1690415 - IRC Chat stuck on "Connecting..." after a network outage. Should show timeout

Status: RESOLVED FIXED

(Chat Core :: General, defect)

Description

[diomede979]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/88.0.4324.104 Safari/537.36

Steps to reproduce:

connect to an IRC server, disconnect Internet connection from the OS (for example by disabling the connection) until the IRC client detects disconnection and then reconnect.

Actual results:

the IRC connections stuck in "Disconnecting..." state. Then everytime I do need to force the reconnection manually.

Expected results:

as in the older versions, after a disconnection a countdown starts until the reconnection.

Comment 1

[WaltS48 [:walts48]]

Could not find a dupe or reproduce using 78.7.1 on Fedora 33 Workstation.

Connected to Chat.
Turned off my wireless connection.
Chat disconnected as expected.
Turned my wireless connection on.
Chat reconnected.

I should reconnect while the application is still showing disconnecting?

If it isn't completely disconnected why should it reconnect?

Comment 2

[diomede979]

(In reply to WaltS48 [:walts48] from comment #1)

Could not find a dupe or reproduce using 78.7.1 on Fedora 33 Workstation.

Connected to Chat.
Turned off my wireless connection.
Chat disconnected as expected.
Turned my wireless connection on.
Chat reconnected.

I should reconnect while the application is still showing disconnecting?

If it isn't completely disconnected why should it reconnect?

In my case, che application stucks in "Disconnecting..." state, even if network come back again.
In older versions, a countdown starts and when fire the connection starts again.

Comment 3

[WaltS48 [:walts48]]

I'm connecting to chat.freenode.net using IRC.

I tested with 68.11.0 on Linux and did see that countdown and reconnected when I either removed my wireless dongle and reinserted it or just disconnected my user.

Testing with 78.7.1 on Windows 10 and disconnecting my wireless account.

When I reconnect the Wi-Fi account, I get an "Error: Peer reports it experienced an internal error" message in my Instant messaging status dialog and the account does not reconnect. I need to click the "Connect" button.

In the Error Console I see:

Bad certificate or SSL connection for (username)@chat.freenode.net:
Peer reports it experienced an internal error. irc.jsm:921

I removed my chat name from that error.

Do you see something similar?

Comment 4

[Rob Lemley [:rjl]]

 NS_ERROR_XPC_NOT_ENOUGH_ARGS: Not enough arguments [nsIRoutedSocketTransportService.createRoutedTransport] socket.jsm:535
    _createTransport resource:///modules/socket.jsm:535
    onProxyAvailable resource:///modules/socket.jsm:367
    openIMAccountWizard chrome://messenger/content/mailCore.js:649
    oncommand chrome://messenger/content/messenger.xhtml:1

Looks like I missed a spot in bug 1692610.

Comment 5

[diomede979]

(In reply to WaltS48 [:walts48] from comment #3)

I'm connecting to chat.freenode.net using IRC.

I tested with 68.11.0 on Linux and did see that countdown and reconnected when I either removed my wireless dongle and reinserted it or just disconnected my user.

Testing with 78.7.1 on Windows 10 and disconnecting my wireless account.

When I reconnect the Wi-Fi account, I get an "Error: Peer reports it experienced an internal error" message in my Instant messaging status dialog and the account does not reconnect. I need to click the "Connect" button.

In the Error Console I see:

Bad certificate or SSL connection for (username)@chat.freenode.net:
Peer reports it experienced an internal error. irc.jsm:921

I removed my chat name from that error.

Do you see something similar?

how to show status dialog? I don't have it.

Comment 6

[WaltS48 [:walts48]]

In the Chat tab click the "Show Accounts" button.

Do you see the same message in the Error Console that I do about a bad certificate?

Comment 7

[diomede979]

(In reply to WaltS48 [:walts48] from comment #6)

In the Chat tab click the "Show Accounts" button.

Do you see the same message in the Error Console that I do about a bad certificate?

No. But i guess it's just a problem related to Freenode service. Try with some other IRC server.

Comment 8

[WaltS48 [:walts48]]

What one are you having a problem with?

Comment 9

[diomede979]

(In reply to WaltS48 [:walts48] from comment #8)

What one are you having a problem with?

