Hi,
I tried to install octave. First I got a checksum error because the port uses SparseSuite-2.1.1 while the current downloaded version is 2.3.1. After changing to the old version the chacksum was ok but I get compilation errors from that same package SparseSuite-2.3.1:
cc -O -c -DCOMPLEX -I../../AMD/Include -I../../COLAMD -I../../BTF/ Include -I../Include -I../../UFconfig ../Source/klu.c -o klu_z.o In file included from /usr/include/math.h:26, from ../Include/klu_internal.h:87, from ../Source/klu.c:52: /usr/include/architecture/ppc/math.h:686: error: duplicate member 'component'
Anybody knows what's happening?
Looks like the same thing that's happening here: https:// svn.macosforge.org/projects/macports/ticket/11014
There's a patch posted that should fix it.
Thanks a lot for pointing me to the patch. Now it worked. However, along the way gcc-4.0.3 was also installed even though I have gcc-4.0.1 available which comes with Xcode. I wonder whether this is really necessary? Is there a way to tell `port' which compiler to use? Joerg -- Jörg Frauendiener | Tel.: +49-7071-2975922 Theoretische Astrophysik | Fax: +49-7071-29-5889 Universität Tübingen | joerg.frauendiener@uni-tuebingen.de Auf der Morgenstelle 10 | http://www.tat.physik.uni-tuebingen.de/ ~joergf D-72076 Tübingen
On Dec 13, 2006, at 10:38, Jörg Frauendiener wrote:
Thanks a lot for pointing me to the patch. Now it worked. However, along the way gcc-4.0.3 was also installed even though I have gcc-4.0.1 available which comes with Xcode.
I wonder whether this is really necessary? Is there a way to tell `port' which compiler to use?
No; that's not for the user to decide; that's for the port author to decide. The octave port specifies that it requires the gcc40 port: depends_run bin:gnuplot:gnuplot \ port:gcc40 And in fact on Intel, it requires the gcc42 port: platform darwin 8 i386 { depends_run-delete port:gcc40 depends_run-append port:gcc42 It looks like this was changed a month ago: http://trac.macosforge.org/projects/macports/changeset/20584 You can ask the committer or the octave maintainer for more information if necessary. You can of course also play around with the portfile yourself, removing the gcc40 line, uninstalling the gcc40 port, and seeing if it still works for you, and telling the committer and the maintainer about this if it does. Note that since this port has some variation between ppc and i386 hardware, it would be best for you to test on both types of machine, lest you break one or the other with your changes.
Hi Joerg, On Dec 13, 2006, at 11:38 AM, Jörg Frauendiener wrote:
Hi,
I tried to install octave. First I got a checksum error because the port uses SparseSuite-2.1.1 while the current downloaded version is 2.3.1. After changing to the old version the chacksum was ok but I get compilation errors from that same package SparseSuite-2.3.1:
cc -O -c -DCOMPLEX -I../../AMD/Include -I../../COLAMD -I../../BTF/ Include -I../Include -I../../UFconfig ../Source/klu.c -o klu_z.o In file included from /usr/include/math.h:26, from ../Include/klu_internal.h:87, from ../Source/klu.c:52: /usr/include/architecture/ppc/math.h:686: error: duplicate member 'component'
Anybody knows what's happening?
Looks like the same thing that's happening here: https:// svn.macosforge.org/projects/macports/ticket/11014
There's a patch posted that should fix it.
Thanks a lot for pointing me to the patch. Now it worked. However, along the way gcc-4.0.3 was also installed even though I have gcc-4.0.1 available which comes with Xcode.
I wonder whether this is really necessary? Is there a way to tell `port' which compiler to use?
