Re: [Dovecot] Wrong message information reported shortly after delivery
On Thu, 2008-03-06 at 08:21 -0800, Ron Avriel wrote:
If an IMAP fetch is issued some "short" time after a message was delivered with dovecot deliver then dovecot reports wrong (zero) values for that new message. Afterward, it reports the information OK.
What OS/filesystem do you use? I don't see how this is possible. The mails are first written to tmp/ directory and then they're atomically rename()d to the new/ or cur/ directory.
I also tested this myself, rapidly sending the same FETCH command 1000 times/sec while delivering 40MB mails to the maildir. The size was always correct.
I'm using standard out of the box Redhat EL 4 with an ext3 filesystem. Since I can reproduce it easily on different RHEL4 machines, tell me what additional information you'd like.
Thanks again, Ron
____________________________________________________________________________________Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs
On Mar 7, 2008, at 9:36 AM, Ron Avriel wrote:
On Thu, 2008-03-06 at 08:21 -0800, Ron Avriel wrote:
If an IMAP fetch is issued some "short" time after a message was
delivered with dovecot deliver then dovecot reports wrong (zero) values for that new message. Afterward, it reports the information OK.What OS/filesystem do you use? I don't see how this is possible. The mails are first written to tmp/ directory and then they're atomically rename()d to the new/ or cur/ directory.
I also tested this myself, rapidly sending the same FETCH command
1000 times/sec while delivering 40MB mails to the maildir. The size was always correct.I'm using standard out of the box Redhat EL 4 with an ext3 filesystem. Since I can reproduce it easily on different RHEL4 machines, tell me
what additional information you'd like.
Can you reproduce it without your IMAP client? For example using
imaptest (http://imapwiki.org/ImapTest) as the client and running:
./imaptest logout=0 - fetch=100 user=testuser pass=testpass
host=imap.host.org
Or if you can give me access to some server where you can instruct me
how to reproduce it, that would work too.
participants (2)
-
Ron Avriel
-
Timo Sirainen