<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
    <title></title>
  </head>
  <body bgcolor="#ffffff" text="#000000">
    Hmm,&nbsp; I had tested it on XP with PV 3.8.1 and didn't have any
    problems. sorry about that, I'll have to try again.<br>
    <br>
    Burlen<br>
    <br>
    On 04/14/2011 07:54 PM, David Partyka wrote:
    <blockquote
      cite="mid:BANLkTinia9=m1Jz7-H1Z8q7gCyw3hoztXg@mail.gmail.com"
      type="cite">Hi Burlen, I had to revert your patch as it doesn't
      compile on Windows..
      <div><br>
      </div>
      <div>You will have to make sure it compiles there as well and
        resubmit your patch. If you need any help please let me know.
        Thanks.<br>
        <br>
        <div class="gmail_quote">On Thu, Apr 14, 2011 at 3:51 PM, David
          Partyka <span dir="ltr">&lt;<a moz-do-not-send="true"
              href="mailto:david.partyka@kitware.com">david.partyka@kitware.com</a>&gt;</span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt
            0.8ex; border-left: 1px solid rgb(204, 204, 204);
            padding-left: 1ex;">
            Thanks Burlen, This is applied for 3.10.2.
            <div>
              <div class="h5"><br>
                <br>
                <div class="gmail_quote">On Thu, Apr 14, 2011 at 2:11
                  PM, Burlen Loring <span dir="ltr">&lt;<a
                      moz-do-not-send="true"
                      href="mailto:bloring@lbl.gov" target="_blank">bloring@lbl.gov</a>&gt;</span>
                  wrote:<br>
                  <blockquote class="gmail_quote" style="margin: 0pt 0pt
                    0pt 0.8ex; border-left: 1px solid rgb(204, 204,
                    204); padding-left: 1ex;">
                    <div bgcolor="#ffffff" text="#000000"> Thanks Dave,<br>
                      <br>
                      Filed the bug report <a moz-do-not-send="true"
                        href="http://www.paraview.org/Bug/view.php?id=12087"
                        target="_blank">http://www.paraview.org/Bug/view.php?id=12087</a><br>
                      <br>
                      I updated the patch for 3.10.0 as well (attached
                      here and on the bug report).<br>
                      <font color="#888888"> <br>
                        Burlen</font>
                      <div>
                        <div><br>
                          <br>
                          On 04/13/2011 11:24 AM, David Partyka wrote:
                          <blockquote type="cite">Humm, I forgot all
                            about this email. I'll stick it in right now
                            for 3.10.2. If you don't mind please file a
                            bug so that it isn't forgotten.<br>
                            <br>
                            <div class="gmail_quote">On Wed, Apr 13,
                              2011 at 2:17 PM, Burlen Loring <span
                                dir="ltr">&lt;<a moz-do-not-send="true"
                                  href="mailto:bloring@lbl.gov"
                                  target="_blank">bloring@lbl.gov</a>&gt;</span>
                              wrote:<br>
                              <blockquote class="gmail_quote"
                                style="margin: 0pt 0pt 0pt 0.8ex;
                                border-left: 1px solid rgb(204, 204,
                                204); padding-left: 1ex;">
                                <div bgcolor="#ffffff" text="#000000">
                                  Hi Dave,<br>
                                  <br>
                                  What is the status on this?<br>
                                  <font color="#888888"> <br>
                                    Burlen</font>
                                  <div>
                                    <div><br>
                                      <br>
                                      On 02/27/2011 02:53 PM, David
                                      Partyka wrote:
                                      <blockquote type="cite">Thanks
                                        Burlen, We'll take a look.<br>
                                        <br>
                                        <div class="gmail_quote">On Sun,
                                          Feb 27, 2011 at 5:18 PM,
                                          Burlen Loring <span dir="ltr">&lt;<a
                                              moz-do-not-send="true"
                                              href="mailto:bloring@lbl.gov"
                                              target="_blank">bloring@lbl.gov</a>&gt;</span>
                                          wrote:<br>
                                          <blockquote
                                            class="gmail_quote"
                                            style="margin: 0pt 0pt 0pt
                                            0.8ex; border-left: 1px
                                            solid rgb(204, 204, 204);
                                            padding-left: 1ex;"> Hi,<br>
                                            <br>
                                            While installing ParaView on
                                            Nautilus, <a
                                              moz-do-not-send="true"
                                              href="http://www.nics.tennessee.edu/computing-resources/nautilus"
                                              target="_blank">http://www.nics.tennessee.edu/computing-resources/nautilus</a>,
                                            I hit a bug in vtkSocket
                                            that prevents ParaView from
                                            running on this machine.
