On Sat, Nov 1, 2014, at 09:47 PM, Bron Gondwana wrote: > So I got sidetracked a bit: > > commit 9bfeeec1d9140ec04e9bbe6119f3626b4238212d > Author: Bron Gondwana <br...@fastmail.fm> > Date: Wed Oct 29 10:46:02 2014 -0700 > > mailbox: support old versions back to v6 transparently > > Makes upgrade from 2.3 not suck so hard > > ================ > > This was a majorly requested feature. It took a lot of work, but I think > it's solid now. It works even with Cyrus 2.2 I hope (I haven't managed to > build it on my machine yet, gotta install an older linux version and run > something up. > > Still to do is making sure that we only write old-format mailboxes.db entries > when writing old format cyrus.index records, and we should be able to > seamlessly upgrade or downgrade for all the known versions, which is pretty > cool. > > The cyrus.cache format hasn't changed appreciably in that time (yet, I have > plans), so it should be fairly backwards compatible. The one thing that did > change is some character encoding, so cyrus.cache based searches might be a > bit strange. > > I still have to do the sync work - sync-based XFER is a nice to have (needed > for CMU, but not a regression) - but I don't want the existing sync code to > be in a release version that we need to support. > > And that's that :) I'm going to keep working on the sync stuff as much as I > can over the next week, and let you know how I'm going by then. I'm sorry > that it's not as released as I had hoped by now - feel free to play with the > master branch at CMU, it's pretty solid already.
Of course, that was assuming no bugs in anything else that went into the 2.5 branch late. Jeroen has been testing 2.5 as itself, and I've been testing the fastmail changes rebased on top of 2.5 - and we're both still hitting lots of issues. I pushed another 20 or so patches back today. I'm really hoping to get 2.5 + our patches into production at FastMail soon, and then I'll feel a lot more confident releasing 2.5 proper. I don't want to give you something half-baked that corrupts data - that's the real fear. So not quite there yet, so sorry. Bron. -- Bron Gondwana br...@fastmail.fm
