R 2.1 and gfortran

Fri, May 20, 2005 12:09 AM #

From the configure output of the latest devel snapshot:

The Fortran compiler in gcc-4.0 defaults to the name 'gfortran'.   
Other packages such as Octave seem to check for that.

Just to suggest that gfortran should be added to the list of possible  
compilers.

As a footnote I have built R 2.1 on MacOS X using a compiler built  
from very recent Apple cvs sources, with patches from the gfortran  
cvs code up to 18th May.  This binary passes all the tests.

Bill Northcott

Brian Ripley

Fri, May 20, 2005 7:30 AM #

The reasons why it is not there *are* in the R-admin manual.  This was
deliberate: we do not include known broken tools in the searches.

Once gcc-4.0.1 is released and commonly used it will be safe to add it,
but not before. (That will most likely be for R-2.2.x, not R-2.1.x.) For
example, the current gfortran in FC3 is pre-4.0.0 and does not work (and
cannot be made to work on AMD64), and Simon Urbanek found an Apple version
of 4.0.0 that did not work.

On Fri, 20 May 2005, Bill Northcott wrote:

Brian D. Ripley,                  ripley@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