[vtk-developers] Failing Test: TestDescriptiveStatistics needs attention

[Marcus D. Hanwell]

On Thursday 31 December 2009 00:57:29 Pebay, Philippe P wrote:
> Hello all
> 
> Thanks Dave for having disabled the test. Yes, I have indeed been off
>  e-mail for the past few days and will return into that mode starting
>  tomorrow.
> 
> I had become aware of that problem last week, and sent a "help" message to
>  the infovis list, for I have not been able to replicate the bug on any of
>  my platforms. Therefore, all feedback from someone who can replicate the
>  issue on his platform will be very much appreciated. Thank you.
> 
I had a quick look at this yesterday as I can reproduce it here. The m2 
variable is zero, so m22 is zero and so it triggers the divide by zero 
exception. You could check for m2 being zero - what would you want the result 
to be in the case that m2 was zero? The line in question,

VTK/Infovis/vtkDescriptiveStatistics.cxx:536

Also, what platform are you on? I have been working with Francois to extend 
the range of platforms that can have divide by zero enabled for testing 
purposes. As far as I know it should work on Linux and Windows. I haven't been 
able to check, but the way I enabled it on Linux should work with recent GCC 
versions on other platforms.

Hope that helps.

Marcus
> 
> ________________________________________
> From: vtk-developers-bounces at vtk.org [vtk-developers-bounces at vtk.org] On
>  Behalf Of Dave Partyka [dave.partyka at kitware.com] Sent: Monday, December
>  28, 2009 9:35 AM
> To: Bill Lorensen
> Cc: VTK Developers
> Subject: Re: [vtk-developers] Failing Test: TestDescriptiveStatistics needs
>  attention
> 
> Okay, I'll leave it for now and send him more rubber chickens ;-)
> 
> On Mon, Dec 28, 2009 at 12:34 PM, Bill Lorensen
>  <bill.lorensen at gmail.com<mailto:bill.lorensen at gmail.com>> wrote: Agree.
>  And it is the holidays for some...
> 
> On Mon, Dec 28, 2009 at 12:29 PM, David Cole 
<david.cole at kitware.com<mailto:david.cole at kitware.com>> wrote:
> > It might be better to just leave it failing, back out the change or fix
> > it yourself... Disabling a test is a bad idea because nobody has an
> > impetus to re-enable the test in the future.
> > We can wait for a fix. Or if "we" can't then "we" can fix it ourselves...
> >
> >
> > On Mon, Dec 28, 2009 at 12:20 PM, Dave Partyka
> > <dave.partyka at kitware.com<mailto:dave.partyka at kitware.com>>
> >
> > wrote:
> >> Hi Bill,
> >>
> >> I emailed Philippe the day this broke but have not heard back from him.
> >> I typically don't cc the vtk list when sending a rubber chicken to those
> >> who break the dashboard ;-). I suspect he is gone for the holidays so I
> >> will just disable this test for the time being.
> >>
> >> On Mon, Dec 28, 2009 at 12:10 PM, Bill Lorensen
> >> <bill.lorensen at gmail.com<mailto:bill.lorensen at gmail.com>>
> >>
> >> wrote:
> >>> Folks,
> >>>
> >>> The subject test is failing on many platforms:
> >>>
> >>> http://www.cdash.org/CDash/testSummary.php?project=11&name=Infovis-Test
> >>>DescriptiveStatistics&date=2009-12-28
> >>>
> >>> According to cdash, the test has been failing since Dec 22 when these
> >>> files were checked in:
> >>> http://www.cdash.org/CDash/viewUpdate.php?buildid=499182
> >>>
> >>> This continuous build was the first to notice the problem:
> >>> http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=498908
> >>>
> >>> I ran the test in the debugger and here is the failing statement:
> >>> Program received signal SIGFPE, Arithmetic exception.
> >>> 0x0095f351 in vtkDescriptiveStatistics::Test (this=0x8764fe8,
> >>>    inData=0x87642a0, inMetaDO=0x8766f40, outMetaDO=0x87672b0)
> >>>    at
> >>> /home/lorensen/Projects/VTKHead/Infovis/vtkDescriptiveStatistics.cxx:53
> >>>6 536         double s = sqrt( n / ( m22 * m2 ) ) * m3;
> >>>
> >>> This is a result of
> >>> Warning: In
> >>> /projects/kmorel/dashboard/VTK/Infovis/vtkDescriptiveStatistics.cxx,
> >>> line 282
> >>> vtkDescriptiveStatistics (0x59a0c0): InData table does not have a
> >>> column Metric 3. Ignoring it.
> >>>
> >>> Warning: In
> >>> /projects/kmorel/dashboard/VTK/Infovis/vtkUnivariateStatisticsAlgorithm
> >>>.cxx, line 122
> >>> vtkDescriptiveStatistics (0x59a0c0): InData table does not have a
> >>> column Metric 3. Ignoring it.
> >>>
> >>>
> >>> The test purposely adds a bogus column. This causes the zero divide.
> >>> Since VTK does not throw exceptions, the bad data needs to be handled
> >>> in an alternative way to avoid the 0 divide.
> >>>
> >>> It is easy to miss a defect like this because the vtk dashboard seems
> >>> to have many tests failing on a continuing basis.
> >>>
> >>> Bill
> >>> _______________________________________________
> >>> Powered by www.kitware.com<http://www.kitware.com>
> >>>
> >>> Visit other Kitware open-source projects at
> >>> http://www.kitware.com/opensource/opensource.html
> >>>
> >>> Follow this link to subscribe/unsubscribe:
> >>> http://www.vtk.org/mailman/listinfo/vtk-developers
> >>
> >> _______________________________________________
> >> Powered by www.kitware.com<http://www.kitware.com>
> >>
> >> Visit other Kitware open-source projects at
> >> http://www.kitware.com/opensource/opensource.html
> >>
> >> Follow this link to subscribe/unsubscribe:
> >> http://www.vtk.org/mailman/listinfo/vtk-developers
> 
> _______________________________________________
> Powered by www.kitware.com
> 
> Visit other Kitware open-source projects at
>  http://www.kitware.com/opensource/opensource.html
> 
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtk-developers
> 

-- 
Marcus D. Hanwell, Ph.D.
R&D Engineer, Kitware Inc.
(518) 881-4937