Found the problem: Not the new MercuryD.dll causes the crash but polling the hotmail account itself. Polling the hotmail with the original dll causes Mercury to crash as well. Maybe you can pass that to David. If he needs anything from me to debug/reproduce this, I'm more than happy to help...

Konrad

Hi All,

I want to add a hotmail account to the MercuryD module but it is not working (network error). I read a few posts from 2009 about this and the problem seems to be that the SSL certificate from Microsoft is not valid. But these "old" posts said something about "being fixed with v4.72 or the4 next release". I do have the latest v4.74 and it still does not work. It seems like Pegasus can handle it. My settings are pop3.live.com, SSL with port 995.

Any ideas about this?

Thx

Konrad

I found a thread on a Thunderbird support forum where folks discovered that POP3 authentication to Hotmail failed if the password was too long.  The reported length limit varied between 10 characters and 14 characters.  Don't know whether this is a Hotmail issue or TB issue but might be worth a try if you use a long password.

Have you done any session logging to see what the logs show?

Hi!

Password has 7 characters, so should not be too long ;-)

Session log for "SSL with direct connection":

12:24:25.454: --- Wed May 30 12:24:25 2012 ---
12:24:25.454: Connect to 'pop3.live.com:995', timeout 30.
12:24:25.063: 22: Error -32 activating SSL session (locus 6014, type 4, code 0, 'Server provided a broken/invalid certificate, try again with')
12:24:25.063: --- Connection closed normally at Wed May 30 12:24:25 2012. ---
12:24:25.063: 

Session log for "STLS command":

12:23:01.266: --- Wed May 30 12:23:01 2012 ---
12:23:01.266: Connect to 'pop3.live.com:995', timeout 30.
12:23:31.438: 8: NB Socket read timeout
12:23:31.438: --- Connection closed normally at Wed May 30 12:23:31 2012. ---
12:23:31.438: 

Not sure which setting for the SSL connection should be the used but according to the session log, the direct command seems to be the one...

btw: Thunderbird has no problem to connect to this hotmail account using SSL 995.

Any ideas?

Konrad

According to the Pegasus Mail help file STLS is the preferred way. 

More thoughts...

Increase the timeout setting to at least 180, may need to go as high as 300.   There is a timeout setting both in the SMTP host configuration and in the General tab of Internet Options.  I don't know which takes precedence so would change both.

Are you using the built-in NIC on an nVidia based motherboard?   If so, there is a known issue with socket read timeout errors caused by an advanced optimization setting called Checksom Offload.  Disabling this setting has resolved SMTP issues in Pegasus Mail.

Consider doing a search on this forum for "socket read timeout" to see what other ideas are out there.

> According to the Pegasus Mail help file STLS is the preferred way.

Yes, but since this is MSoft$ and they do not use the standards properly you need to use direct SSL for POP3.  This is for Pegasus Mail and it does work, if the certificate validation is turned on you'll get an error similar to that provided by the MercuryD session log.  You can use STARTTLS when sending through Hotmail.  I use port 587 and disable certificate validation.

(a) -POP3-

  Server host name: pop3.live.com
  User name: <your_user_name>@hotmail.com
  Password: <your_password>
  Server TCP/IP port: 995
  SSL/TLS: via direct SSL connect
  Enable server certificate fingerprint tracking: unchecked
  Disable certificate validation: checked

(b) - SMTP via SSL -

  Server host name: smtp.live.com
  Server TCP/IP port: 465
  SSL/TLS: via direct ssl connection
  Enable server certificate fingerprint tracking: unchecked
  Disable certificate validation: checked
  SMTP Authentication: Login to the SMTP server using POP3
  username/password (the Hotmail-POP3-definition has been chosen)

       Session log for "SSL with direct connection":

       12:24:25.454: --- Wed May 30 12:24:25 2012 ---
       12:24:25.454: Connect to 'pop3.live.com:995', timeout 30.
       12:24:25.063:

22: Error -32 activating SSL session (locus 6014, type 4, code 0,

'Server provided a broken/invalid certificate, try again with')
       12:24:25.063: --- Connection closed normally at Wed May 30 12:24:25 2012. ---
       12:24:25.063: 

 Passed to David Harris and beta team to determine if there is any way to turn off certificate validation in MercuryD.

      12:24:25.454: --- Wed May 30 12:24:25 2012 ---
      12:24:25.454: Connect to 'pop3.live.com:995', timeout 30.
      12:24:25.063:

22: Error -32 activating SSL session (locus 6014, type 4, code 0,

'Server provided a broken/invalid certificate, try again with')
      12:24:25.063: --- Connection closed normally at Wed May 30 12:24:25 2012. ---
      12:24:25.063: 

 Konrad,

David Harris ginned up a new MercuryD that turns off the certificate verification when using Direct SSL.  You can get a copy at http://tstephenson.com/mercury/mercuryd.zip I've checked it out with my Hotmail account and it seems to work.  Let me know how this works for you.

[quote user="Thomas R. Stephenson"]

David Harris ginned up a new MercuryD that turns off the certificate verification when using Direct SSL.  You can get a copy at http://tstephenson.com/mercury/mercuryd.zip I've checked it out with my Hotmail account and it seems to work.  Let me know how this works for you.

[/quote]

Is this version based on the official 4.74 and can I use it for production?

As soon as I have tested it, I will get back to you!

Thx so far

Konrad

Thomas, the new dll works like a charm ;-)

Thx

> Is this version based on the official 4.74 and can I use it for production?

Yes and I am using it in production.  It is not released as yet though since David wanted to see if it solved you Hotmail problem first.

[quote user="Thomas R. Stephenson"]

Yes and I am using it in production.  It is not released as yet though since David wanted to see if it solved you Hotmail problem first.