                                            While tracking this down I
                                            uncovered a couple related
                                            issues.<br>
                                            <br>
                                            The main issue is that
                                            vtkSocket does not handle
                                            EINTR. EINTR occurs when a
                                            signal is caught by the
                                            application during a
                                            blocking socket call. While
                                            ParaView does not make use
                                            of signals they are used for
                                            asynchronous communication
                                            by some SGI specific
                                            libraries on Nautilus that
                                            are linked in with SGI MPI.
                                            Because Rank 0 pvserver
                                            spends quite a bit of its
                                            time blocked in socket calls
                                            it only takes a few 10s of
                                            seconds for EINTR to occur.
                                            When faced with EINTR
                                            ParaView silently exits
                                            leaving the user wondering
                                            what the heck happened.
                                            Which brings me to the
                                            second issue, a lack of
                                            error reporting in
                                            vtkSocket.<br>
                                            <br>
                                            To solve the first issue
                                            vtkSocket has to handle
                                            EINTR. How EINTR should be
                                            handled depends on the
                                            specific socket call. For
                                            all calls except connect the
                                            call can simply be
                                            restarted. For EINTR during
                                            connect one can't restart
                                            the call on all unix, so
                                            instead one must block in a
                                            select call when connect
                                            fails with EINTR. To be
                                            portable across Unix one
                                            should handle EINTR in all
                                            socket calls, even simple
                                            ones like set/getsockopt.<br>
                                            <br>
                                            The second issue of error
                                            reporting applies to all
                                            socket related errors in
                                            general, my feeling is that
                                            when a socket call fails
                                            vtkSocket should print a
                                            message using vtkErrorMacro,
                                            errno, and strerror(or
                                            windows equivalent) at the
                                            point of failure. I think
                                            this should be done inside
                                            vtkSocket because this is
                                            the only place one can
                                            safely assume errno has
                                            relevant information and
                                            vtkSocket has been
                                            implemented returning a
                                            single error code, -1, so
                                            that returning the real
                                            error code would change the
                                            API and break existing code,
                                            including ParaView. Not to
                                            mention that the values for
                                            error codes are apparently
                                            different on windows and
                                            unix.<br>
                                            <br>
                                            I took a stab at fixing
                                            these issues, patches
                                            attached. I tested them on
                                            my workstation, nautilus,
                                            and laptop running xp. I ran
                                            a dashboard on my linux
                                            workstation and didn't see
                                            any related issues. Would
                                            someone at KW mind taking a
                                            look at the changes and see
                                            if it could be made
                                            permanent?<br>
                                            <br>
                                            By the way after testing all
                                            socket calls for error
                                            returns I uncovered a third
                                            bug, vtkSocket::Close didn't
                                            set the descriptor ivar to
                                            -1 which resulted in
                                            vtkSocket::~vtkSocket
                                            calling close on a closed
                                            socket. Not a disasterous
                                            error, but this reinforces
                                            my opinion that the returns
                                            should be tested and error
                                            messages printed.<br>
                                            <br>
                                            Thanks<br>
                                            <font color="#888888">
                                              Burlen<br>
                                              <br>
                                              <br>
                                              <br>
                                              <br>
                                              <br>
                                              <br>
                                            </font><br>
_______________________________________________<br>
                                            Powered by <a
                                              moz-do-not-send="true"
                                              href="http://www.kitware.com"
                                              target="_blank">www.kitware.com</a><br>
                                            <br>
                                            Visit other Kitware
                                            open-source projects at <a
                                              moz-do-not-send="true"
                                              href="http://www.kitware.com/opensource/opensource.html"
                                              target="_blank">http://www.kitware.com/opensource/opensource.html</a><br>
                                            <br>
                                            Please keep messages
                                            on-topic and check the
                                            ParaView Wiki at: <a
                                              moz-do-not-send="true"
                                              href="http://paraview.org/Wiki/ParaView"
                                              target="_blank">http://paraview.org/Wiki/ParaView</a><br>
                                            <br>
                                            Follow this link to
                                            subscribe/unsubscribe:<br>
                                            <a moz-do-not-send="true"
                                              href="http://www.paraview.org/mailman/listinfo/paraview"
                                              target="_blank">http://www.paraview.org/mailman/listinfo/paraview</a><br>
                                            <br>
                                          </blockquote>
                                        </div>
                                        <br>
                                      </blockquote>
                                      <br>
                                    </div>
                                  </div>
                                </div>
                              </blockquote>
                            </div>
                            <br>
                          </blockquote>
                          <br>
                        </div>
                      </div>
                    </div>
                  </blockquote>
                </div>
                <br>
              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
    </blockquote>
    <br>
  </body>
</html>