<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    Thank you Barry, now it works for me. Bill, thanks you helped to fix
    this issue reported originally by me. I was not able to follow
    petsc-dev nor to provide the stripped code for weeks because of lack
    of time, I wanted to progress now, but I'm pleasantly surprised it
    had been resolved already.<br>
    <br>
    Thanks,<br>
    Vaclav<br>
    <br>
    <br>
    <div class="moz-cite-prefix">Dne 19.3.2014 13:59, William Perkins
      napsal(a):<br>
    </div>
    <blockquote cite="mid:532994C2.8050204@pnnl.gov" type="cite">OK.
      That probably explains some other weirdness I was seeing, but
      could not isolate.
      <br>
      <br>
      Thanks.
      <br>
      <br>
      On 03/18/2014 09:52 PM, Barry Smith wrote:
      <br>
      <blockquote type="cite">
        <br>
           Bill,
        <br>
        <br>
             Fixed in maint and soon to be pushed to next and then
        master
        <br>
        <br>
              Thanks for providing a working program that exhibits the
        problem so well. It was a rather convoluted bug. So long as you
        either avoid the -options_left option or make sure all PETSc
        objects are destroyed before PetscFinalize() it should not bit
        you until you get the fix.
        <br>
        <br>
            Barry
        <br>
        <br>
        On Mar 18, 2014, at 10:46 PM, Barry Smith
        <a class="moz-txt-link-rfc2396E" href="mailto:bsmith@mcs.anl.gov"><bsmith@mcs.anl.gov></a> wrote:
        <br>
        <br>
        <blockquote type="cite">
          <br>
             False alarm, problem is not as general as I stated.
          <br>
          <br>
             I can reproduce the problem in maint with your example code
          when -breakme is on
          <br>
          <br>
              Problem also happens in master
          <br>
          <br>
             Still working on it.
          <br>
          <br>
              Barry
          <br>
          <br>
          <br>
          <br>
          <br>
          <br>
          <br>
          On Mar 18, 2014, at 10:17 PM, Barry Smith
          <a class="moz-txt-link-rfc2396E" href="mailto:bsmith@mcs.anl.gov"><bsmith@mcs.anl.gov></a> wrote:
          <br>
          <br>
          <blockquote type="cite">
            <br>
              This looks like a problem unrelated to viewers, options, 
            etc.
            <br>
            <br>
              In maint, 3.4.3 and 3.4.4 PetscFinalize() crashes if any
            PETSc object (like your viewer) is not destroyed before
            PetscFinalize(). This does not occur in master and is not
            suppose to happen.
            <br>
            <br>
               Trying to track it down but could take a while to
            determine what change broke this.
            <br>
            <br>
              Barry
            <br>
            <br>
            <br>
            <br>
            <br>
            <br>
            On Mar 18, 2014, at 6:20 PM, William Perkins
            <a class="moz-txt-link-rfc2396E" href="mailto:william.perkins@pnnl.gov"><william.perkins@pnnl.gov></a> wrote:
            <br>
            <br>
            <blockquote type="cite">All,
              <br>
              <br>
              I also ran into this today.
              <br>
              <br>
              Attached is a minimal program that exhibited the problem
              for me. Initially, I thought it was because I was using
              non-default formats (e.g. MATLAB) for VecView and MatView,
              but I think any Viewer use causes the problem.
              <br>
              <br>
              Thanks.
              <br>
              <br>
              Bill
              <br>
              <br>
              P.S. I'm using 3.4.3 configured as follows:
              <br>
              <br>
              python ./config/configure.py \
              <br>
                PETSC_ARCH=arch-macosx-complex-opt \
              <br>
                --with-prefix="$prefix" \
              <br>
                --with-mpi=1 \
              <br>
                --with-cc=mpicc \
              <br>
                --with-fc=mpif90 \
              <br>
                --with-cxx=mpicxx \
              <br>
                --with-c++-support=1 \
              <br>
                --with-c-support=0 \
              <br>
                --with-fortran=0 \
              <br>
                --with-scalar-type=complex \
              <br>
                --with-fortran-kernels=generic \
              <br>
                --with-valgrind=0 \
              <br>
                --download-superlu_dist \
              <br>
                --download-parmetis \
              <br>
                --download-metis \
              <br>
                --download-f2cblaslapack=1 \
              <br>
                --with-clanguage=c++ \
              <br>
                --with-shared-libraries=0 \
              <br>
                --with-dynamic-loading=0 \
              <br>
                --with-x=0 \
              <br>
                --with-mpirun=mpiexec \
              <br>
                --with-mpiexec=mpiexec \
              <br>
                --with-debugging=0
              <br>
              <br>
              <br>
              On 03/12/2014 04:30 AM, Matthew Knepley wrote:
              <br>
              <blockquote type="cite">On Wed, Mar 12, 2014 at 4:52 AM,
                Václav Hapla <<a class="moz-txt-link-abbreviated" href="mailto:vaclav.hapla@vsb.cz">vaclav.hapla@vsb.cz</a>
                <br>
                <a class="moz-txt-link-rfc2396E" href="mailto:vaclav.hapla@vsb.cz"><mailto:vaclav.hapla@vsb.cz></a>> wrote:
                <br>
                <br>
                  You're right this is not happening with the petsc
                example. Thus it's
                <br>
                  probably problem in my code. I'll come back when I
                know what's the
                <br>
                  true reason.
