[Paraview] Troubles building mathstuf/qhull subproject using PV 4.4.0 superbuild with Intel >14.x compilers

[Chuck Atkins]

Hi Giovanni,

It seems there's still some compatibility issues with upstream and the
Intel compiler.  Really it's sloppy logic in how the C headers and their
linkage are specified.  I've addressed this in a fork on github:
https://github.com/chuckatkins/qhull.git .  Basically I've made the headers
responsible for defining their C linkage instead of forcing the consuming
C++ code to know about it.  I've got a pull request to merge it upstream
but until then you can try it from my fork.  I tested it on Moonlight and
was able to build with the latest GNU, Intel and PGI compilers.



- Chuck

On Wed, Jan 20, 2016 at 9:34 AM, Ben Boeckel <ben.boeckel at kitware.com>
wrote:

> On Tue, Jan 19, 2016 at 19:05:38 -0700, Giovanni A. Cone wrote:
> >     I'm attempting to get PV 4.4.0 builds of Paraview
> > against Intel 14 and 15 on our HPC systems.  One item that
> > some users have asked us for was the Cosmotools integration.
> > This ultimately needs the Qhull library to be built.  Builds
> > against GCC 4.8.2 went through without a hitch.  However the
> > Intel builds all failed with an odd cryptic syntax error
> > message that I do not fully comprehend (sorry I'm still
> > stuck in Fortranistan).
> >
> > Building CXX object
>            ^^^
>
> See below.
>
> > CMakeFiles/qhullcpp.dir/src/libqhullcpp/Coordinates.cpp.o
> > In file included from
> >
> /tmp/conega/BUILD/lightshow/paraview-4.4.0-osmesa/intel-15_openmpi-1.6.5/BUILD01/qhull/src/qhull/src/libqhull/poly2.c(16):
> >
> /tmp/conega/BUILD/lightshow/paraview-4.4.0-osmesa/intel-15_openmpi-1.6.5/BUILD01/qhull/src/qhull/src/libqhull/qhull_a.h(106):
> > warning #77: this declaration has no storage class or type
> > specifier
> >    template <typename T>
> >    ^
> >
> > In file included from
> >
> /tmp/conega/BUILD/lightshow/paraview-4.4.0-osmesa/intel-15_openmpi-1.6.5/BUILD01/qhull/src/qhull/src/libqhull/poly2.c(16):
> >
> /tmp/conega/BUILD/lightshow/paraview-4.4.0-osmesa/intel-15_openmpi-1.6.5/BUILD01/qhull/src/qhull/src/libqhull/qhull_a.h(106):
> > error: expected a ";"
> >    template <typename T>
> >             ^
> >  From what little I understand of C and C++, I believe that
> > "templates" are a C++ construct.  I'm thinking the Intel
> > compiler is attempting to digest the above as "straight-up
> > C" and failing.
>
> Marked above, it is compiling as C++, but it appears that it is trying
> to parse template as an identifier rather than a keyword. My guess is
> that it is something like:
>
>     struct foo {
>         int member;
>     } // no semicolon
>     template ...
>
> which parses into a function returning a foo struct (declared there as
> well) named 'template'.
>
> > The error messages put the blame on
> > QHULL_SRC_ROOT/src/libqhull/qhull_a.h  around line 105
> >
> > I then decided to snoop on github.com for another qhull
> > repo.  I found one
> > at
> >
> > https://github.com/qhull/qhull
> >
> >  From which I believe the "https://github.com/mathstuf/qhull
> > next branch" repo that Paraview wants to use was forked from.
>
> Not originally, but it is now that qhull has moved from sourceforge to
> github (it was originally forked from George Zegaris' clone which he
> mirrored from sourceforge to github, but I recently re-forked the proper
> repo).
>
> > Point being that the mainline non-forked Qhull has something
> > different in the preprocessor conditional surrounding the
> > above offensive line of source:
> >
> > #if defined(__cplusplus) && defined(__INTEL_COMPILER) &&
> > !defined(QHULL_OS_WIN)
> >
> > while the mathstuf/qhull repo DOES NOT have the above
> > "defined(__cplusplus)" preproc conditional.
>
> My 'next' branch is based off an old release, so differences to upstream
> are not surprising.
>
> > I tested to see if the addition would at least allow the
> > 'mathstuf/qhull next branch' source to properly build with
> > Intel 14 and 15.  It got about to 36% and then failed with
> > the error message:
>
> All my branch does is fix an Xcode compilation errors where std::
> iterator tags are forward declared improperly.
>
> > In file included from
> > /tmp/conega/qhull_SRC/src/libqhullcpp/QhullIterator.h(13),
> > from /tmp/conega/qhull_SRC/src/libqhullcpp/Coordinates.h(13),
> > from /tmp/conega/qhull_SRC/src/libqhullcpp/Coordinates.cpp(11):
> > /tmp/conega/qhull_SRC/src/libqhull/qhull_a.h(106): error:
> > this declaration may not have extern "C" linkage
> >    template <typename T>
> >    ^
>
> It appears that something is doing
>
>     extern "C" {
>     #include "qhull.h"
>     }
>
> which won't work since qhull is a C++ header. Are there any 'extern "C"'
> bits around? Passing -E (or whatever Intel's equivalent is) to dump the
> preprocessed source would help to see. The __cplusplus check may be
> guarding against that line in upstream.
>
> > So at least the conditional helped a bit for what I believe
> > is the "straight up C" part of the build.
> >
> >
> > I'll CC the author of the fork Ben Boeckel about this.
>
> That's me too :) . Seems I forgot to change the author setting for that
> clone :/ . Rewrote the commit with the proper email.
>
> --Ben
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ParaView Wiki at:
> http://paraview.org/Wiki/ParaView
>
> Search the list archives at: http://markmail.org/search/?q=ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/paraview
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20160121/13badec3/attachment.html>