[Dovecot] alpha5: (message_body_search_ctx) assert

[Jeff A. Earickson]

Timo,
   One of the two people who caused this assert has touched their
mailbox via POP (Qualcomm POP, I haven't switched my POP service
to dovecot yet).  The X-UIDL would have been added by either
qpopper or Eudora most likely.  People are warned not to do
POP and IMAP at the same time, but it happens anyway.

Jeff Earickson
Colby College

On Sat, 31 Dec 2005, Timo Sirainen wrote:

> Date: Sat, 31 Dec 2005 15:13:11 +0200
> From: Timo Sirainen <tss at iki.fi>
> To: Jeff A. Earickson <jaearick at colby.edu>
> Cc: dovecot at dovecot.org
> Subject: Re: [Dovecot] alpha5: (message_body_search_ctx) assert
> 
> On Thu, 2005-12-22 at 10:41 -0500, Jeff A. Earickson wrote:
>> Hi,
>>    My second case of this assert, gdb analysis of the core
>> dump attached.  My setup: Solaris 9, dovecot built with
>> gcc 4.0.2, mbox format, imap usage only.  INBOX is NFS
>> mounted from another Solaris 9 system, index files are local.
>> Syslog complaint is:
>>
>> imap(user): file message-body-search.c: line 414 (message_body_search_ctx):
>> assertion failed: (input->v_offset == part->physical_pos +
>> part->header_size.physical_size)
>
> This happens if message headers are modified in mbox. Looks like in your
> case something added 30 bytes there. Could there be something else
> besides Dovecot modifying the mbox? Or if Dovecot added/modified some
> headers that aren't hidden, I'd like to know about that too. The hidden
> headers are: Content-Length, Status, X-IMAP, X-IMAPbase, X-Keywords,
> X-Status and X-UID. Hmm. Maybe X-UIDL should be added to that list..
> Yes, I'll do that.
>
> Anyway, I think I should also add some checks for this instead of
> letting it assert-crash..
>
>