[R-SIG-Mac] 64- and 32-bit versions of R

[Prof Brian Ripley]

Some recent postings have shown some common confusions, so this posting 
is an attempt to clarify.

Recent CRAN binaries of R are x86_64 (64-bit) and i686 (32-bit), and you 
can run either on all but the very oldest support Macs running Leopard 
(any machine since mid-2008, I believe, and Leopard is well past 
end-of-life).

- There are separate installation of R.app called R.app (32-bit) and 
R64.app (64-bit) if a standard install is used.  When updating they are 
supplied separately, both as R.app.

- Command-line R is installed under /usr/bin as R, R32, R64 and Rscript. 
  The idea was that 'R' is R32 on Leopard and R64 on later, but an 
installer script bug made 'R' R32 on Mountain Lion up to 2.15.2 (fixed 
in the nightly builds).   I have gotten into the habit of always using 
R64 (or R32).

R --arch=x64_64 and R --arch=i386 are equivalent to R64 and R32 
respectively.  Note that these come *before* any arguments such as CMD 
INSTALL.

When you install a package from sources, it will usually try to install 
the package for all the installed sub-architectures, that is both x86_64 
and i386 for CRAN binaries.  There are exceptions, including if the 
package has a configure script (as e.g. rgl has).  If you only ever use 
64-bit R, then we suggest you install packages by

R64 CMD INSTALL --no-multiarch mypkg_x.y.tar.gz

Increasingly software for OS X is 64-bit only.  The most pressing 
problem is the Java runtime where security concerns dictate using a very 
recent version from Oracle.  So in a security-conscious environment, 
rJava is 64-bit only.


Future note:
-----------

The intention is that for 3.0.0 due April 3, CRAN builds will require 
Snow Leopard or later and be 64-bit only.  If there is any very good 
reason to maintain 32-bit support in the CRAN binaries please let us 
know.  (I happily ran 64-bit R on a 2008-vintage 2GB MacBook Air until 
last Easter, so 64-bit R does run on minimal hardware.)

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595