Thanks Dirk! The answers from Romain and you show exactly how to make things work again. I wasn’t aware of the possibility to override the Makeconf via a new Makevars. I work with environment modules to be able to load different compilers and to test my package compilation with them. I can now produce the ~/.R/Makevars file during module load. That makes everything more comfortable than adding a CFLAGS to the R CMD INSTALL command.
Best SImon On 27 Oct 2013, at 15:08, Dirk Eddelbuettel <e...@debian.org> wrote: > > On 27 October 2013 at 10:49, Simon Zehnder wrote: > | I wonder though, why it uses in Clays case the llvm-g++-4.2 even if it is > not there anymore? > > That has been explained a number of times, eg in the different SO question > (eg even the one ref'ed by JJ is a duplicate). An earlier one is > > > http://stackoverflow.com/questions/19503995/error-when-with-xcode-5-0-and-rcpp-command-line-tools-are-installed > > In short, R "writes down" the values of CC, CXX, CFLAGS, .. seen at _its_ > compile time and uses it each time R CMD ... is invoked. > > That works, as on most sane operating systems you assume that 'cc' or 'gcc' > will always point to a working compiler. OS X is, ahem, different. And so > it breaks, at least until the default R build catches up with the tool chain. > > Really a question for r-sig-mac and the OS X users here to sort out. > > Dirk > > -- > Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com _______________________________________________ Rcpp-devel mailing list Rcpp-devel@lists.r-forge.r-project.org https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
