On Sat, Sep 11, 2010 at 5:35 PM, Dr. David Kirkby <david.kir...@onetel.net> wrote: > On 09/12/10 12:10 AM, François Bissey wrote: >>> >>> On 11 September 2010 21:48, Ondrej Certik<ond...@certik.cz> wrote: >>>> >>>> When building femhub and packages for femhub, I have to deal with >>>> these fortran issues as well. And I never understood >>>> >>>> a) why sage used g95 in the first place (yes I know it's smaller, but >>>> it's not standard at all imho) >>> >>> Agreed. I can't see the point of it. >>> >>>> b) all this mess with fortran --- lots of packages (like Trilinos) >>>> simply fail to compile thanks to some setup in Sage/FEMhub due to >>>> fortran/blas/atlas. It works in Ubuntu. >>>> >>>> So I just want to give you a big thumbs up to make these fortran >>>> issues less pain, and following standards more. >>>> >>>> Ondrej >>> >>> To follow "standards" in the lose sense of the word, we would drop the >>> name SAGE_FORTRAN and instead use FC like other packages to indicate >>> the path to a Fortran compiler. >>> >>> * We do not have SAGE_C, we use CC instead >>> * We do not have SAGE_C_PLUS_PLUS - we use CXX instead >>> * We DO use SAGE_FORTRAN, when everyone else uses FC now. >>> >>> That's a different issue though. Sage would need more changes to get >>> rid of SAGE_FORTRAN. But it would be worth it in my opinion. Having a >>> script as a compiler is a pain. One can''t see what options are passed >>> to the code. >>> >>> BTW, POSIX does not mandate the use of CC, CXX or FC, but does contain >>> a list of commonly used variables they suggest people do not use. FC >>> is one of them. So whilst it is not a "standard", it is rather >>> commonly used. >>> >> +1 to move to FC. >> >> surely you meant: "they suggest people use" rather than "do not use"? >> >> François >> > > No, I said what I meant. > > CC/CXX/FC are not standards as mandated by the Unix standard. So the Unix > standard does not suggest you use them. What they suggest you do is to not > use commonly used variables for anything else. This is what what is said. > > http://www.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html > > > "It is unwise to conflict with certain variables that are frequently > exported by widely used command interpreters and applications:" > > That list includes CC, CXX and FC. > > > So its agreed FC is widely use - why the hell does Sage have to be any > different?
I would use FC. As to gfortran on the Mac --- my understanding is that the fortran*.spkg package contains the binary of g95 for all supported architectures in the g95 dir. It also contains gfortran binary in the gfortran/ directory. I researched how to get a decent working gfortran binary on the Mac, but I didn't find anything useful, only http://hpc.sourceforge.net/. One more question about Mac --- how do you install gcc in there? Using xcode? So I would just ship gfortran binary that is compatible with gcc on the Mac somehow, and be done with it, and require it on linux, which I think is not a problem these days. Ondrej -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
