or should it be tucked away so that every compile of everything doesn't see 
it? 


#if defined(PETSC_HAVE_XMMINTRIN_H) && !defined(__CUDACC__)
#include <xmmintrin.h>
#endif

   if an example needs this include couldn't the example just include it 
directly?

  Thanks


> Begin forwarded message:
> 
> From: Matthew Knepley <[email protected]>
> Subject: Re: [petsc-maint] Error during linking
> Date: July 31, 2017 at 8:00:59 AM CDT
> To: Oliver Henrich <[email protected]>
> Cc: "[email protected]" <[email protected]>
> 
> On Mon, Jul 31, 2017 at 7:51 AM, Oliver Henrich <[email protected] 
> <mailto:[email protected]>> wrote:
> Hi Matt,
> 
> Many thanks for getting back so quickly and sorry for not including the files 
> (I keep repeating this as well to my users ;-)
> 
> Please find below the log files. As far as I can see it compiles with C.
> 
> I can confirm that commenting out the definition of the header file 
> (petscsys.h line # 1901) after compilation and before linking to my code 
> fixes the problem and delivers correct results.
> 
> I remember vaguely I had this problem in the past. But it took me now all 
> morning to fix it.
> 
> Satish, do we have a Sierra build machine? This looks like a plain vanilla 
> configuration, so we should see this error in the nightly builds. It
> should be easy to get a test that reproduces once I know exactly what 
> environment fails.
> 
>   Thanks,
> 
>      Matt
>  
> Cheers,
> Oliver
> 
> 
> On 31 Jul 2017, at 13:25, Matthew Knepley <[email protected] 
> <mailto:[email protected]><mailto:[email protected] 
> <mailto:[email protected]>>> wrote:
> 
> On Mon, Jul 31, 2017 at 5:18 AM, Oliver Henrich <[email protected] 
> <mailto:[email protected]><mailto:[email protected] 
> <mailto:[email protected]>>> wrote:
> Hi there,
> 
> I get compiling errors when linking PETSc against my code. The problem seems 
> to be related to Apple's marvellous Xcode and clang.
> 
> My system is Mac OS X Sierra 10.12.6 and I’m using Xcode 8.3.3 and clang 
> 8.1.0, but I could reproduce this issue also with older versions of PETSc 
> (3.5.2) and Xcode 8.2.1.
> 
> Here is the output shortly before and including the firstfatal error:
> 
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscksp.h:6:
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscpc.h:6:
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscmat.h:6:
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscvec.h:9:
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscis.h:7:
> In file included from /Users/ohenrich/petsc-3.7.6/include/petscsys.h:1901:
> In file included from 
> /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/clang/8.1.0/include/xmmintrin.h:27:
> /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/clang/8.1.0/include/mmintrin.h:44:24:
>  error:
>       expected ')'
> static __inline__ void __DEFAULT_FN_ATTRS
> 
> By commenting out line #1901 in 
> /petsc-3.7.6/include/petscsys.h:1901***after*** compiling PETSc I could 
> prevent this error from occurring. I’m running tests, but it seems my code 
> delivers the correct result.
> 
> In order to really know what happened, we need configure.log and make.log. 
> However, are you compiling PETSc with C++?
> There was a reported bug with the Intel headers and g++ where it was 
> misinterpreting the preprocessor hints about arguments,
> which seems similar to what is happening here. I believe putting extern "C" 
> in for this header might work as well if the above
> guess is correct. We are working to get configure to recognize and work 
> around this.
> 
>   Thanks,
> 
>      Matt
> 
> Best wishes and many thanks,
> Oliver
> 
> --
> Dr Oliver Henrich
> Department of Physics
> University of Strathclyde, Glasgow
> John Anderson Building 6.29
> Glasgow G4 0NG, Scotland, UK
> 
> Tel: +44 (0)141 548 4012
> Fax: +44 (0)141 552 2891
> --
> The University of Strathclyde is a charitable body, registered in Scotland 
> with registration number SC015263.
> 
> 
> 
> 
> --
> What most experimenters take for granted before they begin their experiments 
> is infinitely more interesting than any results to which their experiments 
> lead.
> -- Norbert Wiener
> 
> http://www.caam.rice.edu/~mk51/ <http://www.caam.rice.edu/~mk51/>
> 
> --
> Dr Oliver Henrich
> Department of Physics
> University of Strathclyde, Glasgow
> John Anderson Building 6.29
> Glasgow G4 0NG, Scotland, UK
> 
> Tel: +44 (0)141 548 4012 <tel:%2B44%20%280%29141%20548%204012>
> Fax: +44 (0)141 552 2891 <tel:%2B44%20%280%29141%20552%202891>
> --
> The University of Strathclyde is a charitable body, registered in Scotland 
> with registration number SC015263.
> 
> 
> 
> 
> -- 
> What most experimenters take for granted before they begin their experiments 
> is infinitely more interesting than any results to which their experiments 
> lead.
> -- Norbert Wiener
> 
> http://www.caam.rice.edu/~mk51/ <http://www.caam.rice.edu/~mk51/>

Reply via email to