470907 - Thunderbird always uses port 119 for NNTP, ignoring port configured in server settings

Status: RESOLVED FIXED

(MailNews Core :: Networking: NNTP, defect)

Description

[Alex Krastelev]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.0.5) Gecko/2008120122 Firefox/3.0.5
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1b3pre) Gecko/20081222 Shredder/3.0b2pre

In Thunderbird 3.0b1 and the latest Shredder build (20081222),
Thunderbird cannot read news messages if a news server uses non-standard NNTP port (as configured in Account Settings --> Server Settings --> Port). 

In my case, the local news server runs on port 2119 instead of the standard port 119. Thunderbird can download and display headers as expected. 
However, clicking on any message to display its body causes the following error:

"The connection was refused when attempting to contact 192.168.1.2:119"
(note port 119, not 2119 in the error message).

This server used to work with Thunderbird 2.0.

Reproducible: Always

Steps to Reproduce:
1. Configure a newsserver using a non-standard port different from 119.
For testing, it is possible to simply forward port 2119 to some other nntp server, using socat utility:
$ socat  TCP4-LISTEN:2119,fork,reuseaddr      TCP:news.microsoft.com:119


2. Configure new newsgroup account in Thunderbird (File-->New Account-->Newsgroup Account, enter the local server name)

3. Open settings of the newly created account, go to Server Settings, change NNTP port to 2119

4. Close account settings, right-click the account, click Subscribe

5. Subscribe to any newsgroup (microsoft.public.access), let Thunderbird download headers from the newsgroup

6. Attempt to open or preview any message by clicking it
Actual Results:  
The news message doesn't open, an error is displayed:
"The connection was refused when attempting to contact 192.168.1.2:119"
the error message shows port 119 instead of the configured port 2119

Expected Results:  
The news message should open, as connection to the server is possible (Thunderbird can obtain list of newsgroups and list of messages in the newsgroup).


Thunderbird 2.0 was working without any problem in the same configuration

Comment 1

[Mark Banner (:standard8)]

This works fine for me with my local tests.

Can you try getting a protocol log as described here: https://wiki.mozilla.org/MailNews:Logging - this should give us more information as to what is going on.

Are you connecting via SSL? It may be worth retrying on today's nightly which is out in the next couple of hours, as there was one configuration issue fixed, although I do doubt it will fix your problem.

Comment 2

[Sascha Grage]

I can confirm this bug with
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1b3pre) Gecko/20081223 Shredder/3.0b2pre

NNTP port at localhost:120

Comment 3

[Sascha Grage]

Attachment: Logfile

Comment 4

[Mark Banner (:standard8)]

That's an interesting log, and explains why I may not have seen the problem.

It looks like we initially get the port right, so that we're able to get information about the messages etc, but then when you select a message, we're reverting to the old port.

Does anyone have any idea when this started to be a problem (just pre or post beta 1 initially would be useful)?

Confirming based on comment 2.

Comment 5

[David :Bienvenu]

I wonder if the reworking of the nntp connection cache might be involved? That was post b1.

Comment 6

[mathewm]

I can confirm this problem with:

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.1b3pre) Gecko/20090223 Thunderbird/3.0b2

Comment 7

[Phil Lacy]

confirming on windows port 1198
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1b3pre) Gecko/20090312 Shredder/3.0b3pre
Unable to log results.

Comment 8

[Mark Banner (:standard8)]

Based on the comments, this is a problem that people will hit and an obvious regression. I think we need to fix this so setting blocking TB3.

Comment 9

[Phil Ringnalda (:philor)]

Either the code or the comment from bug 213729 in nsNntpService::DisplayMessage certainly seems wrong, since the comment says "We need to set the port on the url, just like nsNNTPProtocol::Initialize does" and then proceeds to set it to the default port, unlike nsNNTPProtocol::Initialize which sets it to the port in the URI it was handed.

Comment 10

[Mark Banner (:standard8)]

Attachment: The fix

Fix with testcase. As Phil said in comment 9, we're currently always assuming the default port.

Note, I've also included a fix in nsNNTPNewsgroupList::CleanUp to prevent/fix a debug assertion occurring. The assertion was:

###!!! ASSERTION: invalid min or max param: 'min <= max && min > 0', file /Users/moztest/comm/main/src/mailnews/base/util/nsMsgKeySet.cpp, line 1075

I think this is occurring because with the testcase when we get to nsNNTPNewsgroupList::CleanUp both m_knownArts.set and m_set are empty. A few lines down in CleanUp, firstKnown and lastKnown (obtained from m_knownArts.set) are both 0, when the loop starts processing, firstKnown gets incremented (because its zero) but lastKnown doesn't. Hence the assertion.

