Timo Sirainen wrote:
Fixes:
- Crashes in non-x86 64bit systems
- Bugs in cache file header caching (you probably should delete all dovecot.index.cache files to be sure the bug won't haunt you in future)
- FETCH ENVELOPE patch by Chris Wakelin (I'll try to figure out what to do with the list patch later)
Known bugs left:
- Thunderbird + maildir: moving lots of messages from mailbox to another may cause deadlock (lock timeouts in half a minute or so and fails the operation).
- I guess I should try to look at the Thunderbird + POP3 + lots of messages -> DELE error too again..
- I've heard that 1.0-tests are a lot slower than 0.99.x or 1.0-stable when checking new mail (with kmail), ie. client asking STATUS of all mailboxes. I've looked at their strace outputs and the difference (5-10x slower) comes from 1.0-test waiting longer in poll() for client input, which doesn't really make any sense to me. Can anyone verify this or think of reasons why this could happen? Maybe I'm setting some socket option differently, although I don't think I've changed anything..
After upgrading from 1.0-test76 to 1.0-test78 (i386 on OpenBSD3.7), I immediately started seeing a bunch of these errors in the maillog:
Jul 14 07:44:15 mail dovecot: POP3(marilynearmstrong): file mbox-file.c: line 144 (mbox_file_seek): assertion failed: ((((char *)data - (char *)NULL) & 7) == 0) Jul 14 07:44:35 mail dovecot: POP3(bdjohnston): file mbox-file.c: line 144 (mbox_file_seek): assertion failed: ((((char *)data - (char *)NULL) & 7) == 0) Jul 14 07:47:52 mail dovecot: POP3(tracyh): file mbox-file.c: line 144 (mbox_file_seek): assertion failed: ((((char *)data - (char *)NULL) & 7) == 0) Jul 14 07:48:05 mail dovecot: POP3(nedwilliams): file mbox-file.c: line 144 (mbox_file_seek): assertion failed: ((((char *)data - (char *)NULL) & 7) == 0)
I'm not sure how this affected the users since I quickly reverted back to 1.0-test76 (it's a production environment that I recently migrated from RH Linux to OpenBSD). I'll be glad to provide any other information that may be useful.
Thanks,
-- Emmett "Buddy" Pate