                <br>
                <br>
                <br>
                It could still be our bug. If you can strip down your
                code and send it,
                <br>
                I will look at it.
                <br>
                <br>
                Thanks,
                <br>
                <br>
                    Matt
                <br>
                <br>
                  Matt, sorry for replying to your personal address
                only.
                <br>
                <br>
                  Thank you,
                <br>
                  Vaclav
                <br>
                <br>
                <br>
                  Dne 11.3.2014 16:04, Matthew Knepley napsal(a):
                <br>
                <blockquote type="cite">  On Tue, Mar 11, 2014 at 9:27
                  AM, Václav Hapla <<a class="moz-txt-link-abbreviated" href="mailto:vaclav.hapla@vsb.cz">vaclav.hapla@vsb.cz</a>
                  <br>
                    <a class="moz-txt-link-rfc2396E" href="mailto:vaclav.hapla@vsb.cz"><mailto:vaclav.hapla@vsb.cz></a>> wrote:
                  <br>
                  <br>
                        FWICS PetscFinalize in petsc-current contains
                  the following bug.
                  <br>
                        It calls
                  <br>
                           PetscObjectRegisterDestroyAll()
                  <br>
                        and after that, if -options_table is present, it
                  calls
                  <br>
                       
                  PetscViewerASCIIGetStdout(PETSC_COMM_WORLD,&viewer);
                  <br>
                           PetscOptionsView(viewer);
                  <br>
                        causing
                  <br>
                           [0]PETSC ERROR: No support for this operation
                  for this
                  <br>
                        object type!
                  <br>
                           [0]PETSC ERROR: Only supports ASCII viewer!
                  <br>
                        because the viewer is destroyed already by
                  <br>
                        PetscObjectRegisterDestroyAll.
                  <br>
                  <br>
                        I haven't tested petsc-dev but I can see in it's
                  the same
                  <br>
                        situation in the pinit.c source.
                  <br>
                  <br>
                  <br>
                    I cannot reproduce this on 'next':
                  <br>
                  <br>
                    next
                  $:/PETSc3/petsc/petsc-dev/src/snes/examples/tutorials$
                  ./ex19
                  <br>
                    -options_table -snes_monitor
                  <br>
                    PETSC_OPTIONS="-cublas 0" ./ex19 -options_table
                  -snes_monitor
                  <br>
                    lid velocity = 0.0625, prandtl # = 1, grashof # = 1
                  <br>
                      0 SNES Function norm 2.391552133017e-01
                  <br>
                      1 SNES Function norm 6.839858507066e-05
                  <br>
                      2 SNES Function norm 8.558777232425e-11
                  <br>
                    Number of SNES iterations = 2
                  <br>
                    #PETSc Option Table entries:
                  <br>
                    -options_table
                  <br>
                    -snes_monitor
                  <br>
                    #End of PETSc Option Table entries
                  <br>
                  <br>
                    Can you reproduce this with a PETSc example?
