Re: BUG: Major problem, data loss

[Brian Stafford <brian stafford uklinux net>]

On Thu, 23 August 09:56 Carlos Morgado wrote:
> 
> On 2001.08.22 20:58:50 +0100 M . Thielker wrote:
>
> > I am using SMTP. The string is contained in cyrus imapd and appears there
> > because I am using an IMAP sentbox.

SMTP is totally unconnected with Cyrus IMAP.  There are two ways to get a
message
into a cyrus imapd mailbox.  One is to pipe the message into the delivery
program that sendmail uses.  I believe the delivery program also supports
LMTP - a SMTP derivative without message queueing semantics, the HELO/EHLO
command replaced with LHLO and a few other changes - see RFC 2033.
The alternative is to use IMAP itself, deliver the message as if it were being
copied from another mailbox.  This is probably not such a good idea as the 
semantics are likely to be wrong for message delivery.

> > Apparently, libbalsa_message_queue uses mutt_write_fcc to writ ethe message
> > to the outbox first, then to the fccbox. That second invocation of
> > mutt_write_fcc caused the message, but the email should, at that time, have
> > already been written to the outbox. So chances are good that it actually did
> > go out and I just lost my local copy.
> 
> yeah, that sounds right. hopefully, you just lost your copy.
> 
> > However, this points to a problem in formatting messages for storage in
> > IMAP, probably another case of not quoting strings that may contain spaces.

Hmmmm.... you should be able to deliver/copy syntactically incorrect messages
as a matter of robustness.  Sounds like a bug in the IMAP server.

> yes, imperative now that inbox in imap works.
> it would be good to have a message that breaks before starting to hack at it.
> you have any idea on what exactly chocked on cyrus and caused that message ?
> 
> anyone a cyrus expert ? :)

I looked at the code once.  Just the once, it scared me off.  Still,
at least it's not as ugly as sendmail's code.

Brian