[Paraview] Client-Server Fails on Altix

[Utkarsh Ayachit]

Kent,

Unfortunately there's no easy way around the requirement that the 
client-server versions must match. The client assumes what the server 
side provides in terms of classes, their API etc. These keep on 
changing. For example, the composite dataset stuff changed a great deal 
between 3.2 and the current CVS. Consequently, if a 3.2 client connected 
to a CVS server and started creating old composite data filters, the 
server can do nothing but abort().

Utkarsh


Kent Eschenberg wrote:
> Fixed, mostly. When client and server are built from the exact same 
> source (from CVS) it works when the client uses 32-bit IDs and when the 
> server uses either 32-bit or 64-bit IDs.
> 
> Unfortunately, it seems that our users must now switch their clients to 
> the same CVS version as the server on the Altix. For example, the 3.2.1 
> client fails with this message:
> 
> ===
> Handshake failed. You are probably using mismatching versions of client 
> and server.
> ===
> 
> Could someone involved with this part of ParaView comment, in general, 
> on using a client with a server of a different version?
> 
> Kent
> Pittsburgh Supercomputing Center
> 
> Berk Geveci wrote:
>> To be on the safe side, I would try:
>>
>> * Client/server built from the exact same source
>> * Both set to the same VTK_USE_64BIT_ID (I recommend off)
>>
>> -berk
>>
>>
>> On Tue, Jun 24, 2008 at 3:35 PM, Kent Eschenberg <eschenbe at psc.edu> 
>> wrote:
>>> The client is 32-bit and the server is 64-bit, with VTK_USE_64BIT_ID 
>>> off and
>>> on, respectively.
>>>
>>> However ... I've had no problem using this same client system and 
>>> pvserver
>>> on another 64-bit system with VTK_USE_64BIT_ID on.
>>>
>>> The client and server that work together run CentOS. The other server 
>>> with
>>> the new problem runs SUSE. All use PV CVS but downloaded at slightly
>>> different times.
>>>
>>> Suggestions welcome!
>>> Kent
>>> Pittsburgh Supercomputing Center
>>>
>>> Berk Geveci wrote:
>>>> Are you running the client on the same machine? By any chance, do the
>>>> client and server have different VTK_USE_64BIT_IDS set?
>>>>
>>>> -berk
>>>>
>>>> On Mon, Jun 23, 2008 at 11:29 AM, Kent Eschenberg <eschenbe at psc.edu>
>>>> wrote:
>>>>> The interactive version of PV works fine but when using pvserver (1 
>>>>> or 4
>>>>> processes) all attempts to add to the display (e.g., a source of 
>>>>> box or
>>>>> sphere) only yields this message:
>>>>>
>>>>> ERROR: In .../ParaView3/Servers/Common/vtkPVDataInformation.cxx, 
>>>>> line 971
>>>>> vtkPVDataInformation (0x9e93040): Error parsing data set type.
>>>>>
>>>>> I also get the now-expected warnings about being unable to connect 
>>>>> to an
>>>>> X
>>>>> server.
>>>>>
>>>>> System: SGI Altix 4700, Itanium2 ia64
>>>>> OS: enhanced version of SuSE
>>>>> Compiler: g++ 4.1.2
>>>>> Cmake: 2.6.0
>>>>> ParaView: from CVS around June 13
>>>>> Qt: 4.3.5
>>>>> Mesa: 7.0.1
>>>>>
>>>>> Any ideas?
> _______________________________________________
> ParaView mailing list
> ParaView at paraview.org
> http://www.paraview.org/mailman/listinfo/paraview
>