[R-sig-ME] Workaround for my trouble installing lme4a

[Douglas Bates]

On Thu, Apr 15, 2010 at 9:24 AM, Kevin E. Thorpe
<kevin.thorpe at utoronto.ca> wrote:
> Hello, since my difficulties installing lme4a are buried in another thread,
> I'm creating a new thread more relevant to the problem.  In working through
> this email, I did find a solution (at the bottom), but I still have a
> question after that.
>
> I have fetched the entire project tree using subversion.  Here is my
> sessionInfo.
>
>> sessionInfo()
> R version 2.10.1 Patched (2009-12-29 r50852)
> i686-pc-linux-gnu
>
> locale:
>  [1] LC_CTYPE=en_US       LC_NUMERIC=C         LC_TIME=en_US
>  [4] LC_COLLATE=C         LC_MONETARY=C        LC_MESSAGES=en_US
>  [7] LC_PAPER=en_US       LC_NAME=C            LC_ADDRESS=C
> [10] LC_TELEPHONE=C       LC_MEASUREMENT=en_US LC_IDENTIFICATION=C
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> loaded via a namespace (and not attached):
> [1] tools_2.10.1
>
> I have installed all of the dependencies in lme4a/DESCRIPTION, including
> Rcpp and the correct version of Matrix.  The command
> 'R CMD INSTALL lme4a' fails when it gets to the following instruction:
>
> g++ -I/usr/local/lib/R/include  -I/usr/local/include
> -I"/usr/local/lib/R/library/Matrix/include"
> -I"/usr/local/lib/R/library/Rcpp/include"   -fpic  -g -O2 -c glmFamily.cpp
> -o glmFamily.o
>
> One thing I notice is that the one of the -I switches does not match how my
> system is installed (I have not done anything special in my R installation).
>
> Specifically, my version of Rcpp (0.7.11) does not have an include
> directory.  The header files appear to be in
> /usr/local/lib/R/library/Rcpp/lib
>
> I created a symbolic link for the missing include directory and successfully
> built lme4a.
>
> My question is, how are the -I switches constructed?  Are they hard-coded in
> the package source somewhere?  Did something go wrong in my Rcpp install?

The -I switches were constructed from code in lme4a/src/Makevars that
called Rscript and was difficult to maintain.  We (well, primarily
Dirk and Romain) made a change in where the include files for Rcpp
were stored so that the LinkingTo specification in the DESCRIPTION
file is all that is necessary for the includes.

Obviously I have been testing lme4a against the development version of
Rcpp and not against the released version.  Dirk is suggesting a 0.8.0
release of Rcpp, after which I will bump up the dependence of lme4a on
Rcpp.  Short term you could install the development version of Rcpp,
also on R-forge.