[R-SIG-Mac] YA Mac 64 bit Question

Karl Goiser karl.goiser at anu.edu.au
Mon May 7 14:27:44 CEST 2007


Thank you Simon.

I'll give your suggestions a go!

Karl


On 07/05/2007, at 2:35 AM, Simon Urbanek wrote:

> Karl,
>
> On May 6, 2007, at 3:31 AM, Karl Goiser wrote:
>
>> Hello,
>>
>> I am trying to get a 64 version of R running on my Mac Pro (like
>> several others, it seems).  I
>
> You can try this:
>
> curl http://r.research.att.com/R-25-br-quad.tar.gz -O
> tar fvxz R-25-br-quad.tar.gz -C /
>
> It is a quad-arch framework (including 64-bit x86). To run the 64- 
> bit version, use
> /Library/Frameworks/R.framework/Resources/bin/R64
>
> If you want to use X11, you'll need 64-bit client in /usr/local/ 
> X11R7 (I used Xorg).
>
>
>> t seems there are two options, 10.4 and
>> 10.5:
>>
>> http://r.research.att.com/exp/ says to install Xcode [tick] and GNU
>> Fortran [tick], get the R source (from CRAN, I assume, e.g. http://
>> cran.au.r-project.org/src/base/R-2/R-2.5.0.tar.gz) [tick], and
>> configure with a suggested command line:
>>
>> ./configure CC='gcc -arch x86_64' CXX='g++ -arch x86_64'
>> F77='gfortran -arch x86_64' FC='gfortran -arch x86_64' --with-blas='-
>> framework vecLib' --with-lapack --without-x --with-system-zlib --
>> build=x86_64-apple-darwin8.8.1 --without-readline
>>
>>
>> Tiger:
>>
>> When I run it (without r_arch=x86_64) I get an error:
>> configure: error: --with-iconv=yes (default) and suitable iconv is
>> not available
>>
>> What is this and how do I address it?
>>
>
> There are two possibilities:
> 1) use --without-iconv (untested)
> 2) get 64-bit iconv:
>   a) compile it from Apple's sources
>     or
>   b)
> curl http://r.research.att.com/libiconv.2.dylib -o /tmp/libiconv. 
> 2.dylib \
> && sudo cp /tmp/libiconv.2.dylib /usr/lib/libiconv.2.dylib
>
> It is strange, because Xcode SDK contains 64-bit stub for iconv,  
> but the system library is 32-bit only.
>
>>
>> Leopard:
>> When I run the same configure as per the web site, I get a different
>> error:
>>
>> ./configure: line 61328: test: =: unary operator expected
>>
>> If I then go on to 'make', I get the following:
>> ld64: warning: option -headerpad_max_install_names is obsolete and
>> being ignored
>> ld64-72 failed: duplicate symbol _Rf_mkChar in CommandLineArgs.o and
>> CConverters.o
>>
>
> I'm not sure I can comment on this using a public mailing list, so  
> all I'll say is that the hot fix is to apply this patch to src/ 
> include/Rinternals.h:
>
> Index: Rinternals.h
> ===================================================================
> --- Rinternals.h        (revision 41454)
> +++ Rinternals.h        (working copy)
> @@ -926,7 +926,7 @@
> #define warningcall_immediate  Rf_warningcall_immediate
> #endif
> -#if defined(CALLED_FROM_DEFN_H) && !defined(__MAIN__) && (defined 
> (COMPILING_R) || ( __GNUC__ && !defined(__INTEL_COMPILER) ))
> +#if defined(CALLED_FROM_DEFN_H) && !defined(__MAIN__) && (defined 
> (COMPILING_R) || ( __GNUC__ && !defined(__INTEL_COMPILER) &&  
> (__APPLE_CC__ < 5400) ))
> #include "Rinlinedfuns.h"
> #else
> /* need remapped names here for use with R_NO_REMAP */
>
> I didn't test this in 2.5 yet, but I think you get the idea.
>
> Cheers,
> Simon
>
>



More information about the R-SIG-Mac mailing list