I am installing R 2.15.1 onto RHEL 6, using gcc 4.7.0 with Intel MKL
10.3.7 and the following environment:
export BLAS_LIBS="-Wl,--start-group
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_gf_lp64.a
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_sequential.a
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_core.a -Wl,--end-group
-lpthread"
export LAPACK_LIBS="-Wl,--start-group
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_gf_lp64.a
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_sequential.a
/usr/caen/intel-12.1/mkl/lib/intel64/libmkl_core.a -Wl,--end-group
-lpthread"
./configure --prefix=/home/software/rhel6/R/2.15.1 \
-with-x=no --with-blas --with-lapack --enable-R-shlib
Everything compiles OK, but 'make check' shows some differences. I
tried this first with gcc 4.7.0 and then again with gcc 4.4.6 (RHEL 6
installed). Same results.
There are differences for the package 'tools', and 'stats'. The
first, perhaps, I could ignore because of a change in format, or which
packages are getting installed with recommended, but the one for
'stats' seems not benign because it reverses the sign of all the
loadings for the third extracted principal component. I would think
it reasonable that the Grand Means would agree to four decimal places,
also, no?
Might anyone have a notion why I am seeing these differences?
Copious output follows.
Thanks, -- bennet
Here are the relevant lines copied from the 'make check':
Testing examples for package ?tools?
comparing ?tools-Ex.Rout? to ?tools-Ex.Rout.save? ...
files differ in number of lines:
811a812,828