1480972 - "connection to server timed out" makes message unmovable until restart

Status: RESOLVED DUPLICATE of bug 1469117

(MailNews Core :: Networking: IMAP, defect)

Description

[Magnus Melin [:mkmelin]]

On trunk I've been getting quite frequent "connection to server timed out" messages on IMAP. The message appears almost instantly after trying to move a message so I suspect there is no real timeout.

After that, the message won't move in any way until I restart Thunderbird.

Not sure if this is a recent regression, but I never used to see this and now it's quite frequent.

Comment 1

[Magnus Melin [:mkmelin]]

Attachment: imap-conn-fail-log.txt

Logs for such an incident (the relevant parts, I think)

Comment 2

[Alfred Peters [:infofrommozilla]]

Dupe of Bug 1492007?

Comment 3

[gene smith]

The attached log shows tb responding to an IDLE response and trying to send DONE. But the SendData fails because the transport connection is gone (which may be also why it exited IDLE which would be right before the start of the log file). It looks like the server really did FIN (close) the connection but can't really tell without a true network dump. Or may tb sent the FIN.

Next tb tries the move and then requests STATUS but fails at both since the connection is gone. Then it looks like a new connection is opened successfully and the log stops.

Hard to say if it's a dup of bug 1492007 but maybe. It is reported as specific to Zimbra but this bug is for Exchange according to the log.

I doubt it but it sounds sort of related to Bug 1469117, but I think this regression fix should be in Trunk now.

When you retry the move, is there any error notification then? 

What if you move to another folder, maybe even on another account, and then come back to the problem folder. Does that allow moves to work again?

Are you only unable to move files from one problem folder or is it for all folders on the Exchange account?

Comment 4

[Magnus Melin [:mkmelin]]

I stopped seeing this once bug 1469117 landed.

(Coming back to the folder wouldn't help, only a restart could make it work again, as far as I remember)

Comment 5

[Wayne Mery (:wsmwk)]

So this was ultimately a regression of bug 1430659

Comment 6

[gene smith]

There were no timer problem at v52 before the changes linked here were made: Bug 1469117 comment 4. So these changes caused the regression that I observed. Maybe bug 1430659 was an earlier update that got broken, at least for tb, by the linked changes. 

Of course, this assumes what Magnus saw was actually caused by this socket timeout problem.