[Insight-developers] Broken dashboard -- More info

Hans Johnson hjohnson at psychiatry.uiowa.edu
Tue Feb 15 09:26:48 EST 2005


Gordon,

The errors were not really due to NrrdIO, but rather the compiler
dependancies relating to NrrdIO.  The new version of NrrdIO.h no longer
has airExists_d (and by extension itk_airExists_d).  The previously
built itk IO wrapper (itkNrrdIO.h, itkNrrdIO.cxx, etc --> itkNrrdIO.o)
object files were originally built against a version of NrrdIO/NrrdIO.h
that did declare and define itk_airExists_d.  After the upgrade of
NrrdIO/NrrdIO.h, this should a triggered a rebuild of itkNrrdIO.o and a
recreation of libitkIO.a.  The failure occurred in that the compiler
dependancies did not trigger a rebuild of the object files, which did
not trigger a rebuild of the libitkIO.a libraries, and when the final
executables that depend on libitkIO.a and libNrrdIO.a were linked, the
linking of the final executables failed due to the missing
itk_airExists_d function found the now outdated libitkIO.a.

Hans


On Tue, 2005-02-15 at 07:45, Gordon Kindlmann wrote:
> hi,
> 
> > By manually deleting the Insight/Code/IO/itkNrrdIO*.o files out of my  
> > build directory, a new build of those object files is forced.  This  
> > fixes the faulty builds.
> 
> Thanks dealing with this!  I would have no idea how to debug these  
> problems.
> 
> So that I can be more informed about how the build and testing  
> procedures work, can you point me to the dashboard that was reporting  
> problems for which you were apologizing?
> 
> When I look at
> http://www.itk.org/Testing/Dashboard/20050215-0500-Nightly/
> Dashboard.html
> it seems like very few of the failures have anything to do with NrrdIO.
> 
> Gordon
> 



More information about the Insight-developers mailing list