View Issue Details [ Jump to Notes ] [ Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0011515VTK(No Category)public2010-11-24 18:072012-12-12 12:06
ReporterDavid Gobbi 
Assigned ToDavid Gobbi 
PrioritylowSeveritycrashReproducibilityalways
StatusclosedResolutionfixed 
PlatformOSOS Version
Product Version 
Target VersionFixed in Version5.10.0 
Summary0011515: Pipeline crashes with multiple update extent requests, if different extents are requested
DescriptionWhen the vtkStreamingDemandDrivenPipeline receives update extent requests from more than one downstream algorithm, the extent from the final request is used, even though it may be smaller than the extent from other requests. This can lead to a segfault when the final extent is smaller than the extent required by some of the downstream algorithms.
Steps To ReproduceRun the following tcl example. It will give the following error (assuming a 2-CPU machine) and then crash. I'm guessing that on a 1-CPU machine, it will crash without giving the error.

ERROR: In /Volumes/Work/Kitware/vtk-git/Filtering/vtkImageData.cxx, line 1472
vtkImageData (0x10630fe10): GetScalarPointer: Pixel (0, 128, 0) not in memory.
 Current extent= (0, 127, 0, 127, 0, 0)

===========
package require vtk

# Simple example that crashes when an UpdateExtent request
# from one algorithm is overwritten by a smaller UpdateExtent
# request from another algorithm

# read an image that has an extent of 0 255 0 255 0 0
vtkPNGReader reader
reader SetDataSpacing 0.8 0.8 1.5
reader SetFileName "$VTK_DATA_ROOT/Data/fullhead15.png"

# Uncomment this to make the crash disappear
#reader Update

# clip the image down to 0 127 0 127 0 0
vtkImageClip clip
clip SetInputConnection [reader GetOutputPort]
clip SetOutputWholeExtent 0 127 0 127 0 0

# do an operation on the clipped and unclipped data
vtkImageBlend blend
blend SetInputConnection [reader GetOutputPort]
blend AddInputConnection [clip GetOutputPort]

vtkImageViewer viewer
viewer SetInputConnection [blend GetOutputPort]
viewer SetColorWindow 2000
viewer SetColorLevel 1000
viewer Render
Additional InformationAssigned to myself, because it has been bugging me (hah!) so I've been thinking about it on and off.

One fix would be to add another information key, called COMBINED_UPDATE_EXTENT, that is the union of all update extent requests from all downstream algorithms. Every time an algorithm received REQUEST_UPDATE_EXTENT, it would combine the requested update extent with COMBINED_UPDATE_EXTENT. Then, in REQUEST_DATA (i.e. at the end of the update cycle), the COMBINED_UPDATE_EXTENT would be cleared.

This solution would be specific to 3D extents, i.e. structured data sets like vtkImageData.

Here is a patch that implements the solution described above:
http://review.source.kitware.com/446 [^]
TagsNo tags attached.
Project
Type
Attached Files

 Relationships

  Notes
(0029876)
David Gobbi (developer)
2012-12-12 12:06

 Patch was merged on Feb 25, 2011.

 Issue History
Date Modified Username Field Change
2010-11-24 18:07 David Gobbi New Issue
2010-11-24 18:10 David Gobbi Steps to Reproduce Updated
2010-11-24 18:11 David Gobbi Category Development => (No Category)
2010-11-24 18:13 David Gobbi Steps to Reproduce Updated
2010-11-26 11:46 David Gobbi Assigned To => David Gobbi
2010-11-26 11:46 David Gobbi Status backlog => tabled
2010-11-26 12:06 David Gobbi Additional Information Updated
2010-11-26 12:42 David Gobbi Additional Information Updated
2010-11-26 17:53 David Gobbi Note Added: 0023559
2010-11-26 18:04 David Gobbi Additional Information Updated
2010-11-26 18:20 David Gobbi Note Deleted: 0023559
2010-11-26 18:21 David Gobbi Additional Information Updated
2011-02-25 21:26 David Gobbi Status tabled => @80@
2011-02-25 21:26 David Gobbi Resolution open => fixed
2012-12-12 12:06 David Gobbi Note Added: 0029876
2012-12-12 12:06 David Gobbi Status customer review => closed
2012-12-12 12:06 David Gobbi Fixed in Version => 5.10.0

Copyright © 2000 - 2018 MantisBT Team