[vtk-developers] Site v20n17.pbm.ihost.com needs attention

Dave Partyka dave.partyka at kitware.com
Thu Jan 14 12:57:11 EST 2010


I am actually more inclined to just disable metaIO on that machine as it
doesn't appear to have issues anywhere else. If no one objects I will set
VTK_USE_METAIO off as it defaults to ON for that machine.

On Thu, Jan 14, 2010 at 12:54 PM, Bill Lorensen <bill.lorensen at gmail.com>wrote:

> Not specualting, just doing the best I can with my limited knowledge.
>
> On Thu, Jan 14, 2010 at 12:42 PM, Francois Bertel
> <francois.bertel at kitware.com> wrote:
> > Stop speculating, the issue is in metaio. Here is the reason of the
> > failure and there is unfortunately no trivial solution right now (even
> > for Brad King!):
> >
> > From an private thread of discussion involving me, Luis, Julien Jomier
> > and Brad King, Dec 4 2009:
> >
> > Brad King said:
> >
> > "
> > $ nm libvtkmetaio.so | grep ElementByteOrderFix
> > .vtkmetaio::MetaArray::ElementByteOrderFix() T   271508768         172
> > .vtkmetaio::MetaImage::ElementByteOrderFix(long) T   269262912
> 180
> > ...
> >
> > $ nm libvtkIO.so| grep ElementByteOrderFix
> > ...
> > vtkmetaio::MetaImage::ElementByteOrderFix(long long) U           -
> >
> > Something is causing streamoff to be a different size in one of vtkIO's
> > translation units than in metaio's translation units.  It's probably the
> > KWSys Configure.h that defines _LARGEFILE_SOURCE and related macros (this
> > was something Andy Cedilnik added at some point to "guarantee" LFS
> support
> > in all Kitware projects)."
> >
> > Brad King said:
> >
> > "
> > [I recommends to] teach MetaIO about large file support.
> >
> > Doing it right is a bit tricky.  This is one of those cases that requires
> > users of a library to define macros before including its headers.  We
> have
> > generally avoided this requirement by configuring feature settings into
> > header files such as vtkConfigure.h.  However, the problem here is that
> > enabling large file support means defining a macro like _LARGEFILE_SOURCE
> > in *every* translation unit before any *system* header is included.  This
> > cannot be done reliably in a place like vtkConfigure.h because users
> could
> > write this:
> >
> > #include <stdio.h>
> > #include <vtkObject.h> /* Uh oh, _LARGEFILE_SOURCE is defined too late!
> */
> >
> > By the time our configured header gets included, the system API has
> already
> > been loaded without large file support!  If instead we require users to
> > add -D_LARGEFILE_SOURCE on the command line then it will apply to the
> > system API no matter when it is included.
> >
> > A better approach is this:
> >
> >  # CMakeLists.txt
> >  # ... try compiles to detect proper LFS_MACRO ...
> >  add_definitions(-D${LFS_MACRO})
> >
> >  /* myconfig.h.in */
> >  #ifndef @LFS_MACRO@
> >  # error "This source requires -D at LFS_MACRO@"
> >  #endif
> >
> > The LFS_MACRO value can be put in the FooConfig.cmake file for use by
> > outside projects.
> >
> > Unfortunately we did not understand this problem back when the macros
> were
> > put in KWSys, and changing it there now is very hard for compatibility :(
> > "
> >
> >
> > On Thu, Jan 14, 2010 at 12:30 PM, Dave Partyka <dave.partyka at kitware.com>
> wrote:
> >> Humm I didn't noticed that it was using nightly CMake. I
> most certainly will
> >> give it a shot. Thanks for pointing that out!
> >>
> >> On Thu, Jan 14, 2010 at 12:27 PM, Bill Lorensen <
> bill.lorensen at gmail.com>
> >> wrote:
> >>>
> >>> Dave,
> >>>
> >>> I noticed that this build is using cmake2.9. Have you tried it with
> >>> cmake2.8?
> >>>
> >>> Bill
> >>>
> >>> On Thu, Jan 14, 2010 at 11:54 AM, Dave Partyka <
> dave.partyka at kitware.com>
> >>> wrote:
> >>> > I just fixed all the compile errors late last week and now it just
> has
> >>> > test
> >>> > failures. I'll keep prodding at it.
> >>> >
> >>> > On Thu, Jan 14, 2010 at 11:52 AM, Bill Lorensen
> >>> > <bill.lorensen at gmail.com>
> >>> > wrote:
> >>> >>
> >>> >> Folks,
> >>> >>
> >>> >> Late October was the last time the site v20n17.pbm.ihost.com had a
> >>> >> green dashboard.
> >>> >>
> >>> >> Could the owner of this build please take a look and see if the
> >>> >> problems can be resolved? Currently it is reporting 610 failing
> tests:
> >>> >> http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=514704
> >>> >>
> >>> >> Or, perhaps we no longer support this compiler: AIX00C518-xlC ?
> >>> >>
> >>> >> Bill
> >>> >> _______________________________________________
> >>> >> 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
> >>> >>
> >>> >
> >>> >
> >>
> >>
> >> _______________________________________________
> >> 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
> >>
> >>
> >>
> >
> >
> >
> > --
> > François Bertel, PhD  | Kitware Inc. Suite 204
> > 1 (518) 371 3971 x113 | 28 Corporate Drive
> >                      | Clifton Park NY 12065, USA
> > _______________________________________________
> > 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
> >
> >
> _______________________________________________
> 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
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20100114/d021e185/attachment.html>


More information about the vtk-developers mailing list