I never got your initial bug report
Package: r-core-case
Version: 2.13.1-1
Severity: wishlist
Hello,
The Debian packages for R add a "/usr/local/lib/R/site-library"
directory into which the local sysadmin can install packages, which is
really useful. However, compiled R packages are not always compatible
between versions of R (e.g. going from 2.9 to 2.13), and not having a R
version number in that directory path makes migration to a new R version
without breaking anything a bit complicated, at least for sites which
have multiple machines accessing a common /usr/local via NFS, etc.
Would it be possible to also add a "/usr/local/lib/R/%v/site-library" at
the front of R_LIBS_SITE? That way local packages could be installed in
that directory and there wouldn't be any problems when changing versions
or R, or running multiple versions on different computers, etc. The "%v"
is used in other places already (packages installed in a user's home
directory, for one), and given that the R ABI can change between
version, I was a bit surprised that said "%v" wasn't used for the
/usr/local path too. (Actually, I initially thought that R didn't change
its ABI for modules.)
Thanks,
Christian
as the package is 'r-base-core' -- the typo prevented delivery.
The idea is not bad but I am not going to address this, and decided to close
the report rather than to let it linger:
i) R _usually_ remain compatible across versions, the 2.9 to 2.10 upgrade
was the one exception in the last few years.
ii) We do not have support for several R versions anyway. The suggestion
sometimes comes up to support 'R-devel' as well as the current release
but even if we did that, as per i) there is so little difference / need
for packages to have distinct installation paths...
iii) We'd have to propagete the .libPaths() reliably across user settings
which, while probably possible, is likely to just introduce another bug
here or there.
So in retrospect, I'd prefer things to stay as they are.
Thanks for the report and suggestion though.
Dirk
--
New Rcpp master class for R and C++ integration is scheduled for
San Francisco (Oct 8), more details / reg.info available at
http://www.revolutionanalytics.com/products/training/public/rcpp-master-class.php
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
