[R-SIG-Mac] Updating to Mavericks

[peter dalgaard]

On 31 Oct 2013, at 23:19 , Prof Brian Ripley <ripley at stats.ox.ac.uk> wrote:

> On 31/10/2013 14:57, peter dalgaard wrote:
>> Thanks Brian.
>> 
>> One more item: If building from SVN, you may need to run "svn upgrade" in the source directory (at top level).
>> 
>> I suppose that's a consequence of upgrading Xcode rather than Mavericks per se, but might as well mention it here.
> 
> Yes.  A quick poll of systems I have access to has svn 1.6.x on 10.7, 1.7.x on 10.8 and 10.9.  As AFAICS it is part of Xcode, it depends on the version you had installed and I believe those machines were all fully updated.

At any rate, "svn up" tells you what to do, the only slight confusion is that make claims that your source directory is not an svn checkout until you have upgraded it.

One further issue is that gnutar seems to have gone AWOL. We have, in the CRAN binary, 

/Library/Frameworks/R.framework/Resources/etc/Renviron:TAR=${TAR-'/usr/bin/gnutar'}

so this affects R CMD INSTALL --build. I suppose people might discover this when a teaching package has gotten archived at CRAN. (A colleague bumped into it with languageR.) Looks like the obvious workaround does work around it:

TAR=/usr/bin/tar R CMD etc.





>> Do we have any authoritative advice as to what should go into config.site? Mine seems to work unchanged, but I'm not sure I have a clue about what they are doing anymore.
> 
> I have
> 
> % cat config.site
> CC="clang -mtune=native"
> OBJC=$CC
> F77="gfortran-4.2 -arch x86_64 -mtune=native"
> FC=$F77
> CXX="clang++ -mtune=native"
> CFLAGS="-g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion"
> 
> Those are rather picky C flags, of course: intended to pick up issues with long vectors.
> 
> gcc/g++ are not the same as clang/clang++, although the only differences I have detected are what they report themselves as.  As I don't know if there are any material differences I chose to use clang explicitly. Similarly calling gfortran-4.2 explicitly to avoid any other versions (and I would even consider using /usr/local/bin/gfortran-4.2).
> 
>> -pd
>> 
>> On 30 Oct 2013, at 15:38 , Prof Brian Ripley <ripley at stats.ox.ac.uk> wrote:
>> 
>>> This is an attempt to collect together various pieces of advice.
>>> 
>>> If you update to Mavericks and want to compile packages (or run some) with the CRAN binary R, you will need to
>>> 
>>> - Update R to 3.0.2 or later if you use R.app.
>>> 
>>> - Re-install XQuartz.  The Mavericks update re-populates /usr/X11 with links which tell you to install XQuartz (even if it is currently installed).
>>> 
>>> - Re-install the Apple Java 6 runtime if you use rJava.  Try any of the rJava examples and you will be prompted for an install.
>>> 
>>> - Install Xcode 5.0.1 if you had not previously done so.
>>> 
>>> - Re-install the Xcode command-line tools.  Xcode for Mavericks has moved most things inside Xcode (specifically under /Applications/Xcode.app/Contents/Developer) and the command-line tools previously installed seemed almost to work, but not entirely.
>>> 
>>> xcode-select --install
>>> 
>>> seems to be the recommended way to do so under Mavericks.
>>> 
>>> 
>>> If you previously had them installed, the Xcode 4.6.3 compilers (such as llvm-g++-4.2) should still work.  Otherwise you could select clang as your compiler: see http://cran.r-project.org/doc/manuals/r-release/R-admin.html#OS-X-packages . There are also compilers called 'gcc' and 'g++' in Xcode 5.0.1: these are clang-based but are not quite the same as clang/clang++.  Although g++ reports
>>> 
>>> % g++ --version
>>> Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
>>> 
>>> it looks at the libcxx headers at
>>> /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../lib/c++/v1/ .  And the packages which do not compile under clang++ do not compile under g++ .
>>> 
>>> Packages using a C++ interface may need re-compiling if you use clang++ (or g++) as your C++ compiler.  The external software at http://r.research.att.com/libs/ which I know does is gdal and zeromq. It would be prudent to compile Rcpp with the same compiler as a Rcpp-using package, although not always necessary.
>>> 
>>> 
>>> --
>>> Brian D. Ripley,                  ripley at stats.ox.ac.uk
>>> Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
>>> University of Oxford,             Tel:  +44 1865 272861 (self)
>>> 1 South Parks Road,                     +44 1865 272866 (PA)
>>> Oxford OX1 3TG, UK                Fax:  +44 1865 272595
>>> 
>>> _______________________________________________
>>> R-SIG-Mac mailing list
>>> R-SIG-Mac at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>> 
> 
> 
> -- 
> Brian D. Ripley,                  ripley at stats.ox.ac.uk
> Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
> University of Oxford,             Tel:  +44 1865 272861 (self)
> 1 South Parks Road,                     +44 1865 272866 (PA)
> Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-- 
Peter Dalgaard, Professor
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com