[/quote]

As said, works great ;-)

Thx to you and David for your quick support!

Konrad

[quote user="Thomas R. Stephenson"]

> Is this version based on the official 4.74 and can I use it for production?

Yes and I am using it in production.  It is not released as yet though since David wanted to see if it solved you Hotmail problem first.

[/quote]

In the last few days I had lots of mercury crashes:

12-06-01.1835: Restarting Mercury after apparent abnormal termination
12-06-01.1927: Normal operation restored - resetting counters.
12-06-01.1940: Restarting Mercury after apparent abnormal termination
12-06-01.2027: Normal operation restored - resetting counters.
12-06-01.2347: Restarting Mercury after apparent abnormal termination
12-06-02.0027: Normal operation restored - resetting counters.
12-06-02.0715: Restarting Mercury after apparent abnormal termination
12-06-02.0728: Normal operation restored - resetting counters.
12-06-02.1015: Restarting Mercury after apparent abnormal termination
12-06-02.1028: Normal operation restored - resetting counters.
12-06-03.2250: Restarting Mercury after apparent abnormal termination
12-06-03.2329: Normal operation restored - resetting counters.
12-06-04.2252: Restarting Mercury after apparent abnormal termination
12-06-04.2330: Normal operation restored - resetting counters.
12-06-05.0641: Restarting Mercury after apparent abnormal termination
12-06-05.0730: Normal operation restored - resetting counters.
12-06-06.0054: Restarting Mercury after apparent abnormal termination
12-06-06.0130: Normal operation restored - resetting counters.
12-06-06.0628: Restarting Mercury after apparent abnormal termination
12-06-06.0630: Normal operation restored - resetting counters.

Not sure if this has anything to do with the new dll but the timing would fit. How can I debug this? Is there a stack trace written when mercury crashes or any other log file which will help here?

Konrad

> Not sure if this has anything to do with the new dll but the timing
> would fit. How can I debug this? Is there a stack trace written when
> mercury crashes or any other log file which will help here?

I doubt if this is related to MercuryD at all unless MercuryD is downloading a corrupted file.  This could also happen when your anti-virus software finds what it thinks is a file with a virus and grabs it away from Mercury,

Mercury does write files in the queue when there is a crash to a separate directory.

FWIW, I have 5 systems using the latest version of MercuryD since May 31 and I am getting no crashes at all.

I restored the original MercuryD.dll and so far no crash for more than 24 hours. With the other dll it crashed several times a day... Need to do more tests!

No files or directory in the queue folder. It is empty...

No AV software on the system except ClamAV. But I never had this behavior and I doubt there are many messages coming in through pop3 at all. I only use it for "old" accounts. Maybe one or two mails a day...

There should be stack trace somewhere when mercury crashes. Are you sure that they are stored in the queue directory?! This does not make much sense to me since the queue folder is for mails to be processed, right?

Konrad

btw: The "pop3 client configuration" window is a bit buggy and causes mercury to crash from time to time.

bug #1): The first time I open it after some time or a restart, the scrollbar is missing. This is almost a 100% reproduce able. I have lots of accounts in there in they do not fit in the list box but still there is no scrollbar. If I re-open the window, the scrollbar shows up.

bug #2): Opening this window causes mercury to crash from time to time. This is not a 100% reproduceable but it happens maybe every day or two.

Anyhow: I would like to focus on the crashed which happen without any interaction. How can I debug them?

Thx

Konrad

One more info:

I checked the loader.log to see when mercury was restarted due to a crash (which happens really, really often) and I compared it with the mercury core log to find out what mercury did at the time when it crashed. No entries in the core log at all for the each crash time! This means to me, that mercury does not crash because of processing a mail. It is crashing while doing "nothing"...

How can I track this?

Konrad

> btw: The "pop3 client configuration" window is a bit buggy and causes mercury to crash from time to time.

Strange, I have never seen that.  It is rock solid for me.  I now have a test setup with more than a screen full of entries running to see if the number of entries makes a difference.

> bug #1): The first time I open it after some time or a restart, the
> scrollbar is missing. This is almost a 100% reproduce able. I have
> lots of accounts in there in they do not fit in the list box but still
> there is no scrollbar. If I re-open the window, the scrollbar shows
> up.

Known windows problem,  all you have to do is select one of the entries and then use page up/page down to see all of the entries.  FWIW, how many accounts do you have?

> bug #2): Opening this window causes mercury to crash from time to
> time. This is not a 100% reproduceable but it happens maybe every day
> or two.

Again, I see no MercuryD crashes.  This starts to make me wonder about your MercuryD.dat file.  If this file is not available, or corrupted, when you try to open MercuryD I suspect it might cause a crash.  That is just a guess though since I cannot duplicate your problem.

Here is my MercuryD setup, I just turned on session logging for testing.

[MercuryD]
Scratch : C:\MERCBAYA\Scratch\MercuryD
Timeout : 300
Poll : 300
Session_logging : C:\MERCBAYA\Sessions\MercuryD
Session_logmode : 1

>  
> Anyhow: I would like to focus on the crashed which happen without any interaction. How can I debug them?

Not sure.  I'm leaving my setup running 24/7 to see if I can duplicate your problem.  

Hi!

I deactivated the new mercuryd.dll and my installation did not crash for the last 2 days. After that, I activated the dll again and mercury crashed after a few hours...

I do have about 40 pop3 accounts configured but only 5 or 6 are enabled at the moment. And I do have one hotmail account in there.

Can you contact David and ask him how we can debug this? Mercury should write a stack trace for each crash. This is how crashes are usually handled and debugged!

Thx

Konrad