On Sun, 30 Sep 2007, Timo Sirainen wrote:
On Thu, 2007-09-27 at 17:07 -1000, Julian Cowley wrote:
Lines are broken for clarity. This is dovecot 1.0.5 on CentOS 4.5. .. imap(i_malloc+0x1b) [0x80b0e1b] -> imap [0x8083037] -> imap [0x80833e6] -> imap(index_storage_search_init+0xf4) [0x80836f4] ->
Are you sure it's 1.0.5? This looks a lot like the crash that empty messagesets caused with post-1.0.5 hg version: http://hg.dovecot.org/dovecot-1.0/rev/f85adea35b74
You're right, it's not strictly 1.0.5 since I added the following patches. I'm not using the above empty messagesets patch though. Should I be?
My patch file name hg revision ================================ ============ dovecot-cmd-create-fix.patch 6bbb20cdaf1e dovecot-deliver-dup-db-fix.patch 9d4e89fa62d0 dovecot-mailbox-idle-check.patch afb48abd91c0 dovecot-search-not-fix.patch d144e54f09e6 dovecot-search-opt-p1.patch fa89431f893e dovecot-search-opt-p2.patch 92b2bd6b6c2d dovecot-search-opt-p3.patch 678626efb942
In addition there are a few patches in my spec file which are inherited from the original by Axel Thimm, but these have been in there for donkey's years and are unlikely to be related to the problem.
Getting a proper gdb backtrace of the crash would anyway be better than this raw backtrace: http://dovecot.org/bugreport.html
Or at least I'd like to know what gdb says with:
gdb /usr/local/libexec/dovecot/imap l *0x8083037 l *0x80833e6
Unfortunately, it looks like I've built dovecot with stripped binaries, so there is no symbol information for the above locations. I'll rebuild with -g enabled and see if the bug can be reproduced.
BTW, I have an strace of the crash if you are interested. I haven't posted it here since it contains user information, but I could send it directly to you.