Yes, this is really necessary. The gcc40 port was built so you would have a gfortran that is ABI compatible with your gcc and g++. Apple does not supply gfortran so MacPorts has to build it. (And Octave really does have modules written in c, c++ and fortran :-) ) The Octave port uses Apple's gcc and g++, which means it may well fail to build when Apple's compilers move from 4.0.x to 4.1 or 4.2. The R port takes a more conservative approach and builds with gcc, g++ and gfortran from a single release. We used to have a fortran only port (g77) but there were endless problems in getting executables to link properly with modules compiled by different version of Apple's compilers. Best Wishes, Greg
Joerg -- Jörg Frauendiener | Tel.: +49-7071-2975922 Theoretische Astrophysik | Fax: +49-7071-29-5889 Universität Tübingen | joerg.frauendiener@uni-tuebingen.de Auf der Morgenstelle 10 | http://www.tat.physik.uni-tuebingen.de/ ~joergf D-72076 Tübingen
_______________________________________________ macports-users mailing list macports-users@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo/macports-users
On Dec 13, 2006, at 6:00 PM, Gregory Wright wrote:
Yes, this is really necessary. The gcc40 port was built so you would have a gfortran that is ABI compatible with your gcc and g++. Apple does not supply gfortran so MacPorts has to build it. (And Octave really does have modules written in c, c++ and fortran :-) )
Is gcc40 the standard port to depend upon when one needs a fortran compiler? The reason for all the current py-scipy bugs on trac is that the port doesn't install a fortran compiler as a dependancy. I put up a patch for it, but told it to depend upon gcc41. Should that be changed to gcc40? Thanks, Andy
On Dec 14, 2006, at 12:09 AM, Andy Kish wrote:
On Dec 13, 2006, at 6:00 PM, Gregory Wright wrote:
Yes, this is really necessary. The gcc40 port was built so you would have a gfortran that is ABI compatible with your gcc and g++. Apple does not supply gfortran so MacPorts has to build it. (And Octave really does have modules written in c, c++ and fortran :-) )
Is gcc40 the standard port to depend upon when one needs a fortran compiler? The reason for all the current py-scipy bugs on trac is that the port doesn't install a fortran compiler as a dependancy. I put up a patch for it, but told it to depend upon gcc41. Should that be changed to gcc40?
I think it should be changed to gcc42 since gcc40 doesn't work on Intel Macs (last time I checked). I hope this helps. -Anthony
Thanks, Andy _______________________________________________ macports-users mailing list macports-users@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo/macports-users
For what is worth, As a longtime member of the octave mailing list, and as someone who regularly builds octave without MacPorts or Fink, let me add the current thinking about building octave under Mac OS X 10.4.x: Apple's gcc 4.0.1 together with g95.org's g95 work well under both PPC and Intel machines for building octave 2.9.9. In my experience, gfortran has caused lots of problems. On Dec 14, 2006, at 5:26 AM, Anthony Agelastos wrote:
On Dec 14, 2006, at 12:09 AM, Andy Kish wrote:
On Dec 13, 2006, at 6:00 PM, Gregory Wright wrote:
Yes, this is really necessary. The gcc40 port was built so you would have a gfortran that is ABI compatible with your gcc and g++. Apple does not supply gfortran so MacPorts has to build it. (And Octave really does have modules written in c, c++ and fortran :-) )
Is gcc40 the standard port to depend upon when one needs a fortran compiler? The reason for all the current py-scipy bugs on trac is that the port doesn't install a fortran compiler as a dependancy. I put up a patch for it, but told it to depend upon gcc41. Should that be changed to gcc40?
I think it should be changed to gcc42 since gcc40 doesn't work on Intel Macs (last time I checked). I hope this helps.
-Anthony
Thanks, Andy
Marius -- Marius Schamschula Webmaster The Huntsville Macintosh Users Group www.hmug.org webmaster at hmug dot org marius at schamschula dot com
participants (6)
-
Andy Kish
-
Anthony Agelastos
-
Gregory Wright
-
Jörg Frauendiener
-
Marius Schamschula
-
Ryan Schmidt