I am having problems with all of them, for example: irc.azzurra.org

Comment 10

[WaltS48 [:walts48]]

Well, I've already wasted enough of my time on this and have documented my results.

For review:

• I didn't find a duplicate.
• I have no problem reconnecting using Thunderbird release and beta on Linux.
• I need to click the "Connect" button in the Instant messaging status dialog using those Thunderbird versions on Windows 10.
• I never get stuck in "Disconnecting..." when manually forcing a disconnection.

I guess I'll have to wait for an outage and hope I'm using chat when it occurs.

Do you often have outages in your location?
Does it happen in safe mode, a test profile, with any AV applications disabled?

Comment 11

[diomede979]

(In reply to WaltS48 [:walts48] from comment #10)

Well, I've already wasted enough of my time on this and have documented my results.

For review:

• I didn't find a duplicate.
• I have no problem reconnecting using Thunderbird release and beta on Linux.
• I need to click the "Connect" button in the Instant messaging status dialog using those Thunderbird versions on Windows 10.
• I never get stuck in "Disconnecting..." when manually forcing a disconnection.

I guess I'll have to wait for an outage and hope I'm using chat when it occurs.

Do you often have outages in your location?
Does it happen in safe mode, a test profile, with any AV applications disabled?

there is my mistake, I get stuck on "Connecting...", and not "Disconnecting...". In the past, everytime a network outage occurs, a red countdown starts. Now I do not see that countdown and the status stucks in "Connecting...", then I do need to click on "Disconnect" and then "Connect". In this way it comes back connected. It happens with all of my configured server.

Comment 12

[WaltS48 [:walts48]]

Could not reproduce with chat.freenode.net in my testing of the 78.8.0 release candidate on Windows 10 or Linux.

Comment 13

[diomede979]

(In reply to WaltS48 [:walts48] from comment #12)

Could not reproduce with chat.freenode.net in my testing of the 78.8.0 release candidate on Windows 10 or Linux.

I have the problem on 78.7.1 release.

Comment 14

[WaltS48 [:walts48]]

Please reread comment #1, comment #3 and comment #10.

You have the problem.
I can't find any other reports.
You have a support issue not a bug IMHO.
It would be nice if you answered questions asked.

How often do you experience a network outage?

Comment 15

[Patrick Cloke [:clokep]]

It could be interesting to take a look at the protocols logs (via "Show Accounts" on the "Chat" tab, then right click on the account and click "Copy Debug Log"). It is possible the server is sending odd data we're not expecting.

Note that this log could include "sensitive" data (it will contain at least the domain you're trying to connect to).

Comment 16

[diomede979]

(In reply to Patrick Cloke [:clokep] from comment #15)

It could be interesting to take a look at the protocols logs (via "Show Accounts" on the "Chat" tab, then right click on the account and click "Copy Debug Log"). It is possible the server is sending odd data we're not expecting.

Note that this log could include "sensitive" data (it will contain at least the domain you're trying to connect to).

Connection reset.
[2/24/2021, 8:36:56 PM] DEBUG (@ prpl-irc: disconnect resource:///modules/socket.jsm:216)
Disconnect
[2/24/2021, 8:36:57 PM] DEBUG (@ prpl-irc: connect resource:///modules/socket.jsm:171)
Connecting to: apple.bnc4free.com:1339
[2/24/2021, 8:36:57 PM] DEBUG (@ prpl-irc: onTransportStatus resource:///modules/socket.jsm:560)
onTransportStatus(STATUS_RESOLVING)
[2/24/2021, 8:36:57 PM] DEBUG (@ prpl-irc: onTransportStatus resource:///modules/socket.jsm:560)
onTransportStatus(STATUS_RESOLVED)
[2/24/2021, 8:36:57 PM] DEBUG (@ prpl-irc: onStartRequest resource:///modules/socket.jsm:500)
onStartRequest
[2/24/2021, 8:36:57 PM] DEBUG (@ prpl-irc: onStopRequest resource:///modules/socket.jsm:509)
onStopRequest (2152398878)

Comment 17

[diomede979]

After any further investigation, I discovered that it is a behavior caused by a network adapter named "vEthernet (Default Switch)" component of Hyper-V on Windows 10.
Looks like that Thunderbird, when disconnecting from internet, recognizes this adapter like a working internet connection and stuck on it trying to wait connection. This network adapter can't be disabled and must be removed from Windows component.
On the following link there are instruction on how to disable it.
(https://superuser.com/questions/1282014/how-to-remove-all-the-vethernet-default-switch-once-and-for-all).

