On 12 Aug, 2005, at 03:01, Bill Northcott wrote: > On 12/08/2005, at 4:26 AM, Brian Granger wrote: > >> What version of gcc are you using? I have not had any luck >> getting scipy and its dependencies working with gcc 4.0. >> >> Maybe this has changed? Has anyone had success with gcc 4.0 yet? >> >> If you need to change the default version of gcc on Mac OS X use >> the command: >> >> sudo gcc_select 3.3 >> > > This is probably the solution. > Why? > because you can only use g77 with gcc-3.x OR gfortran with > gcc-4.x. You cannot use g77 with gcc-4 they are incompatible. > So if you have a Python/R/Perl/you-name-it binary containing > Fortran objects built with g77 you cannot extend it with code > compiled with gcc-4. This is true on any platform.
Why doesn't the setup.py script for scipy take care of this detail? Or how do I configure setup.py to take care of this dependency. So to be specific I select gcc 3.3 before I run scipy's setup.py build and setup.py install and then I set it back. The mac scipy install instructions list the following packages that scipy is dependent on. Some are downloadable binaries some you have to build. Do I need to redo all these packages with gcc 3.3 or are some of them smart enough to select the correct compiler? Numeric Numarray G77 FFTW F2PY GNUPlot AquaTerm basically I have to start over? Is DarwinPorts smart enough to use the correct compiler? Thanks for you help do you have documented how you set up scipy. I am trying to figure out what approach to use. I wanted to use framework python because I wanted to do some mac gui work as well but may I should just cut my losses and use DarwinPorts for scientific stuff. Any good reasons not to? > > The issue arises mainly on Tiger because it does not have an > included Fortran compiler and out of the box it defaults to using > gcc-4. Binary packages are shipped built with g77 and sometimes > including g77. These are bound to fail on Tiger unless the default > compiler is changed using the gcc_select command as above. > > I have built Scipy with gcc-4, but you must use gfortran to compile > all the Fortran code for that to work. > > I do wish someone on Python, R-project or some other web site would > document this issue. I must have posted at least twenty messages > saying the same thing. > > Bill Northcott > ********************************************************************** Samuel M. Smith Ph.D. 2966 Fort Hill Road Eagle Mountain, Utah 84043 801-768-2768 voice 801-768-2769 fax ********************************************************************** "The greatest source of failure and unhappiness in the world is giving up what we want most for what we want at the moment" ********************************************************************** _______________________________________________ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig
