[Dovecot] 1.0-stable loses flag changes sometimes?

[Chris Wakelin]

Thanks! I'm trying the dovecot-stable fix now (so far so good!)

I've also re-run my test69 test (three messages delivered at once whilst 
idle) using test70 and I get the same as before. This time I've included 
an extra flag fetch first so you can see the extra fetch responses 
really are bogus, the flags haven't changed.

. SELECT INBOX
* FLAGS (\Answered \Flagged \Deleted \Seen \Draft)
* OK [PERMANENTFLAGS (\Answered \Flagged \Deleted \Seen \Draft \*)] 
Flags permitted.
* 717 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1111502437] UIDs valid
* OK [UIDNEXT 24799] Predicted next UID
. OK [READ-WRITE] Select completed.
. FETCH 691:700 (FLAGS)
* 691 FETCH (FLAGS (\Seen))
* 692 FETCH (FLAGS (\Seen))
* 693 FETCH (FLAGS (\Seen))
* 694 FETCH (FLAGS (\Seen))
* 695 FETCH (FLAGS (\Seen))
* 696 FETCH (FLAGS (\Seen))
* 697 FETCH (FLAGS (\Seen))
* 698 FETCH (FLAGS (\Seen))
* 699 FETCH (FLAGS (\Seen))
* 700 FETCH (FLAGS (\Seen))
. OK Fetch completed.
. IDLE
+ idling
* 693 FETCH (FLAGS (\Seen))
* 694 FETCH (FLAGS (\Seen))
* 695 FETCH (FLAGS (\Seen))
* 720 EXISTS
* 3 RECENT
DONE
. OK Idle completed.
. LOGOUT
* BYE Logging out
. OK Logout completed.

This doesn't happen with dovecot-stable. I'm wondering whether it might 
be a different issue, after all. If it's only giving spurious but 
correct responses, then I guess a real client/user wouldn't notice.

Best Wishes,
Chris

Timo Sirainen wrote:
> On Fri, 2005-05-13 at 15:21 +0100, Chris Wakelin wrote:
> 
>>It seems similar to the 1.0-stable issue, only the spurious FETCH 
>>responses seem to give the wrong flags then (but I can't reproduce this 
>>on demand).
> 
> 
> I think the basic idea is this:
> 
> - Update the flags for the last message
>   --> because of lazy-writing Dovecot just marks the message dirty
> - New message arrives on the mailbox
>   --> Dovecot re-reads the last message, forgets the dirty state
> - close the mailbox
>   --> Because the dirty state is forgotten, the flag is never written to
> mbox
> 
> Fixed in CVS now for both 1.0-test and 1.0-stable.
> 


-- 
--+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+-
Christopher Wakelin,                           c.d.wakelin at reading.ac.uk
IT Services Centre, The University of Reading,  Tel: +44 (0)118 378 8439
Whiteknights, Reading, RG6 2AF, UK              Fax: +44 (0)118 975 3094