I don't know if it is still a Thunderbird bug, in the sense that Thunderbird should be able to recognize a working internet connection before trying to connect trough it. But for sure now we know what is causing the issue and how to workaround that.

Comment 18

[Magnus Melin [:mkmelin]]

Thanks for clarifying. Let's close then.

Comment 19

[diomede979]

I'm sorry, but this bug han't been resolved. I think that "Connecting..." state should have a timeout. Right now it keeps this state

Comment 20

[diomede979]

I investigated about the reason of the issue.

this is the DEBUG log when you try to connect to a server when you are offline:

---

(Thunderbird 78.12.0 (20210712120515), Gecko 78.12.0 (20210712120515) on Windows NT 10.0; Win64; x64)
[8/7/2021, 5:48:10 PM] DEBUG (@ prpl-irc: connect resource:///modules/socket.jsm:171)
Connecting to: **********
[8/7/2021, 5:48:10 PM] DEBUG (@ prpl-irc: onTransportStatus resource:///modules/socket.jsm:560)
onTransportStatus(STATUS_RESOLVING)
[8/7/2021, 5:48:10 PM] DEBUG (@ prpl-irc: onTransportStatus resource:///modules/socket.jsm:560)
onTransportStatus(STATUS_RESOLVED)
[8/7/2021, 5:48:10 PM] DEBUG (@ prpl-irc: onStartRequest resource:///modules/socket.jsm:500)
onStartRequest
[8/7/2021, 5:48:10 PM] DEBUG (@ prpl-irc: onStopRequest resource:///modules/socket.jsm:509)
onStopRequest (2152398878)

according to https://searchfox.org/comm-central/source/suite/chatzilla/js/lib/connection-xpcom.js#9

2152398878 code number which is in posted DEBUG log corresponds to "NS_ERROR_UNKNOWN_HOST = NS_ERROR_MODULE_NETWORK + 30"

The problem of the is here:

https://searchfox.org/comm-central/source/chat/modules/socket.jsm#460

is needed to manage the "NS_ERROR_UNKNOWN_HOST" state which occurs when a connection tries to connect when offline.
I suggest to start a timeout retry or disconnect it when this status event happens.

In this way we will not see che connection stuck on "connecting..." state even when the machine goes online.

thank you.

Comment 21

[diomede979]

Attachment: 1690415-add-unknown-host.patch

Comment 22

[Richard Marti (:Paenglab)]

Comment on attachment 9235847 [details] [diff] [review]
1690415-add-unknown-host.patch

I added Patrick as reviewer to let it go further.

Comment 23

[Thomas D. (:thomas8)]

diomede979 (aka :wallbroken), congratulations and thank you for offering your first patch after our mentoring sessions!
The patch looks formally correct at first glance, review will check if it does the right thing.

Comment 24

[Thomas D. (:thomas8)]

Comment on attachment 9235847 [details] [diff] [review]
1690415-add-unknown-host.patch

Review of attachment 9235847 [details] [diff] [review]:
-----------------------------------------------------------------

Just a drive-by comment in response to assignee's question on Matrix...

::: chat/modules/socket.jsm
@@ +460,5 @@
>        this.onConnectionReset();
>      } else if (aStatus == NS_ERROR_NET_TIMEOUT) {
>        this.onConnectionTimedOut();
> +    } else if (aStatus == NS_ERROR_UNKNOWN_HOST) {
> +      this.onConnectionReset();

On Matrix, :wallbroken asked:
> just a question: in my development, i added another `else if()` statement instead of adding just an `or` condition in already existing first one. line 463 could have [been] directly added in [line] 459. that's the same?

The net effect will be the same, but for the sake of reducing code redundancy, if both conditions are meant to run the same code, you should indeed combine your new condition with the existing condition. That said, I'm not sure if `NS_ERROR_UNKNOWN_HOST` should really call `this.onConnectionReset()` which looks specific to `NS_ERROR_NET_RESET` (just like `NS_ERROR_NET_TIMEOUT` has `this.onConnectionTimedOut();`). Your reviewer will tell you more.

