[vtk-developers] VTK test failures

[Mathieu Coursolle]

Hi,

First, to be able to build PPC only on an intel machine, you will need
the latest CVS cmake I think.

From what I can tell, the problem was there for a long time ago, and we were
just lucky it did not cause any tests to fail. 

I don't understand completely the fix, but that commit seems to solve
the issues:

<http://public.kitware.com/cgi-bin/viewcvs.cgi/Utilities/kwsys/
ProcessUNIX.c?rev=1.71&view=log>

Thanks.

Mathieu


>I'm trying to replicate this problem on my Intel Mac, and have so far
>been unsuccessful.  It may be that I'm not compiling PPC binaries
>correctly.  How do you compile as PPC binaries and verify the type of
>binaries being used?
>
>Also, I noticed that on today's dashboard, all of the tests that started
>failing on the 8th are now working again.  Why is that?
>
>-Ken
>
>> -----Original Message-----
>> From: Mathieu Coursolle [mailto:mcoursolle at rogue-research.com]
>> Sent: Thursday, January 10, 2008 4:47 PM
>> To: Mathieu Coursolle; Moreland, Kenneth; vtk-developpers
>> Subject: Re: [vtk-developers] VTK test failures
>> 
>> After some investigations, it seems to fail when built on PPC as well.
>> 
>> Here are the 3 configurations I tested:
>> 
>> Build as Intel, On intel : PASS
>> Build as PPC, On intel : FAIL
>> Build as PPC, On PPC : FAIL
>> 
>> Any one noticed that failure too?
>> 
>> Thanks.
>> 
>> Mathieu
>> 
>> >Hi,
>> >
>> >Since 2008-01-08, some new test failures appeared on one of our
>> dashboard:
>> ><http://www.vtk.org/Testing/Sites/RogueResearch4/Mac10.5-VTKCocoa-gcc-
>> >dbg-rosetta/20080108-0300-Nightly/Test.html>
>> >
>> >I was able to track down the cause to the following changes:
>> >
>> ><http://public.kitware.com/cgi-bin/viewcvs.cgi/Filtering/
>> >vtkColorTransferFunction.cxx.diff?cvsroot=VTK&r1=1.72&r2=1.73>
>> ><http://public.kitware.com/cgi-bin/viewcvs.cgi/Filtering/
>> >vtkColorTransferFunction.h.diff?cvsroot=VTK&r1=1.54&r2=1.55>
>> >
>> >This dashboard is build as Big endian (ppc) on a little endian system
>> >(intel), so it could be an endianness problem.
>> >
>> >Does any body knows why this change is causing those new test failures?
>> >
>> >Thanks.
>> >
>> >Mathieu
>> >
>> >
>> >--
>> >____________________________________________________________
>> >Mathieu Coursolle                   mcoursolle at rogue-research.com
>> >Rogue Research                      www.rogue-research.com
>> >Montréal, Québec, Canada
>> >_______________________________________________
>> >vtk-developers mailing list
>> >vtk-developers at vtk.org
>> >http://www.vtk.org/mailman/listinfo/vtk-developers
>> >
>> 
>
>
>