[ITK] Segmentation Fault at SpatialObject::New()

[Jinwoo Choi]

Hi Matt, thank you for the reply. 

Initially the segfault was happening randomly as a part of a large software. I was trying to see if another statement somewhere in the codebase was causing this to segfault.
I tried to isolate the object creation code to see if it still segfaults. Hence the simple code I have posted. 

Is there anyway to make sure that the atomic decrement barrier works as intended? 

Thank you again, 

Jin









-------- Original message --------
From: Matt McCormick <matt.mccormick at kitware.com> 
Date:05/01/2014  4:45 PM  (GMT-05:00) 
To: Jinwoo Choi <jinwoo.choi at mail.utoronto.ca> 
Cc: community at itk.org 
Subject: Re: [ITK] Segmentation Fault at SpatialObject::New() 

Hi Jin,

The OSAtomicDecrement64Barrier [1] might not be behaving, but why is
this code being created?  Why is the only statement the creation of an
object in a "while(true)"?  Does it still crash with other statements
in the loop?

Thanks,
Matt

[1] http://itk.org/gitweb?p=ITK.git;a=blob;f=Modules/Core/Common/src/itkLightObject.cxx;h=ce95746f71180f7084a5cdcc7d9f332356b049d3;hb=HEAD#l204

On Thu, May 1, 2014 at 2:24 PM, Jinwoo Choi
<jinwoo.choi at mail.utoronto.ca> wrote:
> Dear all,
>
> I have the following simple code:
>
> #include <itkSpatialObject.h>
> int main(int argc, char *argv[])
> {
>   typedef itk::SpatialObject<2> SOT;
>   while (true) {
>     SOT::Pointer sot = SOT::New();
>   }
> }
>
> Which results in a segmentation fault with the following stack trace:
>
> #0  0x00007fff8986bce2 in __pthread_kill ()
> #1  0x00007fff8dd237d2 in pthread_kill ()
> #2  0x00007fff8dd14a7a in abort ()
> #3  0x00007fff8dd7384c in free ()
> #4  0x00000001093c2f69 in itk::LightObject::~LightObject (this=0x115f00370)
> at /Users/jin/MyVTK/ITK/ITK/Modules/Core/Common/src/itkLightObject.cxx:254
> #5  0x00000001093c2e7e in itk::LightObject::UnRegister (this=0x115f00370) at
> /Users/jin/MyVTK/ITK/ITK/Modules/Core/Common/src/itkLightObject.cxx:206
> #6  0x0000000100038e14 in itk::SmartPointer<itk::VectorContainer<unsigned
> long, itk::Point<double, 2u> > >::UnRegister (this=0x7fff5fbff7b0) at
> itkSmartPointer.h:157
> #7  0x0000000100038dc9 in itk::SmartPointer<itk::VectorContainer<unsigned
> long, itk::Point<double, 2u> > >::~SmartPointer (this=0x7fff5fbff7b0) at
> itkSmartPointer.h:65
> #8  0x0000000100038da5 in itk::SmartPointer<itk::VectorContainer<unsigned
> long, itk::Point<double, 2u> > >::~SmartPointer (this=0x7fff5fbff7b0) at
> itkSmartPointer.h:64
> #9  0x000000010004cc04 in itk::BoundingBox<unsigned long, 2, double,
> itk::VectorContainer<unsigned long, itk::Point<double, 2u> > >::BoundingBox
> (this=0x115f00270) at itkBoundingBox.hxx:124
> #10 0x000000010004cb15 in itk::BoundingBox<unsigned long, 2, double,
> itk::VectorContainer<unsigned long, itk::Point<double, 2u> > >::BoundingBox
> (this=0x115f00270) at itkBoundingBox.h:125
> #11 0x0000000100028982 in itk::BoundingBox<unsigned long, 2, double,
> itk::VectorContainer<unsigned long, itk::Point<double, 2u> > >::New () at
> itkBoundingBox.h:90
> #12 0x0000000100028097 in itk::SpatialObject<2u>::SpatialObject
> (this=0x115f00000) at itkSpatialObject.hxx:35
> #13 0x0000000100027e35 in itk::SpatialObject<2u>::SpatialObject
> (this=0x115f00000) at itkSpatialObject.h:57
> #14 0x0000000100027b32 in itk::SpatialObject<2u>::New () at
> itkSpatialObject.h:137
> #15 0x00000001000279df in main (argc=1, argv=0x7fff5fbffb30) at
> /Users/jin/vurtigo/rtMain.cpp:64
> #16 0x0000000100027924 in start ()
>
> I have this issue with ITK4.2.1 and ITK 4.5.2, happens both Debug and
> Release compilation.
>
> Oddly, when I put a break point on (using gdb) itkSpatialObject.h:57 and
> print m_Bounds, delete all break points, and continue (continue in the main
> while loop), the segmentation fault stops occurring.
>
> My platform is OSX 10.7.5, and I am using /usr/bin/c++ (which I believe is
> clang), clang -v -version outputs the following:
> Apple LLVM version 4.2 (clang-425.0.28) (based on LLVM 3.2svn)
> Target: x86_64-apple-darwin11.4.2
> Thread model: posix
>
> Any insights into this issue will be greatly appreciated!
>
> Thank you,
>
> Jin
> _______________________________________________
> Community mailing list
> Community at itk.org
> http://public.kitware.com/cgi-bin/mailman/listinfo/community
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/community/attachments/20140502/7b814609/attachment-0002.html>