Dumpstream data appears to be malformed again * something a lot of conversion scripts are bitten by, apparently, the svnadmin: File already exists: filesystem 'test/db', transaction '1-1', path 'blah' message is basically SVN’s version of Clearcase’s evil-twin error message. It means that a node has Node-action ‘add’ when a node with the same Node-path already exists in the repository, or in the dump file being loaded. * it’s not particularly sensitive to whitespace (huge relief). Extra lines/spaces at the end of a node dump or the dump file itself are ignored, and I’ve tried replacing the space in Node-path: trunk with a tab, with success. On the other hand, throwing an extra space into a properties field without incrementing the property length is not allowed, for good reason.

Armed with that, and a proper test environment, I’m hopeful this will start working soon.

(Posts on my Paris trip to come — I have one that’s been ready for several days that I haven’t uploaded in the hopes that I’d put it up with pics. Never mind…it’s going up today without.)