Skip to content

R.matlab

4 messages · Anestis Antoniadis, Henrik Bengtsson, stefano iacus

#
[Sorry if this is a repost, but I got an odd error message previously trying
to reply to 'r-sig-mac-request@stat.math.ethz.ch']

Hi. I (the author) got a similar problem report two weeks ago and it was
also from a person running R v2.0.0 on Mac OS X. I do not have access to
this system so unfortunately I cannot troubleshoot it myself. 

The solution that worked for this person was to download the bundle manually

http://www.maths.lth.se/help/R/R.classes_0.55.tar.gz

and run

R CMD INSTALL R.classes_0.55.tar.gz

from the system's command prompt.

Does this work? 

If it does, it is a good starting point for troubleshooting. The problem
seems to be that R.oo is not available (==installed) when it tries to
install R.audio, which depends on R.oo. I'm not 100% sure from your log
(below), if R tried (failed, succeeded?) to install the R.oo package first,
and then tried to install R.audio (your error message), or if it starts with
R.audio first. 

Have someone else on the R-SIG-Mac list experienced this? Try

install.packages("R.classes", contriburl="http://www.maths.lth.se/help/R")

Currently, it seems to me that this is a OSX problem only (based on sample
size n=2). If not, I will check with the r-devel list.

Best wishes

Henrik Bengtsson
#
Dear Henrik,
dear all,
same problem for me on Mac OSX R v 2.0.0.
However it works with Jan de Leeuw version.
Stefano Iacus is aware of the problem.
Best
Anestis
On Nov 5, 2004, at 11:31 AM, Henrik Bengtsson wrote:

            
====================================================
Prof. Anestis ANTONIADIS
Laboratoire de Modelisation et Calcul(LMC-IMAG)
Universite Joseph Fourier
Tour IRMA
B.P.53
38041 Grenoble CEDEX 9
FRANCE

Tel:(33) (0) 476514306
Fax:(33) (0) 476631263
#
Thank you. I understand from your reply that there are a few MacOSX versions
maintained (or similar) and that the one maintained by Stefano Iacus is the
one giving the problem. 

To Stefano Iacus, if you not already know, I suspect that it has to do with
the order the packages are installed. Under "Packages bundles"
(http://cran.r-project.org/doc/manuals/R-exts.html#Package%20bundles) it
says

"The Contains field [in DESCRIPTION] lists the packages, which should be
contained in separate subdirectories with the names given. During building
and installation, packages will be installed in the order specified. Be sure
to order this list so that dependencies are appropriately met."

With the introduction of lazy loading this is important. Before, it did not
seem to be a requirement in practice. Is this order strictly followed by the
MacOSX installation scripts?

Just a though.

Henrik
#
It seems like the problem is in the fact that if R_LIBS is not defined  
the aqua GUI does not define it.
By default, the aqua GUI uses ~user_home/R/library and R.oo is in fact  
installed there (temporarily, then removed on failing).
One temporary solution is the following:
1. authenticate as root from within the Aqua GUI (i.e. click on the  
"lock")
2.  
install.packages("R.classes",contriburl="http://www.maths.lth.se/help/ 
R",lib=.libPaths[2])
this will install the bundle inside the framework.

and I have to figure out how to pass the lib argument to the script.

For some reason, you don't see the output of the package installation  
due to lack of synchronization when you run a script as a root.

FYI: the aqua GUI uses the same install.packages as the console  
version. The problem is that R CMD INSTALL (which is called by that  
function) is not aware of R_LIBS env var which is to the contrary  
needed in this bundle case.

stefano
On Nov 5, 2004, at 1:48 PM, Henrik Bengtsson wrote: