On Wed, Aug 15, 2012 at 06:42:11PM +0200, Luca Capello wrote: > Hi there!
Sorry for the long delay. I took come VAC ;-) > However, I get duplicate emails for no obvious reasons. I know that at > least one duplicate come from my synced machine because of its filename: > > Maildir/cur/1345032816.1118_0.mantissa:2, > Maildir/cur/1345042069.M833566P8146.gismo,S=3250,W=3324:2, > Maildir/cur/1345042092.M850696P3594.mantissa,S=3250:2, > Maildir/cur/1345042092.M850697P3594.mantissa,S=3250:2, Can you be more explicit? Here I see 4 messages, 2 of which have a very similar file name. Where is the problem? Are they all the same message? > I have not done anything special with the above mailbox except accessing > it directly from the server between two smd-pull runs. This is > something vital for me: I want to be able to access any copy of my > mailbox whenever/wherever I want. It is designed to let you do that, so in principle it is possible. > While the above happened on gismo (my main laptop, sid with no fancy DE) > with mantissa being the server (with 1.2.5-1~bpo60+1), I experienced the > same on a wheezy "standard" installation (GNOME3, see #682737) where > smd-loop was started by the smd-applet: I am sure that at least at the > first synchronization I got duplicates. Given that my mailbox is quite > big (see #659263), I copied it to the new machine before starting > smd-pull: however, this latter took more than 2 hours and I needed > different tries to have it finished (the problem was the network > connectivity). I would guess that stopping an smd run leaves cruft > behind, which results in duplicates. Hum, killling smd on the way should not be a problem. If it leaves cruft, then it is in the tmp/ directory of the maildir, that is not even displayed by MUAs. However, after some thinking, I think I can trick smd to duplicate a mail in the following way: I start a sync from A -> B, it copies file m from host A to host B, say a message marked as NEW. I kill smd before it finishes. Then I log on B, I read my email, I mark m as OLD. Then I start the sync again from A -> B, m is copied to B again, since it is not there (I moved it from new/ to cur/, possibly altering its header too) and since the premature death of smd prevented it from updating its metadata with the info the m was on B too. So yes, killing smd _while_ changing the mailbox status may lead to duplicates, but you are relly playing against it. Is that what happened? If this is not the case, I've few remarks on the first sync: - If you start copying the mailbox, then no duplicates should pop up. If it does there is a bug, and I need more data to reproduce it. But this is very unlikely. - If you start from 2 similar mailboxes, say kept in sync using offlineimap, then some duplicates may pop up during the first sync, since the _same_ email could be named differently on the two hosts. Indeed in 1.2.5 I've added smd-uniform-names to cope with that. It generates a script to uniform names, to be run before the first sync. Cheers -- Enrico Tassi -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
