Skip to content

R chek of package VR is broken (PR#1683)

6 messages · Brian Ripley, Peter Dalgaard, Dirk Eddelbuettel +1 more

Original
Brian Ripley
# 
That's not an error in the VR bundle (sic), but it is an error *by you* in
the usage of R CMD check (sic).  From `Writing R extensions'

<p>Using <code>R CMD check</code>, the R package checker, one can test whether
<em>source</em> R packages work correctly.

and the under-qualified pilot applied it to a *binary* installed package,
despite the emphasized warning.
On Mon, 17 Jun 2002 juergen@henge-ernst.de wrote:

            

      
      
      
    

        

  
    

      
      
    

  


  


    

    

      
      

        


  

        

      Peter Dalgaard
    

    

      
      #
    

  


  

      
ripley@stats.ox.ac.uk writes:

            

      
      
      
    

        
Or, to put it differently, what you *should* have done was

<install R + recommended packages>
tar xfz Recommended/VR_7.0-3.tar.gz 
R CMD check VR

(A couple of gotchas in that: You cannot check MASS by itself, only
the entire VR bundle, and you need the other packages installed or the
check chokes on library(nlme))

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Dirk Eddelbuettel
    

    

      
      #
    

  


  

      
On Mon, Jun 17, 2002 at 10:00:11PM +0200, ripley@stats.ox.ac.uk wrote:

            

      
      
      
    

        
I wonder if "R CMD check ... " could abort with an error message if it is
applied to a binary package?  
 
Dirk

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Mon, 17 Jun 2002, Dirk Eddelbuettel wrote:

        

            

      
      
      
    

        
I am not sure how simple that is.  It used to be hard to tell source and
binary packages apart.  What we can do is to test for Built: in the
DESCRIPTION file, which will cope with people who have built the package
recently.

The trouble with idiot-proofing is anticipating the idiot: as far as I am
aware this is a first.

In all other cases I tried the INSTALL part of R CMD check failed.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Kurt Hornik
    

    

      
      #
    

  


  

          

      
      
      
    

        
This was introduced in 1.4 I think (?) so it seems we can change the
heuristics now to rely on the correspondence Built: <-> binary.  I will
soon (once I have time again) change INSTALL accordingly.

            

      
      
      
    

        
We need better integration of INSTALL into check anyway.  But we can
have more idiot-proofing in check as well.  Something in the works is a
reorganization of INSTALL/check/build with something else needed for
creating front-matter stuff, and we also need to rethink the logic of
check (e.g., verify correctness of meta-information prior to trying to
install).

-k
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel 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-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Tue, 18 Jun 2002, Kurt Hornik wrote:

        

            

      
      
      
    

        
Not necessarily, as people do manually install packages.  I've added the
test for Built in R CMD check for 1.6.0, which should catch a lot of
misuses.