Comment 25

[Patrick Cloke [:clokep]]

Comment on attachment 9235847 [details] [diff] [review]
1690415-add-unknown-host.patch

Review of attachment 9235847 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good overall, I think this mostly just needs a comment!

Could you update the commit entry to say "Bug 1690415 - Handle NS_ERROR_UNKNOWN_HOST error during connection by retrying. r=clokep"

Thank you!

::: chat/modules/socket.jsm
@@ +460,5 @@
>        this.onConnectionReset();
>      } else if (aStatus == NS_ERROR_NET_TIMEOUT) {
>        this.onConnectionTimedOut();
> +    } else if (aStatus == NS_ERROR_UNKNOWN_HOST) {
> +      this.onConnectionReset();

Combining it with the `NS_ERROR_NET_RESET` clause probably makes sense.

Can you add a comment saying why we do this? I think the following would be accurate: "If the host cannot be resolved, reset the connection to attempt to reconnect."  Does that sound reasonable to you?

I had to remind myself a bit about how all these codes work, a connection reset gets handled in the IRC code and will cause a message to the user ("Lost connection with server"), it will kill the connection with `ERROR_NETWORK_ERROR`, which should start the reconnect timer. So this seems like the correct thing to do. (XMPP does similar things.)

Comment 26

[Richard Marti (:Paenglab)]

Attachment: 1690415-add-unknown-host.patch

Updated the patch.

Comment 27

[Martin Giger [:freaktechnik]]

Comment on attachment 9236252 [details] [diff] [review]
1690415-add-unknown-host.patch

Review of attachment 9236252 [details] [diff] [review]:
-----------------------------------------------------------------

::: chat/modules/socket.jsm
@@ +457,5 @@
>      }
>      this.disconnected = true;
> +    // If the host cannot be resolved, reset the connection to attempt to
> +    // reconnect.
> +    if (aStatus == NS_ERROR_NET_RESET || NS_ERROR_UNKNOWN_HOST) {

This will not work as expected and always be true.

Comment 28

[Richard Marti (:Paenglab)]

Attachment: 1690415-add-unknown-host.patch

Now with correct check.

Comment 29

[Patrick Cloke [:clokep]]

Comment on attachment 9236257 [details] [diff] [review]
1690415-add-unknown-host.patch

Review of attachment 9236257 [details] [diff] [review]:
-----------------------------------------------------------------

Seems like it should work!

Comment 30

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/2675ff5a0d6b
Handle NS_ERROR_UNKNOWN_HOST error during connection by retrying. r=clokep

Comment 31

[Uplift Request]

Time for uplift requests? NI'ing the reviewer since the assignee might not be familiar with the process.

Comment 32

[Patrick Cloke [:clokep]]

Comment on attachment 9236257 [details] [diff] [review]
1690415-add-unknown-host.patch

[Approval Request Comment]
Regression caused by (bug #): N/A
User impact if declined: In some situations chat accounts might not reconnected properly.
Testing completed (on c-c, etc.): This has been on Daily for a few weeks and is in TB 93.0b1.
Risk to taking this patch (and alternatives if risky): I don't see how this can make the situation work -- it attempts to reconnect in more situations.

Comment 33

[Wayne Mery (:wsmwk)]

Comment on attachment 9236257 [details] [diff] [review]
1690415-add-unknown-host.patch

[Triage Comment]
Approved for esr91

Comment 34

[Wayne Mery (:wsmwk)]

I don't see how this can make the situation work

I assume you mean "worse"

Comment 35

[Patrick Cloke [:clokep]]

(In reply to Wayne Mery (:wsmwk) from comment #34)

I don't see how this can make the situation work

I assume you mean "worse"

Yes, sorry about that!

Comment 36

[Rob Lemley [:rjl]]

Thunderbird 91.1.1
https://hg.mozilla.org/releases/comm-esr91/rev/90901e4e57c5

Comment 37

[Wayne Mery (:wsmwk)]

diomede979, please test 91.1.1 is now available.