[Rd] Install R 2.0 package on R 1.9.1

[Paul English]

On Tue, 5 Apr 2005, stefano iacus wrote:

>>  I'm getting this error which seems to be known issue:
>> 
>> library("quantreg")
>> Error in firstlib(which.lib.loc, package) :
>>         couldn't find function "lazyLoad"
>> In addition: Warning message:
>> package quantreg was built under R version 2.0.1
>> Error in library("quantreg") : .First.lib failed
>> 
>> However, on CRAN, the Mac OS X "1.9" directory is just a symlink to the Mac 
>> OS X "2.0" directory - there is no separate directory for 1.9 packages as 
>> best I can tell.
> this is not true,
> have a look at
> http://cran.r-project.org/bin/macosx/2.0
> http://cran.r-project.org/bin/macosx/1.9
> they are just different (and 1.9 has not been updated since oct 2004)

Aah, the mirrors are very slow, so I only tried one. cran.r-project 
doesn't show the problem, so now I've got a R 1.9 package - thanks!
http://cran.cnr.Berkeley.edu
http://www.bioconductor.org/CRAN
do not have the problem either.

but:

http://lib.stat.cmu.edu/R/CRAN/

has this problem. Even if you go directly to:

http://lib.stat.cmu.edu/R/CRAN/bin/macosx/1.9/

The index lists "2.0." The way I got to it was by going from the R 
homepage CRAN->http://lib.stat.cmu.edu->Mac OS X->1.9


>> I did also try building the source package, but that failed. I emailed the 
>> packager asking for help, but it can't hurt to ask here as well. The 
>> problem is that the package build can't find -lfrtbegin. On Mac OS X this 
>> comes from fink so it is installed in /sw. So I untarred the package and 
>> edited Makevars to tell the build to look at -L/sw/lib, but interestingly 
>> enough that only worked for one gcc line, and not the final linking stage. 
>> As you can see below, the final linking stage is also looking for 
>> /usr/local/lib/gcc/powerpc-apple-darwin6.6/3.4 which doesn't exist. I'm not 
>> sure where it is getting that from. I can only guess that autoconf or 
>> something is messed up on this machine:
<snip build errors>
>
> this problem comes from the fact that you have system configured differently 
> from how the R binaries has been built, so config files distributed with it 
> reflect these differences (in particular this folder is relative to the gcc 
> shipped with OS X 10.2, you should replace it with the one pointing to the 
> right gcc for your system)

That makes good sense. I see there is an Aqua version out now that my 
users might like.

> My suggestions is to build R from sources as well following the notes on the 
> Faq for Mac OS X or asking to R-sig-Mac.

I'm setting up 2.0 in parallel on a separate machine, so I will do that, 
and whenever my user is ready I'll upgrade her R install.

> In any case, you should give us more details on the version of R you are 
> using, the OS X version, etc to leave us be helpful.

OS X ver 10.3.8
R 1.9.1

Although you've already answered all my questions. :-)