[R] Help executing R on High Performance Cluster

[Ivan Krylov]

Hi Eric!

В Tue, 4 Oct 2022 13:36:59 +0000
"Rhon Calderon, Eric" <Eric.RhonCalderon using pennmedicine.upenn.edu> пишет:

> cd /home/ericrhon/labsoftware/R-devel/

> ./configure --prefix=/home/ericrhon/labsoftware/R-devel

I'm not sure, but the reason for the confusing behaviour could be
that the installation prefix is set to be the same as the directory
where you had unpacked the source code. This may have caused file path
conflicts and maybe some kind of recursion when building the final
executable path.

R can run straight from the build directory, without installation.
If you start from scratch, does bin/R work without `make install`? If
not, do you get any useful output if you run bin/R under sh -x?

(Additionally, R supports building away from the source code directory.
For example, if you navigate to the source code directory, then run
mkdir build && cd build && ../configure $CONFIGURE_ARGUMENTS, the build
will be mostly self-contained to that directory, which may be useful
when experimenting.)

Why are you building R with --with-recommended-packages=no?

R is supposed to be easy to build and start working with, but there may
be important details to consider, especially on an HPC, in the "R
installation and administration" guide:
https://cran.r-project.org/doc/manuals/r-release/R-admin.html

> Here is the exec script of R, line 272 is highlighted:
<...>
>	[[alternative HTML version deleted]]

This mailing list strips the HTML version of the e-mails; here's what
we see: https://stat.ethz.ch/pipermail/r-help/2022-October/475956.html

One way to link to R source code is to use the GitHub mirror:
https://github.com/r-devel/r-svn

-- 
Best regards,
Ivan