[AGENT++] uxsnmp

[Frank Fock]

Hi Richard,

OK, thank you for provinding this clarification. I was already fixing 
the issue
but was unsure if it indeed will fix the problem for sure. But now I see 
that
it will.
The updated version should be available in about two weeks.

Best regards,
Frank

Am 13.05.2015 um 16:18 schrieb Straus, Richard:
> Hi Frank,
>
> Order of our operations is as follows:
>
> 1. Call a 'Terminate' external function in the DLL from the client.
> 2. Terminate() involves:
> 	a) calling Snmp::stop_poll_thread()
> 	b) deleting the Snmp and v3MP instances
> 	c) calling Snmp::socket_cleanup()
> 3. Client unloads DLL after Terminate() completes.
>
> The problem was caused by the m_hThreadEvent signaling the completion of the worker thread when in fact it hadn't yet closed. The key is to ensure the thread exits before stop_poll_thread exits. This is guaranteed by having WaitForSingleObject wait on the thread handle itself which is signaled when the thread exits.
>
>
>
> Richard Straus
>
>
> -----Original Message-----
> From: AGENTPP [mailto:agentpp-bounces at agentpp.org] On Behalf Of Frank Fock
> Sent: May-12-15 6:09 PM
> To: agentpp at agentpp.org
> Subject: Re: [AGENT++] uxsnmp
>
> Hi Richard,
>
> Wouldn't it be better to have a callback when the thread ended and then unload the DLL.
> Your workaround might be working but I do not see a reason why that should prevent a crash for sure. A race condition may still occur, doesn't it?
>
> Best regards,
> Frank
>
> Am 07.05.2015 um 17:05 schrieb Straus, Richard:
>> Hi,
>>
>> We have encountered a crash in the Snmp::stop_poll_thread() call. This crash includes the latest version of snmp++ 3.3.5.
>>
>> We have determined the issue to be caused by WaitForSingleObject call:
>>
>> #ifdef WIN32
>>       ::WaitForSingleObject(m_hThreadEndEvent, INFINITE);
>>       CloseHandle(m_hThread);
>>
>> The SNMP calls are made in a DLL that is loaded and then unloaded. The problem is the thread keeps running for a brief period of time after setting m_hThreadEndEvent and during that time, the DLL unload sometimes completes - taking away the object code from the still-executing thread.
>>
>> We have been able to resolve the issue for WIN32 by removing all code referring to m_hThreadEndEvent. That means deleting the CreateEvent, SetEvent, ResetEvent calls pertaining to m_hThreadEndEvent as they are unnecessary. The above code in stop_poll_thread() is then replaced with this code:
>>
>> #ifdef WIN32
>>                   if (NULL != m_hThread && INVALID_HANDLE_VALUE != m_hThread)
>>                   {
>>                                   ::WaitForSingleObject(m_hThread, INFINITE);
>>                                   CloseHandle(m_hThread);
>>                                   m_hThread = NULL;
>>                   }
>>
>> Regards
>>
>> Richard Straus
>> Senior Software Developer
>> Equitrac MFP and Print Management Solutions Nuance Communications
>>
>> _______________________________________________
>> AGENTPP mailing list
>> AGENTPP at agentpp.org
>> https://oosnmp.net/mailman/listinfo/agentpp
> --
> ---
> AGENT++
> Maximilian-Kolbe-Str. 10
> 73257 Koengen, Germany
> https://agentpp.com
> Phone: +49 7024 8688230
> Fax:   +49 7024 8688231
>
> _______________________________________________
> AGENTPP mailing list
> AGENTPP at agentpp.org
> https://oosnmp.net/mailman/listinfo/agentpp
> _______________________________________________
> AGENTPP mailing list
> AGENTPP at agentpp.org
> https://oosnmp.net/mailman/listinfo/agentpp

-- 
---
AGENT++
Maximilian-Kolbe-Str. 10
73257 Koengen, Germany
https://agentpp.com
Phone: +49 7024 8688230
Fax:   +49 7024 8688231