On Wed, Nov 29, 2017 at 4:37 PM, Waleed Almutiry <walmu...@uoguelph.ca> wrote: > Hi all, > > I tried to upload my package to CRAN and did not pass the the incoming > checks. This was because the use of ieee_arithmetic module in a fortran > subroutine in the src folder. I built and checked the package using (R CMD > build) and (R CMD check) under IOS system and it is working fine without > errors. The gcc version in my system is (GNU Fortran (GCC) 5.2.0) while the > gcc version in the CRAN system is 4.9.3. This is the message that I have got > from the CRAN: [snip] > > I used the ieee_arithmetic in some subroutines to read and write results of > Infinity values. I used it inside the subroutine as > > IF (ieee_support_inf(Inf)) THEN Inf = ieee_value(Inf, ieee_positive_inf) END > IF >
Hello, Waleed. I handled Inf and NaN by writing a C function that passes the magic code words R_NaN and INFINITY and called them as external functions in Fortran. See the source code of 'utils_and_wrappers.c' and 'delaporte.f95' at [1] for a working example of "set_nan" and "set_inf"ideas. Originally, I used the actual bitpatterns to make the functions pure, if not elemental, but Soldaris SPARC through a fit since it was the opposite endian from the rest of the known universe. Using the magic R words will ensure that the proper values will be passed regardless of platform. Hope that helps, Avi [1] https://bitbucket.org/aadler/delaporte ______________________________________________ R-package-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-package-devel
