[Paraview-developers] [EXTERNAL] ParaView Windows Socket Issue

[Utkarsh Ayachit]

Yes, Windows only with MSVC compiler. Server was running on the same
machine as well. The test takes care of starting the server and client and
then running the actions that seem to cause the failures -- anything with
large data movements is what I am tending to believe, but not sure how true
that is.


On Mon, May 12, 2014 at 7:55 PM, Scott, W Alan <wascott at sandia.gov> wrote:

>  Windows, correct?  Visual compiler?  Where are you running the server?
>
>
>
>
>
>
>
> *From:* Utkarsh Ayachit [mailto:utkarsh.ayachit at kitware.com]
> *Sent:* Monday, May 12, 2014 5:54 PM
> *To:* Scott, W Alan
> *Cc:* ParaView Developers
> *Subject:* Re: [EXTERNAL] [Paraview-developers] ParaView Windows Socket
> Issue
>
>
>
> It is true that it seems like it;s failing only on this machine. If you
> want to test it, just check out git/next and build it as Release or
> RelWithDebInfo and then run the following test:
>
>
>
> ctest -R pvcs.RenderNan
>
>
>
> It hangs randomly for me.
>
>
>
> Thanks
>
> Utkarsh
>
>
>
> On Mon, May 12, 2014 at 5:31 PM, Scott, W Alan <wascott at sandia.gov> wrote:
>
> Timing?  Could you have hardware that is becoming flaky?  Try on another
> box?
>
>
>
> If you want to send me bits, I could give it a try?
>
>
>
> Alan
>
>
>
> *From:* Paraview-developers [mailto:
> paraview-developers-bounces at paraview.org] *On Behalf Of *Utkarsh Ayachit
> *Sent:* Monday, May 12, 2014 3:11 PM
> *To:* ParaView Developers
> *Subject:* [EXTERNAL] [Paraview-developers] ParaView Windows Socket Issue
>
>
>
> I am running into a strange issue with sockets on Windows. Here are the
> symptoms:
>
>
>
> + VS9 (VS2008), 64 bit, Windows 7
>
> + Fails only with Release or RelWithDebInfo builds (never in Debug)
>
> + client-server tests fail when receiving large chunks of data (although
> sometimes it happens for small chunks too)
>
> + "recv" call fails with error WSAEFAULT (most of the time). I am assured
> that the corresponding "send" does indeed happen. In other words, not a
> race condition or some other deadlock.
>
> + I can overcome the problem by simply doing a memset(buffer, 0,
> lengthofbuffer) before the "recv" call.
>
>
>
> I already tried chunking the recv is 8KB chunks, doesn't help. Has anyone
> ever seen something like this, or has any ideas that could work?
>
>
>
> Thanks
>
> Utkarsh
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview-developers/attachments/20140512/4daf6b40/attachment.html>