                  <br>
                  <br>
                      Thanks,
                  <br>
                  <br>
                         Matt
                  <br>
                  <br>
                        Best regards,
                  <br>
                        Vaclav
                  <br>
                  <br>
                  <br>
                        --
                  <br>
                  <br>
                        Vaclav Hapla
                  <br>
                        Junior researcher / PhD student
                  <br>
                        IT4Innovations <a class="moz-txt-link-rfc2396E" href="http://www.it4i.eu/"><http://www.it4i.eu/></a> /
                  Dept. of Applied
                  <br>
                        Mathematics <a class="moz-txt-link-rfc2396E" href="http://am.vsb.cz/"><http://am.vsb.cz/></a>
                  <br>
                  <br>
                        _office PI 328_
                  <br>
                        tel.: (+420) 597 329 078
                  <a class="moz-txt-link-rfc2396E" href="tel:%28%2B420%29%20597%20329%20078"><tel:%28%2B420%29%20597%20329%20078></a>
                  <br>
                        Podnikatelsky inkubator VSB-TUO
                  <br>
                        Studentska 6202/17
                  <br>
                        708 00 Ostrava-Poruba
                  <br>
                        Czech Republic
                  <br>
                  <br>
                        _contact address_
                  <br>
                        IT4Innovations
                  <br>
                        VSB - Technical University of Ostrava
                  <br>
                        17. listopadu 15/2172
                  <br>
                        708 33 Ostrava-Poruba
                  <br>
                        Czech Republic
                  <br>
                  <br>
                  <br>
                  <br>
                  <br>
                    --
                  <br>
                    What most experimenters take for granted before they
                  begin their
                  <br>
                    experiments is infinitely more interesting than any
                  results to
                  <br>
                    which their experiments lead.
                  <br>
                    -- Norbert Wiener
                  <br>
                </blockquote>
                <br>
                  --
                <br>
                <br>
                  Vaclav Hapla
                <br>
                  Junior researcher / PhD student
                <br>
                  IT4Innovations <a class="moz-txt-link-rfc2396E" href="http://www.it4i.eu/"><http://www.it4i.eu/></a> / Dept. of
                Applied Mathematics
                <br>
                  <a class="moz-txt-link-rfc2396E" href="http://am.vsb.cz/"><http://am.vsb.cz/></a>
                <br>
                <br>
                  _office PI 328_
                <br>
                  tel.: (+420) 597 329 078
                <a class="moz-txt-link-rfc2396E" href="tel:%28%2B420%29%20597%20329%20078"><tel:%28%2B420%29%20597%20329%20078></a>
                <br>
                  Podnikatelsky inkubator VSB-TUO
                <br>
                  Studentska 6202/17
                <br>
                  708 00 Ostrava-Poruba
                <br>
                  Czech Republic
                <br>
                <br>
                  _contact address_
                <br>
                  IT4Innovations
                <br>
                  VSB - Technical University of Ostrava
                <br>
                  17. listopadu 15/2172
                <br>
                  708 33 Ostrava-Poruba
                <br>
                  Czech Republic
                <br>
                <br>
                <br>
                <br>
                <br>
                --
                <br>
                What most experimenters take for granted before they
                begin their
                <br>
                experiments is infinitely more interesting than any
                results to which
                <br>
                their experiments lead.
                <br>
                -- Norbert Wiener
                <br>
              </blockquote>
              <br>
              --
              <br>
              Bill Perkins
              <br>
              Research Engineer
              <br>
              Hydrology Group
              <br>
              <br>
              Pacific Northwest National Laboratory
              <br>
              902 Battelle Boulevard
              <br>
              P.O. Box 999, MSIN K9-36
              <br>
              Richland, WA  99352 USA
              <br>
              Tel:  509-372-6131
              <br>
              Fax: 509-372-6089
              <br>
              <a class="moz-txt-link-abbreviated" href="mailto:william.perkins@pnnl.gov">william.perkins@pnnl.gov</a>
              <br>
              <a class="moz-txt-link-abbreviated" href="http://www.pnnl.gov">www.pnnl.gov</a>
              <br>
              <viewer.cpp><viewer.out>
              <br>
            </blockquote>
            <br>
          </blockquote>
          <br>
        </blockquote>
        <br>
      </blockquote>
      <br>
    </blockquote>
    <br>
    <div class="moz-signature">-- <br>
      <div id="contact-info">
        <p style="font-size:9pt; font-family: Consolas, monospace">
          Vaclav Hapla<br>
          Junior researcher / PhD student<br>
          <a href="http://www.it4i.eu/">IT4Innovations</a> /
          <a href="http://am.vsb.cz/">Dept. of Applied Mathematics</a>
        </p>
        <p style="font-size:9pt; font-family: Consolas, monospace">
          <u>office PI 328</u><br>
          tel.: (+420) 597 329 078<br>
          Podnikatelsky inkubator VSB-TUO<br>
          Studentska 6202/17<br>
          708 00 Ostrava-Poruba<br>
          Czech Republic<br>
        </p>
        <p style="font-size:9pt; font-family: Consolas, monospace">
          <u>contact address</u><br>
          IT4Innovations<br>
          VSB - Technical University of Ostrava<br>
          17. listopadu 15/2172<br>
          708 33 Ostrava-Poruba<br>
          Czech Republic<br>
        </p>
      </div>
    </div>
  </body>
</html>