[Rd] R-Forge SVN repositories: R CMD build/check error on Windows machines

[Prof Brian Ripley]

The difference is in INSTALL, not build/check.

You are right that the Unix INSTALL was changed in r25808 (Aug 2003), but 
AFAICS this was not documented at the time in [O]NEWS, nor anywhere else.

Can you point me to the documentation you used to implement this?


On Mon, 28 Apr 2008, Duncan Murdoch wrote:

> On 28/04/2008 12:22 PM, Stefan Theussl wrote:
>> Dear R-devel,
>> 
>> One of our R-Forge developers pointed out that it is not possible to build 
>> packages under Windows using the R-Forge repository structure: a package 
>> resides in ./pkg - not in a directory with the same name as the package 
>> name.
>> 
>> Under Linux 'R CMD build pkg' or 'R CMD check pkg' work pretty well (I 
>> think Kurt Hornik fixed that in R 2.5.1 or so) whereas under Windows one 
>> gets the following error (this is the example sent by the user):
>> 
>> c:\work\packages\spdep>R CMD check pkg
>> * checking for working pdflatex ... OK
>> * using log directory 'C:/work/packages/spdep/pkg.Rcheck'
>> * using R version 2.7.0 (2008-04-22)
>> * using session charset: ISO8859-1
>> * checking for file 'pkg/DESCRIPTION' ... OK
>> * this is package 'spdep' version '0.4-21'
>> * package encoding: latin1
>> * checking package name space information ... OK
>> * checking package dependencies ... OK
>> * checking if this is a source package ... WARNING
>> Subdirectory 'pkg/src' contains object files.
>> * checking whether package 'spdep' can be installed ... ERROR
>> Installation failed.
>> See 'C:/work/packages/spdep/pkg.Rcheck/00install.out' for details.
>> 
>> which is:
>> 
>> installing R.css in C:/work/packages/spdep/pkg.Rcheck
>> 
>> 
>> ---------- Making package pkg ------------
>>   adding build stamp to DESCRIPTION
>>   installing NAMESPACE file and metadata
>>   making DLL ...
>>   ... DLL made
>>   installing DLL
>>   installing R files
>>   installing inst files
>>   installing data files
>>   preparing package pkg for lazy loading
>> Loading required package: tripack
>> Loading required package: sp
>> ...
>> Error in findpack(package, lib.loc) : *there is no package called 'pkg'*
>> Calls: <Anonymous> -> findpack
>> Execution halted
>> make[2]: *** [lazyload] Error 1
>> make[1]: *** [all] Error 2
>> make: *** [pkg-pkg] Error 2
>> *** Installation of pkg failed ***
>> 
>> I could verify this on our 'Windows package building machine' not only for 
>> this package but also for others.
>> Therefore, it seems to me that the (Windows) R CMD build/check scripts are 
>> not considering the package name in the DESCRIPTION file but rather take 
>> the directory name as package name.
>> 
>> Or are we just doing something completely wrong?
>
> You're right, on Windows there's an assumption that package foo is in 
> directory foo.  But I don't see why this is a big problem.  Can't you just 
> check out pkg into foo, e.g.
>
> svn co svn://svn.r-forge.r-project.org/svnroot/foo/pkg foo
>
> (Not that I'd be against accepting a patch to remove this restriction; 
> occasionally it might be nice to have two versions of the same package 
> side-by-side.)
>
> Duncan Murdoch
>
>> 
>> I used R-patched 2.7.0 on R-Forge to reproduce this error.
>> 
>> Best regards,
>> Stefan
>> 
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595