325063 - Occasional connection timeouts when message is moved to Sent Folder on IMAP Server

Status: RESOLVED INCOMPLETE

(Thunderbird :: Mail Window Front End, defect)

Description

[Walter F.J. Mueller]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.12) Gecko/20051010 Firefox/1.0.7 (Ubuntu package 1.0.7)
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.12) Gecko/20051010 Firefox/1.0.7 (Ubuntu package 1.0.7)

Setup: using an IMAP account, 'Sent Folder' located on the IMAP server
Action: sending an email
Problem: get occasionally a 'connection timeout' error when the message is copied to the Sent folder. After that, Thunderbird tends respond with connection timeouts for almost all transactions.

Reproducible: Sometimes

Steps to Reproduce:
This behaviour was observed with Mozilla 1.7.x, Thunderbird 1.0.x and 1.5.
It is reported under  http://forums.mozillazine.org/viewtopic.php?p=2018348 and was also seen by others.

Unclear what really triggers this. Here some hints:
* happens rarely with the first email after startup
* happens more often after a long time of inactivity
* when the 'Sent folder' is not on an IMAP server not local, I very rarely observe IMAP connection timeouts on other transactions, but they also happen.

Comment 1

[Aaron Caskey]

I'm having similar problems on a number of win64 boxes running Thunderbird  1.5.0.4. The Windows machines are a 2x 2.4G Dualcore Opterons.

The mail server is running cyrus v2.2.3

After a period of working fine Thunderbird will lock up when trying to copy the message into the Sent folder. Examining the server it seems that the IMAP server is hanging on a fcntl64 call to obtain a write lock on the users Sent folder. Any other actions on that user, from any other client (and the the cyrus deliver program) time out while the server is in this state. Outlook Express works fine on the win64 machines, and the same version of Thunderbird has been working happily on all of our win32 machines for a while, although we have no win32 machines with that hardware.

Comment 2

[David :Bienvenu]

it could be that timeouts aren't working correctly on the dual processor machines, though I thought a fix for at least some platforms had gone into a 1.5.0.2 build. The dual processor bug we've been seeing is ordinary options timing out too quickly, not failed operations never timing out.

See bug 307527 for more info and possible workarounds.

Comment 3

[Wayne Mery (:wsmwk)]

walter wrote: "All problems I described occurred on single processor systems (single core, no hyperthreading). The the issue raised in Comment 2 can be only part of the full story."

Aaron, what's your present status?

Comment 4

[Wayne Mery (:wsmwk)]

Walter, do you see the same problem when using a trunk build?
  http://ftp.mozilla.org/pub/mozilla.org/thunderbird/nightly/latest-trunk/

Note: Aaron's email address is dead

Comment 5

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

No additional useful information since last comment -> resolving incomplete.

Please comment if the issue still occurs in the latest supported 2.0.0.14 / trunk nightlies.