As m_set->FirstMissingRange will just return if we're not in debug mode, and as a result nothing would end up happening, I think its reasonable to do the same if both m_knownArts.set and m_set are empty.

I couldn't reproduce the problem on my debug build when subscribing to a new newsgroup so it could just be the way the testcase is written (i.e. not doing a proper display of the message or something), but I'm still not sure we should assert in this case.

Comment 11

[David :Bienvenu]

I got a test failure when running on windows with this patch:

*** exiting
*** TEST-UNEXPECTED-FAIL | ../../../mozilla/_tests/xpcshell/test_news/unit/test_
getNewsMessage.js | From: John Doe <[e-mail]>
Date: Sat, 24 Mar 1990 10:59:24 -0500
Newsgroups: test.subscribe.simple
Subject: H2G2 -- What does it mean?
Message-ID: <[e-mail]>

What does the acronym H2G2 stand for? I've seen it before...
 == From: John Doe <[e-mail]>
Date: Sat, 24 Mar 1990 10:59:24 -0500
Newsgroups: test.subscribe.simple
Subject: H2G2 -- What does it mean?
Message-ID: <[e-mail]>

What does the acronym H2G2 stand for? I've seen it before...

JS frame :: C:/mozilla-build/msys/tbirdcheck/mozilla/testing/xpcshell/head.js ::
 do_throw :: line 101
JS frame :: C:/mozilla-build/msys/tbirdcheck/mozilla/testing/xpcshell/head.js ::
 do_check_eq :: line 120
JS frame :: ../../../mozilla/_tests/xpcshell/test_news/unit/test_getNewsMessage.
js :: anonymous :: line 41
native frame :: <unknown filename> :: <TOP_LEVEL> :: line 0
JS frame :: ../../../mozilla/_tests/xpcshell/test_news/unit/test_getNewsMessage.
js :: anonymous :: line 32
*** exiting
*** TEST-UNEXPECTED-FAIL | ../../../mozilla/_tests/xpcshell/test_news/unit/test_
getNewsMessage.js | From: John Doe <[e-mail]>
Date: Sat, 24 Mar 1990 10:59:24 -0500
Newsgroups: test.subscribe.simple
Subject: H2G2 -- What does it mean?
Message-ID: <[e-mail]>

What does the acronym H2G2 stand for? I've seen it before...
 == From: John Doe <[e-mail]>
Date: Sat, 24 Mar 1990 10:59:24 -0500
Newsgroups: test.subscribe.simple
Subject: H2G2 -- What does it mean?
Message-ID: <[e-mail]>

What does the acronym H2G2 stand for? I've seen it before...

JS frame :: C:/mozilla-build/msys/tbirdcheck/mozilla/testing/xpcshell/head.js ::
 do_throw :: line 101
JS frame :: C:/mozilla-build/msys/tbirdcheck/mozilla/testing/xpcshell/head.js ::
 do_check_eq :: line 120
JS frame :: ../../../mozilla/_tests/xpcshell/test_news/unit/test_getNewsMessage.
js :: anonymous :: line 41
*** FAIL ***
WARNING: nsExceptionService ignoring thread destruction after shutdown: file C:/
mozilla-build/msys/tbirdcheck/mozilla/xpcom/base/nsExceptionService.cpp, line 19
4
recorder: started(36), aborted(31), completed(5), different header(0), trees tra
shed(0), slot promoted(0), unstable loop variable(1), breaks(2), returns(0), uns
tableInnerCalls(0)

Comment 12

[Mark Banner (:standard8)]

Attachment: The fix v2

We seem to be sending everything across the interface right (i.e. using crlf), its just the storage at the mailnews side varies crlf according to platform. Therefore I've just added a crlf -> lf replacement before the comparison function.

Comment 13

[David :Bienvenu]

Comment on attachment 368220 [details] [diff] [review]
The fix v2

sadly, this is still failing for me on windows. The code looks right, though. I'm sure I've got the right patch and I have rebuilt...I'll try it in an other tree just to be sure.

Comment 14

[David :Bienvenu]

Comment on attachment 368220 [details] [diff] [review]
The fix v2

test fails on an other machine as well

Comment 15

[Mark Banner (:standard8)]

Attachment: The fix v3

Ok, this should use the form of the replace function that works...

Comment 16

[David :Bienvenu]

Comment on attachment 368274 [details] [diff] [review]
The fix v3

great, that worked.

one nit - can use existing declared rv here instead of declaring a new one.

+        nsresult rv = server->GetSocketType(&socketType);

Comment 17

[Mark Banner (:standard8)]

Checked in: http://hg.mozilla.org/comm-central/rev/1d0fcdefa375