Created attachment 337113 [details]
Patch for src/scripts/R.sh.in (found in R-2.8.1-7.fc10.src.rpm in the expanded R-2.8.1.tar.gz)

Description of problem:
R-core-2.8.1-7.fc10.x86_64 was installed by yum-updatesd. There are new files:
/etc/profile.d/R.sh
/etc/profile.d/R.csh

If environment variable R_HOME is already set to a value other than the value used by R, then `R RHOME` produces a warning:

$ R RHOME
WARNING: ignoring environment value of R_HOME
/usr/lib64/R

The result is that the setting of R_HOME in /etc/profile.d/R.sh and /etc/profile.d/R.csh gets a wrong value (in case of bash) or produces a syntax error from setenv (in case of tcsh (csh)).
In case of tcsh (csh) the syntax error stops the execution of the remaining init scripts, which results in an incomplete environment and settings of the tcsh (login) shell.

Here are some results.
Note that the value of R_HOME is
"WARNING: ignoring environment value of R_HOME /usr/lib64/R"
in the error case, and in tcsh
"setenv R_HOME WARNING: ignoring environment value of R_HOME /usr/lib64/R"
procudes a syntax error.

$ bash
$ export R_HOME=
$ bash
$ echo $R_HOME
/usr/lib64/R
$ exit
$ export R_HOME=/usr/local/lib/R
$ bash
$ echo $R_HOME
WARNING: ignoring environment value of R_HOME /usr/lib64/R
$ exit
$ export R_HOME=/usr/lib64/R    
$ bash
$ echo $R_HOME
/usr/lib64/R
$ exit
$ tcsh
tcsh$ exit
exit
$ export R_HOME=/usr/local/lib/R
$ tcsh
setenv: Too many arguments.
tcsh-incomplete $ exit


How could this be solved?

1. The init scripts /etc/profile.d/R.sh and /etc/profile.d/R.csh
   should not set R_HOME is it is already set
   because it may be that the user has set it on his own.

2. R should write warning messages to standard error and not to standard output.
   The waring message listed above is written to standard output.
   Then `R RHOME` returns only the path and not also the error message.


Version-Release number of selected component (if applicable):
R-core-2.8.1-7.fc10.x86_64

How reproducible:


Steps to Reproduce: See above.

Additional info:
This bug breaks login environment with tcsh as login shell and R_HOME already set.
This is the reason why I declare the severity as high
and why this bug should be fixed very soon.
I provide patches which can be applied to the rpm sources
so a new package can be build and released very soon.

For the future I request the developers to consider and to test all possible standard cases (like the case that a variable is already set) before releasing an script (such as an init script) to the public.
Thanks in advance!