[Mono-dev] WCF Fail with System.Net.Sockets.SocketException: Connection reset by peer

Thu Jun 28 12:39:56 UTC 2012

I mean zero RELATED unit tests were failing (in System.ServiceModel)


On 06/28/2012 08:38 AM, Rob Wilkens wrote:
> Worth noting, with Combined patch applied (or both other applied
> separately), zero unit tests are failing.  So apparently, it doesn't
> make anything worse, at least as far as tests go.
>
> -Rob
>
> On 06/28/2012 08:20 AM, Rob Wilkens wrote:
>>
>> Either of the attached patches illustrates what i was trying to say
>> in that last patch
>>
>> (1) attempt2.patch  : This attempts to patch just the below concern,
>> and is based on a guess.
>> -or- (not both)
>> (2) combined.patch : This combines attempt2.patch with the patch i
>> sent last night since they were on the same branch.
>>
>> I'm not able to test these anything beyond "does it compile" because
>> i don't have sample code to reproduce this with.
>>
>> -Rob
>>

>> On 06/28/2012 07:55 AM, Rob Wilkens wrote:
>>> Re : The stacktrace below...
>>>
>>> This occurs when an exception is raised in ChannelDispatcher.cs on
>>> line 601.  It tries to send back an exception message to the client
>>> here, i believe.  But when it does that, it uses the existing
>>> RequestContext. 
>>>
>>> It's apparent that some data is apparently being sent, such as
>>> headers, on the RequestContext (rc) before we get to this exception.
>>>
>>> If we're dealing with the case of SocketException, which caused us
>>> to fail mid-send on the RequestContext, perhaps, again, we shouldn't
>>> handle this like every other exception and not reply.  i.e. in the
>>> exception handler here, if exception is typeof(SocketException)
>>> don't reply, what might be more interesting, if this is
>>> reproducable, would be to - as debugging - print the exception
>>> message and/or stacktrace to the screen to see what exception caused
>>> this.
>>>
>>> Did you file a bug report on this?  The discussion on this
>>> particular issue (or any particular bug) is probably better stored
>>> in the bug report comments than on the whole mailing list.  PLus
>>> comments like the above would stay in the bug report rather than get
>>> lost in the list.  IF you file a bug report, post a link to the bug
>>> report in this thread (the bug # should be enough).
>>>
>>> -Rob
>>>
>>>
>>> On 06/27/2012 01:02 PM, shahbour wrote:
>>>>
>>>>
>>>>         Hello After more testing between Mac and Windows this is
>>>>         what i got Crash Windows Mac Linux Without ErroHandler No
>>>>         Yes Yes With ErrorHandler (return false ) No Yes Yes With
>>>>         ErrorHandler (return true) No No No Before i was always
>>>>         returning false in IErrorHandler implementation because i
>>>>         only implemented for logging purpose but when i return true
>>>>         for the HandleError , the application fire the error and
>>>>         log it but never crash. Now i trying to debug the
>>>>         application under MonoDevelop and repreduce the error,
>>>>         below is what i got
>>>>
>>>>
>>>>             System.InvalidOperationException: Cannot be changed
>>>>             after headers are sent. at
>>>>             System.Net.HttpListenerResponse.set_ContentType
>>>>             (System.String value) [0x00027] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System/System.Net/HttpListenerResponse.cs:107
>>>>             at
>>>>             System.ServiceModel.Channels.Http.HttpStandaloneResponseInfo.set_ContentType
>>>>             (System.String value) [0x00000] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System.ServiceModel/System.ServiceModel.Channels.Http/HttpContextInfo.cs:274
>>>>             at
>>>>             System.ServiceModel.Channels.Http.HttpRequestContext.InternalReply
>>>>             (System.ServiceModel.Channels.Message msg, TimeSpan
>>>>             timeout) [0x00046] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9
>>>>
>>>
>>>>             /mcs/class/System.ServiceModel/System.ServiceModel.Channels.Http/HttpRequestContext.cs:140
>>>>             at
>>>>             System.ServiceModel.Channels.Http.HttpRequestContext.Reply
>>>>             (System.ServiceModel.Channels.Message msg, TimeSpan
>>>>             timeout) [0x00000] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System.ServiceModel/System.ServiceModel.Channels.Http/HttpRequestContext.cs:101
>>>>             at
>>>>             System.ServiceModel.Channels.Http.HttpRequestContext.Reply
>>>>             (System.ServiceModel.Channels.Message msg) [0x00000] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System.ServiceModel/System.ServiceModel.Channels.Http/HttpRequestContext.cs:96
>>>>             at
>>>>             System.ServiceModel.Dispatcher.ListenerLoopManager.ProcessRequest
>>>>             (IReplyChannel reply,
>>>>             System.ServiceModel.Channels.RequestContext rc)
>>>>             [0x0003b] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System.ServiceModel/System.ServiceModel.Dispatcher/ChannelDispatcher.cs:601
>>>>             at
>>>>             System.ServiceModel.Dispatcher.ListenerLoopManager.TryReceiveRequestDone
>>>>             (IAsyncResult result) [0x0001a] in
>>>>             /private/tmp/monobuild/build/BUILD/mono-2.10.9/mcs/class/System.ServiceModel/System.ServiceModel.Dispatcher/ChannelDispatcher.cs:575
>>>>
>>>> Reproducing the error is very simple, just host any application
>>>> under console app and in any service function put Thread.Sleep(..)
>>>> to give you time to close the browser before it reply. Then call
>>>> this function from any client and close it before getting the
>>>> response. In my live program i don't put Thread.sleep it is only to
>>>> give us time between calling the function and closing the browser.
>>>> Under windows we got the bellow that don't crash the application
>>>> error.Message "An operation was attempted on a nonexistent network
>>>> connection" error.InnerException {"An operation was attempted on a
>>>> nonexistent network connection"} System.Exception :q
>>>> {System.Net.HttpListenerException} error.ErrorCode 1229 BR Shahbour
>>>> ------------------------------------------------------------------------
>>>> View this message in context: Re: WCF Fail with
>>>> System.Net.Sockets.SocketException: Connection reset by peer
>>>> <http://mono.1490590.n4.nabble.com/WCF-Fail-with-System-Net-Sockets-SocketException-Connection-reset-by-peer-tp4650173p4650210.html>
>>>> Sent from the Mono - Dev mailing list archive
>>>> <http://mono.1490590.n4.nabble.com/Mono-Dev-f1517221.html> at
>>>> Nabble.com.
>>>>
>>>>
>>>> _______________________________________________
>>>> Mono-devel-list mailing list
>>>> Mono-devel-list at lists.ximian.com
>>>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>>>
>>>
>>
>>
>
>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ximian.com/pipermail/mono-devel-list/attachments/20120628/5e600766/attachment.html>