On Thu, 2007-02-01 at 15:44 -0800, Dan Price wrote:
I have some users who have volunteered to experiment with the server, and they are starting to do so. I've also seen the "flags" problems others have mentioned on the list, and I do not think I could deploy to production with that bug affecting the users.
I'll release rc20 in a few hours, and it's fixed there.
I also have some issues which I'm not sure have been previously reported: One user (of my 4 test users) is occasionally seeing a bug in which mutt 1.4.2.1 claims to be "[downloading xxx/30724]" headers. Upon getting very near to the 30724 total, the total rises to something like: "[downloading xxx/50132]".
Never heard of this before..
"Everything was working fine until i quit out of my mail app (Mac mail) and restarted. Then i saw weird things:
- messages wouldn't get marked unread - messages from eons ago started showing up in my Drafts folder
I think I once saw something similar to this when the mailbox's UIDVALIDITY changed. Mail.app didn't understand that change, and instead of fetching the new message with UID 1234, it looked up the old UID 1234 from its cache and showed it to me instead.
The reason for the UIDVALIDITY change then has probably something to do with you still running UW-IMAP(?) also. Or was the user only using Dovecot?
- Inbox would keep spinning/loading - then eventually Inbox just hung showing me no messages"
Don't know about these. I've been using Mail.app fine..
Any thoughts about how to approach these issues? I don't really have a feel for how to go from user-observed-app-doing-X to imapd-is-doing-Y.
rawlog could be helpful in knowing what exactly is happening. http://dovecot.org/bugreport.html
The big challenge for our site may be mailbox size: the median mailbox size is 62M, the average is 176M and the largest is 1.1GB. I'd be curious to know what sort of mailbox sizes other sites have experienced, and how widespread the use of the Dovecot MDA is-- is this likely to make a big difference for us, and will we have to migrate our (approximately 25) procmail users to sieve?
mbox_very_dirty_syncs=yes would at least help if you're not going to use Dovecot MDA.
Having read the recent discussions about "1.0", I think my team would really appreciate knowing how to obtain something that was thoroughly vetted and worked highly reliably. If major changes are coming, I think it'd be great to have a 1.0.x stability focused tree, and a 1.1.x (or 2.x.x) feature tree.
There are major changes in CVS HEAD already. I've mostly been trying to fix the last few bugs, but looks like I managed to break some things with those fixes in rc19..