MantisBT - ParaView
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0006140ParaViewBugpublic2007-12-10 14:402009-09-09 21:55
ReporterKen Moreland 
Assigned ToUtkarsh Ayachit 
PriorityhighSeverityminorReproducibilityalways
StatusclosedResolutionunable to reproduce 
PlatformOSOS Version
Product Version 
Target Version3.6.2Fixed in Version 
Project
Topic Name
Type
Summary0006140: Extents wrong in .pvd file with .pvti part
DescriptionI have attached a .pvd file that points to a .pvti part which can cause a crash on a parallel server. To see this, connect to a parallel server with 4 processes. Load the attached .pvd file. Now add a contour filter and Apply the default values. The server crashes after emitting the following errors:

Process id: 3 >> ERROR: In /projects/kmorel/dashboard/ParaView3Experimental/VTK/Filtering/vtkImageData.cxx, line 2082
vtkImageData (0xbb7ed0): GetPointer: Pixel (0, 33, 33) not in current extent: (50, 67, 0, 33, 0, 33)

Process id: 1 >> ERROR: In /projects/kmorel/dashboard/ParaView3Experimental/VTK/Filtering/vtkImageData.cxx, line 2082
vtkImageData (0xbb7b00): GetPointer: Pixel (0, 33, 0) not in current extent: (0, 33, 0, 33, 16, 33)
Steps To Reproduce
Additional InformationI think this problem is with extents. If you load the pvti file directly, you don't get the crash but the contour is made on only 1/4 of the data.
TagsNo tags attached.
Relationships
Attached Filestar ironProt_parallel.tar (337,920) 2007-12-10 14:40
https://www.vtk.org/Bug/file/6246/ironProt_parallel.tar
? test8.pvti (383) 2008-06-27 11:17
https://www.vtk.org/Bug/file/6565/test8.pvti
? test8a.vti (444) 2008-06-27 11:17
https://www.vtk.org/Bug/file/6566/test8a.vti
? test8b.vti (444) 2008-06-27 11:17
https://www.vtk.org/Bug/file/6567/test8b.vti
png pvti_test2.png (33,122) 2008-06-27 11:18
https://www.vtk.org/Bug/file/6568/pvti_test2.png
png

png pvti_test3.png (33,815) 2008-06-27 11:18
https://www.vtk.org/Bug/file/6569/pvti_test3.png
png
Issue History
Date ModifiedUsernameFieldChange
2007-12-10 14:40Ken MorelandNew Issue
2007-12-10 14:40Ken MorelandFile Added: ironProt_parallel.tar
2007-12-10 18:40Ken MorelandNote Added: 0009853
2008-01-24 19:00Ken MorelandNote Added: 0010296
2008-01-24 19:00Ken MorelandPrioritynormal => high
2008-04-08 15:39Ken MorelandStatusbacklog => tabled
2008-04-08 15:39Ken MorelandAssigned To => Ken Moreland
2008-04-08 15:39Ken MorelandAssigned ToKen Moreland => Berk Geveci
2008-04-08 15:39Ken MorelandCategory => 3.4
2008-06-27 11:16Kent EschenbergNote Added: 0012557
2008-06-27 11:17Kent EschenbergFile Added: test8.pvti
2008-06-27 11:17Kent EschenbergFile Added: test8a.vti
2008-06-27 11:17Kent EschenbergFile Added: test8b.vti
2008-06-27 11:18Kent EschenbergFile Added: pvti_test2.png
2008-06-27 11:18Kent EschenbergFile Added: pvti_test3.png
2008-07-02 16:19Berk GeveciNote Added: 0012602
2009-02-16 17:37Ken MorelandAssigned ToBerk Geveci => Burlen
2009-02-16 17:37Ken MorelandCategory3.4 => 3.6
2009-05-05 14:59Utkarsh AyachitAssigned ToBurlen => Berk Geveci
2009-05-05 15:00Utkarsh AyachitNote Added: 0016298
2009-05-05 15:00Utkarsh AyachitNote Deleted: 0016298
2009-05-05 15:01Utkarsh AyachitNote Added: 0016299
2009-05-13 13:44Utkarsh AyachitTarget Version => 3.6
2009-08-25 18:31Utkarsh AyachitAssigned ToBerk Geveci => Utkarsh Ayachit
2009-08-25 18:31Utkarsh AyachitCategory => Bug
2009-08-25 18:31Utkarsh AyachitTarget Version3.6 => 3.6.2
2009-09-01 14:15Utkarsh AyachitNote Added: 0017286
2009-09-01 14:15Utkarsh AyachitStatustabled => @80@
2009-09-01 14:15Utkarsh AyachitResolutionopen => unable to reproduce
2009-09-09 21:55Alan ScottNote Added: 0017364
2009-09-09 21:55Alan ScottStatus@80@ => closed

Notes
(0009853)
Ken Moreland   
2007-12-10 18:40   
Also check out parallel regular grids (.pvtr) and parallel structured grids (.pvts). I see similar extent related problems with those. I have example files if you need them.
(0010296)
Ken Moreland   
2008-01-24 19:00   
More information: The data set in question happens to be split into 16 parts, and I see both the .pvd and .pvti readers both work fine when run on 16 processes. This leads me to believe that this filter is being far too naive about loading the files. Each process is just loading one file. Ghost cells are also incorrectly handled.
(0012557)
Kent Eschenberg   
2008-06-27 11:16   
More information and a very simple test case. This suggests an almost complete failure of the logic in the pvti file reader.

File test8.pvti references test8a.vti (extents 0 1 0 1 0 1) and test8b.vti (extents 1 2 0 1 0 1).

When loaded in the order of pvti, a, b I see image pvti_test2.png. After restarting and loading in order b, a, pvti I see image pvti_test3.png. It looks like all 6 of the displayed objects are wrong. The min value of "temp" is 0 though the values are actually 10 to 24.

This was done using standalone, serial PV on a 32-bit workstation compiled from a download from CVS on June 25 at 4:42pm.
(0012602)
Berk Geveci   
2008-07-02 16:19   
NOTE: There are two issues reported here. The one reported by Kent is not a bug. The files test8a.vti and test8b.vti are wrong. In each of the pieces, the WholeExtent and the Piece Extent have to match. The pvti file itself is the only one that should have the actual WholeExtent. So instead of

<VTKFile type="ImageData" version="0.1">
  <ImageData
    WholeExtent="0 2 0 1 0 1"
    Origin="0 0 0"
    Spacing="1 1 1"
  >
    <Piece Extent="0 1 0 1 0 1">
 
it should be

<VTKFile type="ImageData" version="0.1">
  <ImageData
    WholeExtent="0 1 0 1 0 1"
    Origin="0 0 0"
    Spacing="1 1 1"
  >
    <Piece Extent="0 1 0 1 0 1">
(0016299)
Utkarsh Ayachit   
2009-05-05 15:01   
Reminder sent to: Berk Geveci

Berk, do you think you can take a look at this bug?
(0017286)
Utkarsh Ayachit   
2009-09-01 14:15   
Tried with 4 server (remote and local render). Didn't produce any errors for me.
(0017364)
Alan Scott   
2009-09-09 21:55   
Works for me. Tested with 6 remote servers.