[Rd] Problem with R CMD INSTALL and package versions (PR#2878)

rpeng at stat.ucla.edu rpeng at stat.ucla.edu
Wed Apr 30 09:23:14 MEST 2003

Full_Name: Roger Peng
Version: 1.7.0
OS: Linux (Red Hat 8.0)
Submission from: (NULL) (

There seems to be a problem with R CMD INSTALL using --with-package-versions. 
First, when using --save and --with-package-versions together, I get the
following output (in this case, for the `snow' package):


marla:> R CMD INSTALL --save --with-package-versions snow_0.1-1.tar.gz 
* Installing *source* package 'snow' as 'snow_0.1-1' ...
** R
** inst
** save image
cat: /home/rpeng/install/R-1.7.0/lib/R/library/snow_0.1-1/R/snow: No such file
or directory

mv: cannot stat `/home/rpeng/install/R-1.7.0/lib/R/library/snow_0.1-1/R/snow':
No such file or directory
** help
 >>> Building/Updating help pages for package 'snow'
     Formats: text html latex example 
  snow-cluster                      text    html    latex   example
  snow-internal                     text    html    latex
  snow-parallel                     text    html    latex   example
  snow-rand                         text    html    latex   example
  snow-startstop                    text    html    latex   example
* DONE (snow)



When trying to load the package in R, I get:

> library(snow, version = "0.1-1")
Error in .path.package("snow") : none of the packages are loaded
Error in library(snow, version = "0.1-1") : 
        .First.lib failed

Using install.packages() with `installWithVers = TRUE' produces the same
problem.  Removing the --with-package-versions flag makes the package install as
expected.  Removing the --save flag (but keeping the --with-package-versions
flag) gets rid of the complaint during the install process, but I get the same
error when trying to load the package in R.  

I haven't tried this with too many packages but I alwasy get the same behavior. 
I first noticed this when trying to install `pixmap' and `gpclib' packages with
versions since they are both installed using --save.


More information about the R-devel mailing list