[R] RSPerl on OS X Server 10.4.11

[Gregory Downs]

Hello,

I tried the R-Sig-Mac list with this query, but had no takers.  I hope 
that the following isn't too far off the mark for this list.  Many 
thanks in advance if someone can help me out!

Quick summary:  I can't get RSPerl working on a PPC G5 with the 
pre-compiled binary for Mac OS X, but it does work if I compile R from 
source with the "--enable-R-shlib" flag.  If I need to compile from 
source...what other options are recommended?

More details...

I'm having some trouble getting RSPerl working on an XServe cluster.  Using:

perl v5.8.6
R v2.6.2 (installed from binaries from CRAN)
RSPerl v0.92-1
OS X Server 10.4.11, PowerPC G5

I'm installing RSPerl with:

R CMD INSTALL -c -l '/home/gdowns/R' --configure-args='--with-in-perl' 
RSPerl_0.92-1.tar.gz

but when I try to load the library I get:

 > library("RSPerl")
Error in dyn.load(file, ...) :
   unable to load shared library '/home/gdowns/R/RSPerl/libs/ppc/RSPerl.so':
   dlopen(/home/gdowns/R/RSPerl/libs/ppc/RSPerl.so, 6): Symbol not 
found: _boot_R
   Referenced from: /home/gdowns/R/RSPerl/libs/ppc/RSPerl.so
   Expected in: dynamic lookup

Error: package/namespace load failed for 'RSPerl'

So, I tried building R from scratch, using:

./configure --prefix=/tmp/R/ --enable-R-shlib

I then repeated the install of RSPerl with this special version of R:

/tmp/R/R.framework/Resources/bin/R  CMD INSTALL -c -l '/home/gdowns/R' 
--configure-args='--with-in-perl' RSPerl_0.92-1.tar.gz

RSPerl *does* work with this version of R.  I'm confused because 
question 1.3 of the RMacOSX-FAQ says:

"Note that the CRAN binary includes R shared library, so there is no 
need to re-compile R with --enable-R-shlib. "

I also note that when installing RSPerl with the binary distribution of 
R, the R CMD INSTALL process does not generate the perl modules (in the 
RSPerl/perl directory).  They do get created with the 
compiled-from-source version of R (which is, I suspect, why it works!).

I'm not sure where I'm going wrong with this.  I'd highly prefer to 
stick with the pre-compiled binary to make upgrading easier.  Plus I'm 
sure that the version I built today is not as well optimized as the 
binary distribution!

If I really do need to compile R from scratch, can anyone recommend the 
appropriate configure flags?

many thanks and best regards,
Greg.
-- 
Gregory Downs
CRSC Room 219/414
Department of Plant Agriculture
University of Guelph
Guelph, Ontario, Canada, N1G 2W1
519-824-4120 x58164