Re: [Evolution] send/receive race condition on sending mail

[Not Zed <notzed ximian com>]

On Thu, 2003-05-29 at 01:25, Mike Godfrey wrote:
> I'm running evo 1.3.92 on redhat 9, but this is a long-standing problem.
>
> I read my mail using IMAP over SSL; I send outgoing mail to the same
> server using SMTP (no SSL).
>
> Problem #1:
>
> Sometimes, my IMAP connection dies and I get a string of error messages
> to the effect that the message/folder I want to access doesn't exist. If
> I want it to fix itself, I either wait 10 minutes, or if I hit the
> send/receive button and it renews the imap connection right away.

i think we've tried fixing the problem but not with much success.  we
want to rewrite imap - and hopefully address this sort of issue - but
its a big task.  if it gets held off again, we'll probably have to look
at the existing stuff again.  at least the workaround is simple.

> Problem #2:
>
> If I send mail while evo thinks the imap connection is dead, the
> outgoing mail sits (correctly) in my local outbox until the connection
> is brought back up, and then it is magically delivered.  However,
> sometimes it sits for quite a while.
>
> If I hit send/receive, the connection is renewed and the outgoing mail
> is sent on its merry way.  However, sometimes hitting send/receive
> causes more than one copy to be sent out (ie evo managed to send it out
> already, but left the message sitting in the Outbox queue).
>
> My guess is that the locking mechanisms are not correctly done, causing
> a race condition, but I'm not a mail protocol implementor, so maybe
> that's the "correct" behaviour.  But it is annoying that the mail gets
> sent, but evo lets it sit in my mailbox so that when I hit send/receive
> a few minutes later, a second copy is sent out.  Surely, some sanity
> checking should be done with the outgoing queue to make sure a message
> isn't being sent multiple times.

i would think that the problem is that your sent folder is on the imap
server.  so it sends the message, and then tries to append it to your
sent folder.

still, it shouldn't try sending again while the last sending task is
still running.  there's probably a bug in the sent folder failing case -
its supposed to just delete the mail from outbox anyway, even if it
couldn't append to the sent folder, but it could send it.

submit a bug report to bugzilla.ximian.com

some CAMEL_VERBOSE_DEBUG=1 logs might be useful.

> Alternatively, perhaps send/receive (what outlook does) could be split
> up (what mozilla does).

well in reality it is split.  there's no real reason why sending should
be conflicting with receieving, except that one or the other might mean
the dialogue is displayed longer.

> Any comments/suggestions?
>
> PS this ongoing issue of imap connections dying led one of my colleagues
> to drop using evo altogether, which is a shame since one of his grad
> students is an ex-ximian employee.

"oh well"