[Dovecot] Dovecot 1.1.1 killed with SIGABRT

Timo Sirainen tss at iki.fi
Mon Jul 21 03:28:37 EEST 2008


On Mon, 2008-07-21 at 01:24 +0200, Andreas M. Kirchwitz wrote:
> First, the "small" problem with Pine 4.64:
> 
> Jul 20 03:17:04 linux dovecot: imap-login: Login: user=<user1>, method=CRAM-MD5, rip=138.201.2.8, lip=138.201.2.2, TLS
> Jul 20 03:17:53 linux dovecot: IMAP(user1): UIDVALIDITY changed (1215137127 -> 1216072777) in mbox file /home/user1/Mail/Trash
> Jul 20 03:17:53 linux dovecot: Panic: IMAP(user1): file mail-index-transaction.c: line 548 (mail_index_transaction_get_next_uid): assertion failed: (recs[count-1].uid >= next_uid)

Fixed: http://hg.dovecot.org/dovecot-1.1/rev/ea00b1553ef1

> However, instead of this panic operation, Dovecot maybe could issue a
> warning or error response to the client so that the user (hopefully)
> gets a clue about what's going wrong.

This should be transparent to the client. After the above fix it still
logs the warning and also a couple of more errors that I should probably
look into why they're happening.

> Jul 12 01:04:45 linux dovecot: Panic: IMAP(user2): file index-sync.c: line 39 (index_mailbox_set_recent_uid): assertion failed: (seq_range_exists(&ibox->recent_flags, uid))

This seems to be happening with some mbox users. I guess I should try to
look at it again, but I haven't been able to reproduce it myself so far.
If you've any idea how to somewhat easily reproduce it, I'd like to
know.

If it still happens randomly, you could just comment out that line.
That'll cause recent flags to be somewhat wrong sometimes, but
Thunderbird doesn't care about them.

> INBOX with 350 messages in total, 3 of them new. Started Thunderbird,
> index page is displayed fine, then clicked on one of the 3 new messages
> to read it. For no apparent reason, Dovecot dies with signal 6.
> Thunderbird reconnected automatically, and then Dovecot dies again.
> Nice endless loop ...

If this still happens, I'd really like to have:

1. Output of "find ~/Maildir" (i.e. a list of all files in maildir)
2. dovecot-uidlist
3. dovecot.index and dovecot.index.log

None of those should contain anything sensitive about the mailbox
contents.

> Stopping and restarting the complete Dovecot subsystem with the help
> of the init.d script didn't help. Finally the crashes stopped after
> removing the cache files in /home/fubar/Mail/.imap/INBOX.
> 
> It's obvious that the cache files got corrupted somehow. There was
> plenty of disk space free. And the cache files have been freshly
> created from scratch with Dovecot 1.1.0. Other folders were not
> affected. Except Dovecot, nobody ever touches these cache files.

By "cache files" do you mean dovecot.index*? The dovecot.index.cache
file should have nothing to do with this problem.

> Well, the good news is, it stopped crashing after removing the cache
> files. The bad news is, without the cache files (which I removed)
> the problem cannot be reproduced.

Yes, it would have been nice to have them.. :)

Was this a completely new Dovecot installation, so Dovecot v1.0 was
never run?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://dovecot.org/pipermail/dovecot/attachments/20080721/b004597a/attachment.bin 


More information about the dovecot mailing list