possible self-deadlock in idle signal handler

[Michael Bacon]

Just as a follow-up on this, I think I've found a way to at least decrease 
the incidence of the problem with a stupid conf file trick.

The problem appears to emerge when the client decides to do something 
simultaneously with the idle alarm going off.  Since most clients (in the 
case of this user, it was Thunderbird) do things in multiples of minutes, I 
changed the imapidlepoll setting from 60 to 65.  (67 might be even better, 
come to think of it.)  I'll have to let it run a few weeks before I know if 
it made any difference, but it may have the effect of de-syncing activity 
on the client and server ends, which would probably help.

-Michael

--On Tuesday, March 31, 2009 9:38 AM +1100 Bron Gondwana 
<brong at fastmail.fm> wrote:

> On Sun, Mar 29, 2009 at 09:12:55PM -0400, Wesley Craig wrote:
>> On 28 Mar 2009, at 23:27, baconm at email.unc.edu wrote:
>> > This looks like something much more than "well, let me fiddle with
>> > it a bit and submit a patch."  Has anyone with the requisite design
>> > perspective taken a hack at solving it?  I'm assuming that running
>> > idled doesn't ameliorate the problem.
>>
>> Running idled probably makes the problem worse, not better.  I don't
>> know that anyone besides me has even reviewed this code.  And you can
>> see how many people rushed to comment on the issue :)
>
> Yep ;)
>
> Bron ( taking a week off Cyrus to work on some other stuff, mostly )