The CRAN package check page now shows results on Sparc Solaris 10 using a server donated by Sun.

This box tests several different aspects from the existing Intel-based test boxes:

- the CPU is big-endian.

- it does not have extended-precision doubles in FPU registers. (It does support long doubles if selected explicitly.)

- the C/C++ header set is from a different tradition.

- the Sun Studio compiler is used (as has been on one of the Linux boxes, and we may phase the latter out in due course).

- the toolset is AT&T-based Unix rather than GNU or BSD. In particular Solaris make, sed and tar and an actual Bourne shell are used.

There is only a limited set of additional software installed. A few things have had to be compiled with gcc, notably the packages using Gtk+ (and Solaris' own installation of Gtk+ is too old for RGtk2) and MCMCpack.

The server does not have a graphics card and this is causing some problems with the X11 installation, including bitmap devices and rgl, that we are still working on.

Amongst the package issues shown up are

- unterminated final lines in src/Makevars[.in] cause failures.

- there are many packages which require GNU make and fail to declare it. These include Cairo JavaGD Rcpp RcppExamples RInside farmR highlight phyclust png rJava.

- the assumption that 'tar' accepts compressed archives (gdata).

- 60-odd C++ -using packages are not correct C++, often because they use GNU extensions or have missing headers (and C++ requires headers for prototypes), or use C99 functions that are not in C++ and hence not in the headers when used from C++ -- Solaris is rather strict about the latter.


Brian Ripley

--
Brian D. Ripley,                  [email protected]
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

______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

Reply via email to