MantisBT - ParaView
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0013210ParaView(No Category)public2012-05-29 16:042016-08-12 09:59
ReporterSven Buijssen 
Assigned ToChris Harris 
PriorityhighSeverityminorReproducibilityhave not tried
StatusclosedResolutionmoved 
PlatformOSOS Version
Product Version3.14.1 
Target VersionFixed in Version 
ProjectTBD
Topic Name0013210_parallel_exodus_fix
Typecrash
Summary0013210: vtkExodusIIWriter segfaults while attempting to save wavelet source
DescriptionRefactoring of Exodus writer in mid May 2012 has introduced a regression into git/master:

Sources - Wavelet - Apply
File - Save Data - File name: foo.ex2 - Files of type: Exodus II file - Ok
Ok

=> segfault in vtkExodusIIWriter.cxx:1645-1646 due to attempted array out of bounds access as 'numElts' is 0:

    if (numNodes == 0)
      {
      numPoints = blockIter->second.EntityNodeOffsets[numElts - 1]
                + blockIter->second.EntityCounts[numElts - 1];
      }
Steps To Reproduce
Additional Information
TagsNo tags attached.
Relationships
Attached Files
Issue History
Date ModifiedUsernameFieldChange
2012-05-29 16:04Sven BuijssenNew Issue
2012-05-30 09:35Chris HarrisAssigned To => Chris Harris
2012-05-30 09:51Chris HarrisNote Added: 0028658
2012-05-31 08:35Chris HarrisTopic Name => 0013210_parallel_exodus_fix
2012-05-31 08:35Chris HarrisStatusbacklog => gatekeeper review
2012-06-01 14:40Utkarsh AyachitFixed in Version => git-master
2012-06-01 14:41Utkarsh AyachitStatusgatekeeper review => customer review
2012-06-01 14:41Utkarsh AyachitNote Added: 0028661
2012-07-19 16:41Alan ScottNote Added: 0028875
2012-07-19 16:41Alan ScottStatuscustomer review => todo
2012-07-19 16:42Alan ScottNote Added: 0028876
2012-07-19 16:42Alan ScottPrioritynormal => high
2012-07-24 16:19Chris HarrisNote Added: 0028910
2012-07-24 17:29Sven BuijssenNote Added: 0028916
2012-07-25 10:32Chris HarrisNote Added: 0028921
2012-07-25 11:56Chris HarrisNote Added: 0028923
2012-07-25 13:37Chris HarrisNote Added: 0028925
2012-10-29 17:04Utkarsh AyachitFixed in Versiongit-master => 3.98.0
2016-08-12 09:59Kitware RobotNote Added: 0038220
2016-08-12 09:59Kitware RobotStatustodo => closed
2016-08-12 09:59Kitware RobotFixed in Version3.98.0 =>
2016-08-12 09:59Kitware RobotResolutionopen => moved

Notes
(0028658)
Chris Harris   
2012-05-30 09:51   
A fix for this has been pushed to VTK:

http://review.source.kitware.com/#/t/747/ [^]
(0028661)
Utkarsh Ayachit   
2012-06-01 14:41   
Merged in master, if applicable.
(0028875)
Alan Scott   
2012-07-19 16:41   
Chris,
It dies remote server.
Linux, remote server, Master. I just followed your excellent directions below.
(0028876)
Alan Scott   
2012-07-19 16:42   
Making it high priority, since if it doesn't work remote server, it doesn't work. Further, this is the main way we write out datasets.
(0028910)
Chris Harris   
2012-07-24 16:19   
I am unable to recreate this problem. I can connect to a remote server and write out a file without a crash. Is the remote server running master?
(0028916)
Sven Buijssen   
2012-07-24 17:29   
Using a debug build of the git master branch on Linux x86_64, the segfault is gone when having connected to a single, separately started pvserver (both 'mpirun -np 1 /path/to/pvserver' and 'pvserver' work).

But having connected to a dual server, pvserver crashes while trying to save with:

% mpirun -np 2 /scratch/pv-git-build/bin/pvserver
Waiting for client...
Connection URL: cs://host:11111
Accepting connection(s): host:11111
Client connected.
Warning: In /scratch/pv.git/VTK/IO/Exodus/vtkExodusIIWriter.cxx, line 780
vtkPExodusIIWriter (0x1c9ed90): Attempting to proceed without metadata

Warning: In /scratch/pv.git/VTK/IO/Exodus/vtkExodusIIWriter.cxx, line 780
vtkPExodusIIWriter (0x18990d0): Attempting to proceed without metadata

[host:13327] *** Process received signal ***
[host:13327] Signal: Segmentation fault (11)
[host:13327] Signal code: Address not mapped (1)
[host:13327] Failing at address: 0xfffffffffffffffc
[host:13326] *** Process received signal ***
[host:13326] Signal: Segmentation fault (11)
[host:13326] Signal code: Address not mapped (1)
[host:13326] Failing at address: 0xfffffffffffffffc
[host:13327] [ 0] /lib/libpthread.so.0(+0xf8f0) [0x7f086c6338f0]
[host:13327] [ 1] /scratch/pv-git-build/lib/libvtkIOExodus-6.0.so.1(_ZN17vtkExodusIIWriter21WriteBlockInformationEv+0x19e) [0x7f087fed75ba]
[host:13327] [ 2] /scratch/pv-git-build/lib/libvtkIOExodus-6.0.so.1(_ZN17vtkExodusIIWriter9WriteDataEv+0xdce) [0x7f087fed0ce0]
[host:13327] [ 3] /scratch/pv-git-build/lib/libvtkIOExodus-6.0.so.1(_ZN17vtkExodusIIWriter11RequestDataEP14vtkInformationPP20vtkInformationVectorS3_+0xc5) [0x7f087fecfe7d]
[host:13327] [ 4] /scratch/pv-git-build/lib/libvtkIOExodus-6.0.so.1(_ZN17vtkExodusIIWriter14ProcessRequestEP14vtkInformationPP20vtkInformationVectorS3_+0x225) [0x7f087fecfc9d]
[host:13327] [ 5] /scratch/pv-git-build/lib/libvtkCommonExecutionModel-6.0.so.1(_ZN12vtkExecutive13CallAlgorithmEP14vtkInformationiPP20vtkInformationVectorS3_+0x8d) [0x7f086860b30b]
[...]

When using Multi-Core via AutoMPI instead (Edit - Settings - General - Use Multi-Core), the server crashes while saving with a different error message:
% /scratch/pv-git-build/bin/paraview -dr
Process started
AutoMPI: SUCCESS: command is:
 "/scratch/openmpi/gcc4.4.3/1.4.3/bin/mpiexec" "-np" "2" "/scratch/pv-git-build/bin/pvserver" "--server-port=50577"
AutoMPI: starting process server
-------------- server output --------------
Waiting for client...
AutoMPI: server sucessfully started.
ERROR: In /scratch/pv.git/VTK/Parallel/Core/vtkSocketCommunicator.cxx, line 812
vtkSocketCommunicator (0x32db740): Could not receive tag. 188969
(0028921)
Chris Harris   
2012-07-25 10:32   
Ok, so this is a problem with the parallel implementation. I can recreate this using the instruction above.
(0028923)
Chris Harris   
2012-07-25 11:56   
This appear not to be a bug introduced by the refactoring for VTK modular. I ran the same test with 3.14 and see the same behavior.
(0028925)
Chris Harris   
2012-07-25 13:37   
Also fails with 3.12
(0038220)
Kitware Robot   
2016-08-12 09:59   
Resolving issue as `moved`.

This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.