uw-imap & Cygwin secrets revealed, at least a bit!

[Robert Schmidt]

Christian Weinberger wrote:
> The initial problem was:
> - uw-imap on cygwin ran well with Outlook Express
> - but hung upon connection when using Outlook 2003
> 
> In the end I made it to get it work with the following workaround:
> In inetd.conf, I don´t start the daemon directly, but use the following wrapper 
> script.
> 
> #!/bin/sh
> tee NUL | /usr/sbin/imapd

That's an interesting workaround.  My problem sounds slightly different, 
but may be related:

I'm using Outlook XP (2002), and the uw-imapd processes *occasionally* 
hang at very unpredictable times.  When they hang, they consume 100% 
CPU, until terminated.

My current workaround is a script running in the background, kicking in 
every 15 minutes, killing off uw-imapd processes that have been 
consuming more than 10 minutes of CPU time.  This is sort of OK, as I 
don't use this server for interactive work.  I've never noticed any 
corruption in the IMAP store.

Before I had this script, however, the problem was exasperating...  I 
had to ssh to my server about twice a day to manually kill off 5-10 
uw-imapd processes that had gone haywire during the day.  You can 
imagine that just logging in remotely was a pain in these cases, with a 
1000% load on the CPU... :-(

For the record, I've been updating my cygwin installation regularly in 
the last 2 years, and the problem has been constant ever since I started 
using uw-imapd about a year ago.  Also, I always access the IMAP port 
through a ssh tunnel - not sure if that's an important factor.

I'll try your suggestion as well.

Thanks,
Robert

> This slows down communication a bit, but I can connect and had no more hangs up 
> till now. So the original problem seems to be timing specific.
> 
> My guesses are:
> - the cygwin port of uw-imap may have a problem here. I couldn´t find any 
> similar error reports for any other platform uw-imap is running on.
> - cygwin causes the problem itself
> - inetd is out of the line since I also tested with xinetd and had similar 
> effects
> 
> I´d be glad to assist one of the "gurus" going deeper into this.
> 
> Even more, I´d like to invest more time again to debug it myself, but my gdb 
> knowledge is non-existing. If someone would show me how to start, the I 
> definitely would do so. My concern is that usual debugging approaches will fail 
> because of the timing related cause of the problem.
> 
> Thanks to all of you so far!
> Christian
> 
> 


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/