Skip to content

Difficulty with R CMD INSTALL

6 messages · Frank E Harrell Jr, Paul Gilbert, Brian Ripley +1 more

Original
Frank E Harrell Jr
# 
I am running Linux Mandrake 7.2 with the following:

platform i586-pc-linux-gnu
arch     i586
os       linux-gnu
system   i586, linux-gnu
status
major    1
minor    2.3
year     2001
month    04
day      26
language R         

To build a package I ran R CMD build Hmisc, apparently with success:

* checking for file `Hmisc/DESCRIPTION' ... OK
* preparing `Hmisc':
* cleaning src
* checking whether `INDEX' is up-to-date ... NO
* use `--force' to overwrite the existing `INDEX'
* removing junk files
* building `Hmisc_1.0.tar.gz'

The source directory had a single large R source file under R
and many .Rd files under man.  It did not have a test directory,
and it have an extraneous s directory and an extraneous text
file in its root area.

Then I ran, under root, R CMD INSTALL Hmisc_1.0.tar.gz, and got:

Installing source package `Hmisc' ...
 libs
g77   -fPIC  -O3 -fomit-frame-pointer -pipe -s -march=pentium
-ffast-math -fexpensive-optimizations -fno-fast-math -c cidxcn.f -o
cidxcn.o
g77   -fPIC  -O3 -fomit-frame-pointer -pipe -s -march=pentium
-ffast-math -fexpensive-optimizations -fno-fast-math -c cidxcp.f -o
cidxcp.o
g77   -fPIC  -O3 -fomit-frame-pointer -pipe -s -march=pentium
-ffast-math -fexpensive-optimizations -fno-fast-math -c hoeffd.f -o
hoeffd.o
g77   -fPIC  -O3 -fomit-frame-pointer -pipe -s -march=pentium
-ffast-math -fexpensive-optimizations -fno-fast-math -c largrec.f -o
largrec.o
gcc -I/usr/lib/R/include  -I/usr/local/include -mieee-fp
-D__NO_MATH_INLINES  -fPIC  -O3 -fomit-frame-pointer -pipe -s
-march=pentium -ffast-math -fexpensive-optimizations -fno-fast-math -c
ranksort.c -o ranksort.o
g77   -fPIC  -O3 -fomit-frame-pointer -pipe -s -march=pentium
-ffast-math -fexpensive-optimizations -fno-fast-math -c rcorr.f -o
rcorr.o
gcc -shared  -o Hmisc.so cidxcn.o cidxcp.o hoeffd.o largrec.o ranksort.o
rcorr.o -L/usr/local/lib  -lg2c -lm
-L/usr/lib/gcc-lib/i586-mandrake-linux/2.95.3
-L/usr/i586-mandrake-linux/lib -lm
 R
                                
At this point the install program became stuck.  It had created
/tmp/R.INSTALL.xxxx/Hmisc OK, apparently, with a file
Hmisc.so in Hmisc/src that appears valid, and several .o files.

I would appreciate advice on how to best debug this.
Thanks -Frank
Brian Ripley
#
On Sun, 3 Jun 2001 fharrell at virginia.edu wrote:

            

      
      
      
    

        
With what extension?

At that point in the install process we have

    if test -d R; then
      echo " R"
      mkdir -p ${lib}/${pkg}/R
      cat `ls R/*.[RSqrs] R/unix/*.[RSqrs] 2>/dev/null` \
        > ${lib}/${pkg}/R/${pkg}
    fi


This will get stuck if there is an R directory, but the files in it
do not have extensions R S q r or s.  See section 1.1.3 of `Writing R
extensions' for the rules.

That's a bug (getting stuck) fixed in R-devel and hence will be fixed in
1.3.0: from the CVS logs

revision 1.61
date: 2001/05/03 07:18:41;  author: ripley;  state: Exp;  lines: +6 -4
INSTALL failed if dir R was empty or had no suitable files


As a more general comment, R CMD build is intended to be the last stage of
the process, used after R CMD check (which checks that installation works).
When developing a package I tend to INSTALL it several times and run a few
tests each time, then use R CMD check --no-clean to refine it (not
re-installing from scratch each time), and finally R CMD check then R CMD
build.

Brian

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Paul Gilbert
    

    

      
      #
    

  


  

      
Brian

BDR>When developing a package I tend to INSTALL it several times and run
a few
BDR>tests each time, then use R CMD check --no-clean to refine it (not
BDR>re-installing from scratch each time), and finally R CMD check then
R CMD
BDR>build.

I've been assuming the logic was to run R CMD check  first and then R
CMD INSTALL.   I don't think check uses the installed version, so I
guess it doesn't matter, but is there some other reason for doing
install before check? (Or are you just distinguishing development mode
from the normal mode once a package is developed?)

Paul Gilbert

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Mon, 4 Jun 2001, Paul Gilbert wrote:

        

            

      
      
      
    

        
I usually want to check things run before I worry that the documentation
is correct, and R CMD check starts with lots of checks that are irrelevant
until the code parses and runs some simple examples.

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    
1 day later

    

      
      

        


  

        

      Harry Mangalam
    

    

      
      #
    

  


  

      
Hi All,

We're having some problems with the Mac OSX port of R V1.2.3 binary that came from CRAN in that a (Perl) scripted installation of the xgobi library halts unexpectedly.  The commandline that fails is:

R --silent INSTALL --library=/usr/local/genex/lib/R/library \
 /Users/dan/sw/src/GeneX-Server-1.0.3/CyberT-dist/xgobi_1.2-2.tar.gz

The comment from the affected user is:

  It still hangs just after "Installing source package 
  'xgobi'", then " R" appears on the next line (including the space).

I saw the thread on " Difficulty with R CMD INSTALL " and was wondering if others have had a similar problem on OSX.  

Using R 1.2.3 on x86 linux (RH & Debian), I've never seen this problem, so I'm wondering if it's a porting artefact.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Tue, 5 Jun 2001, Harry Mangalam wrote:

        

            

      
      
      
    

        
*That* happens if there are no Rq... files in the R directory.  Is tar
working correctly?  I would try untaring the package, checking that the
R/* files are there and then installing the unpacked version.

BTW, the hanging won't happen in the next release.