R help - Jan 2008 - Fortran level mismatch between package and R

We have a user here who is attempting to compile a package which  
requires f90 (gfortran will do). Our version of R was built with f77  
(g77). When the user attempts to build the package, f77 is always  
invoked and fails due to the presence of f90 features.

Is there a simple method or some documentation on how to force a  
package build to use a different compiler than the one used in  
building R? If so, are opening ourselves up to any problems? In  
particular, we're concerned about API mismatches and fortran library  
mismatches, but there may be other problems we've not considered.

We're using R 2.4.1, linux 2.6/x64, gcc/gfortran 4.0.

To pre-answer some possible questions/suggestions -

Doing a full conversion of our R installation is almost certainly not  
feasible; we also have legacy packages in place which don't build  
with f90/f95.

Yes, we've thought of installing R built with gfortran and retaining  
the old with f77; that's probably our first fallback. Unfortunately  
that shuts legac package users from R improvements and will shut them  
out from packages which start require R2.6 and up features. We do  
*not* wish to run a f77 version of R2.6 and a f95 version of R2.6.  
That's probably our last fallback.

Neither Suleman nor I are on the r-help mailing lists; please try to  
cc us on responses. We'll use the maillist archives to try and catch  
the responses which don't do that.

Advance thanks,

Steve Simmons

Hi,

I recompiled R using gfortran in my personal laptop. But I think it
may help if you build an agent C program which calls the FORTRAN
program, and then call the C program from R.

I'm also curious on how to solve this problem without too much work.
Please let me know if you find a good solution.

On Jan 19, 2008 12:58 AM, Steve Simmons <scs at umich.edu>
wrote:
> We have a user here who is attempting to compile a package which
> requires f90 (gfortran will do). Our version of R was built with f77
> (g77). When the user attempts to build the package, f77 is always
> invoked and fails due to the presence of f90 features.
>
> Is there a simple method or some documentation on how to force a
> package build to use a different compiler than the one used in
> building R? If so, are opening ourselves up to any problems? In
> particular, we're concerned about API mismatches and fortran library
> mismatches, but there may be other problems we've not considered.
>
> We're using R 2.4.1, linux 2.6/x64, gcc/gfortran 4.0.
>
> To pre-answer some possible questions/suggestions -
>
> Doing a full conversion of our R installation is almost certainly not
> feasible; we also have legacy packages in place which don't build
> with f90/f95.
>
> Yes, we've thought of installing R built with gfortran and retaining
> the old with f77; that's probably our first fallback. Unfortunately
> that shuts legac package users from R improvements and will shut them
> out from packages which start require R2.6 and up features. We do
> *not* wish to run a f77 version of R2.6 and a f95 version of R2.6.
> That's probably our last fallback.
>
> Neither Suleman nor I are on the r-help mailing lists; please try to
> cc us on responses. We'll use the maillist archives to try and catch
> the responses which don't do that.
>
> Advance thanks,
>
> Steve Simmons
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
???
Hesen Peng