Building R with f2c - still needed?

Prof Brian Ripley <ripley at stats.ox.ac.uk> writes:

Does anyone have a need to use f2c rather than a Fortran compiler to build 
R?

It is yet one more thing to test, and as it only works on 32-bit platforms 
it is something that I will shortly no longer be able to test.

The only candidate that I can think of is the PDA scene, which R has
been just a little too hard to build for till now (Linux based Zaurus
excepted). For those platforms you could be stuck with only a C/C++
compiler, but probably also without all the rest of the toolchain, so
building will be a major headache anyway.

O__  ---- Peter Dalgaard             ?ster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907

Does anyone have a need to use f2c rather than a Fortran compiler to build 
R?

We used f2c up until a few months ago as a last line of defence against
crappy Fg77 code on m68k/arm (one or both, it changed at times).

I'd be hesitant about throwing it out, but it is of course your call.

Thanks, Dirk

Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison

Thank you both for the comments.  The problem is that we have a lot of 
special-casing for f2c in the configure code, and I think it is becoming 
unmaintainable.  (I've been doing some cleaning up after discovering 
several errors in little-used cases, e.g. a substitute for alloca was 
compiled iff --enable-R-shlib was not specified.)

A cleaner solution would be to use a wrapper for f2c as a pseudo-compiler. 
I'll test that, and if it works then start to strip the special-casing 
out.

On Tue, Dec 13, 2005 at 09:48:56PM +0000, Prof Brian Ripley wrote:

Does anyone have a need to use f2c rather than a Fortran compiler to build
R?

We used f2c up until a few months ago as a last line of defence against
crappy Fg77 code on m68k/arm (one or both, it changed at times).

I'd be hesitant about throwing it out, but it is of course your call.

Thanks, Dirk

-- 
Hell, there are no rules here - we're trying to accomplish something.
                                                 -- Thomas A. Edison

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