Skip to content

repeated - R package - Compilation Error

4 messages · Rick Bilonick, Martin Maechler, Brian Ripley

Rick Bilonick
# 
I'm not sure what your point is. I'm getting a compilation error for a
package that should compile without errors. The error message doesn't say
anything about needing anything - it doesn't complain about
"dependencies." Now once I got repeated to compile, it did give a
*warning" message about needing rmutils. But rmutils won't compile and
gives several error messages (in my other post). I've installed many R
packages and I've never seen problems like this before.

Rick B.
9 days later
Martin Maechler
# 
.................

    rab45> I'm not sure what your point is. I'm getting a compilation error for a
    rab45> package that should compile without errors. The error message doesn't say
    rab45> anything about needing anything - it doesn't complain about
    rab45> "dependencies." Now once I got repeated to compile, it did give a
    rab45> *warning" message about needing rmutils. But rmutils won't compile and
    rab45> gives several error messages (in my other post). I've installed many R
    rab45> packages and I've never seen problems like this before.

well, probably the others where CRAN or bioconductor packages ?

You know, there *is* a reason why we (actually, the repository
maintainers) require that a package runs  "R CMD check" without
a warning.  
If package authors are not willing to clean up and document
their code well enough to be accepted by CRAN,  you have to
expect hardship when trying to install / use the package....

Martin Maechler
Rick Bilonick
# 
I think the real problem is that Fedora Core 4 has switched from f77 to
gfortran.

Rick B.
Brian Ripley
#
On Sun, 21 Aug 2005 rab45+ at pitt.edu wrote:

            

      
      
      
    

        
The REAL problem is that some packages are not written in standard 
Fortran.  Compare the issues you found with the Fortran standards (which 
can be found via http://developer.r-project.org/Portability.html).

Martin's point is that some of the R volunteers put a lot of effort into 
helping (cooperative) package authors avoid such portability problems, and 
that you should be grateful (and give due credit) for such work.

For further discussion see recent threads on the r-devel list, where this 
sort of thing belongs (do check the posting guide).