374058 - IMAPS connection exceeded (sporadic, reproducible)

Status: RESOLVED DUPLICATE of bug 392846

(Thunderbird :: General, defect)

Description

[Filippo Carletti]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.0.3) Gecko/20060523 Ubuntu/dapper Firefox/1.5.0.3
Build Identifier: Version 1.5.0.10 (20070221)

After automatic update to 1.5.0.10 on all windows xp workstations began to appear sporadic but frequent connection errors to our local imap server.
Server is SME Server (smeserver.org), using sslio from ipsvd (smarden.sunsite.dk/ipsvd) linked with matrixssl.
Upon client connection, server logs a first good connection, then, on second connection:
2007-03-14 12:42:29.760993500 sslio[14433]: fatal: ssl decode error: illegal parameter

It never happened before the 1.5.0.10 update.

A similar problem is described in bug 352530 so I tried to disable tls and the problem disappeared.

Server accepts max 12 concurrent connections from pc.
TB is configured for 5 and has only one account open on server.

Reproducible: Sometimes

Steps to Reproduce:
1.Connect to smeserver in imaps
2. Change folder

Actual Results:  
See TB popup about connection closed


Using TB 1.5.0.10 on a mac os x client doesn't trigger the problem.

Comment 1

[Filippo Carletti]

I added tls support to my secure imap server and I no longer experience connection problems.
But the client shouldn't use TLS, I choose SSL.
The decode error usually gets logged on the second connection, the first seems to always succeed (I did check the log carefully but it's not easy).

Comment 2

[Filippo Carletti]

How could I show that TB doesn't use SSL even if configured to do so?

This is a regression, it appeared in 1.5.0.10.

Comment 3

[Filippo Carletti]

In bug 368611, Charlie Brady identified a bug in MatrixSSL affecting SMTP over SSL. Using the same patched version of MatrixSSL for imap connections fixes this problem.

Comment 4

[Andrew Sutherland [:asuth] (he/him)]

Marking as a duplicate of bug 392846 since it sounds like the actual problem was a bug in MatrixSSL, but bug 392846 implements a work-around for the MatrixSSL bug in NSS.  If I somehow misunderstood the situation, please re-open.