R devel - Nov 2009 - Problem building package for R 2.10.0 on Mac OS X

Hi

I have submitted a package (rioja) to CRAN.  It checks OK for all R versions and
OS's except r-release-macosx-ix86 where it fails when checking the examples.
Specifically, it fails because R can't find the package vegan which is
needed in a function. Here is the snippet from the check results:

### Begin snippet

checking examples ... ERROR
Running examples in 'rioja-Ex.R' failed.
The error most likely occurred in:
> ### * chclust
>
> flush(stderr()); flush(stdout())
>
> ### Name: chclust
> ### Title: Constrained hierarchical clustering
> ### Aliases: chclust plot.chclust bstick.chclust
> ### Keywords: cluster hplot
>
> ### ** Examples
>
> data(RLGH)
> diss <- dist(sqrt(RLGH$spec/100)^2)
> clust <- chclust(diss)
> bstick.chclust(clust, 10)
Loading required package: vegan
Warning in library(package, lib.loc = lib.loc, character.only = TRUE,
logical.return = TRUE, :
there is no package called 'vegan'
Error in bstick.chclust(clust, 10) : could not find function "bstick"
Execution halted

### End snippet

The function bstick.chclust listed below loads the package vegan which contains
bstick.default.

bstick.chclust <- function(n, ng=10, plot=TRUE, ...) {
   if (n$method != "coniss")
      stop("bstick cannot display conslink results")
   require(vegan)
   disp <- rev(n$height)
   tot.disp <- disp[1]
   disp <- abs(diff(disp))
   nobj <- length(n$height)
   bs <- bstick(nobj, tot.disp)
   yR <- range(disp[1:(ng-1)], bs[1:(ng-1)])
   if (plot) {
      plot(2:ng, disp[1:(ng-1)], type="o", ylim=yR, ylab="Sum of
Squares", xlab = "Number of groups")
      lines(2:ng, bs[1:(ng-1)], col="red", type="o")
   }
   invisible(data.frame(nGroups = 2:(ng), dispersion=disp[1:(ng-1)], bstick =
bs[1:(ng-1)]))
}

vegan is included as a Suggests in the DESCRIPTION file and the package checks
OK on r-oldrel-macosx-ix86.  It also checks OK with R 2.10.0 on OS X on my own
Mac.  My code could be improved and I should probably trap the warning issued by
require, but I don't understand why vegan is not loading when checking on
CRAN.  Any help will be gratefully received!

Many thanks, Steve

Steve Juggins
School of Geography, Politics & Sociology
Newcastle University   Tel: +44 (0)191 222 8799
Newcastle upon Tyne    Fax: +44 (0)191 222 5421
NE1 7RU, UK            Mobile:   +44 (0)7740054905
http://www.staff.ncl.ac.uk/staff/stephen.juggins/

Please ask such questions on R-sig-mac, or directly to the Mac package 
builder (Simon Urbanek).

As you will see from

http://cran.r-project.org/web/checks/check_summary.html

at least 150 packages are in the same boat (not building because 
dependencies are missing), so something has gone wrong on the build 
machine.  There seem to be results for 'vegan', but they are for a 
pre-release (2.10.0 alpha), not the released version.

Perhaps related, there have been no R-patched builds on 
r.research.att.com for 'leopard' (the main R build for R 2.10.0) since 
Nov 4.

Simon has been travelling, and will no doubt get to this when he has 
time.  Meanwhile, Mac users can fairly easily build packages from the 
sources (provided they have installed Xcode), so this is not something 
to fret about.

One point specific to Mac is that the distributed packages are 
tri-architecture (i386, x86_64, ppc) and so need to work on all three 
archs to get distributed (at least in principle).  There have been 
quite a few packages (including earlier versions of yours) that failed 
on 64-bit platforms.

On Sun, 15 Nov 2009, Stephen Juggins wrote:
> Hi
>
> I have submitted a package (rioja) to CRAN.  It checks OK for all R
versions and OS's except r-release-macosx-ix86 where it fails when checking
the examples. Specifically, it fails because R can't find the package vegan
which is needed in a function. Here is the snippet from the check results:
>
> ### Begin snippet
>
> checking examples ... ERROR
> Running examples in 'rioja-Ex.R' failed.
> The error most likely occurred in:
>
>> ### * chclust
>>
>> flush(stderr()); flush(stdout())
>>
>> ### Name: chclust
>> ### Title: Constrained hierarchical clustering
>> ### Aliases: chclust plot.chclust bstick.chclust
>> ### Keywords: cluster hplot
>>
>> ### ** Examples
>>
>> data(RLGH)
>> diss <- dist(sqrt(RLGH$spec/100)^2)
>> clust <- chclust(diss)
>> bstick.chclust(clust, 10)
> Loading required package: vegan
> Warning in library(package, lib.loc = lib.loc, character.only = TRUE,
logical.return = TRUE, :
> there is no package called 'vegan'
> Error in bstick.chclust(clust, 10) : could not find function
"bstick"
> Execution halted
>
> ### End snippet
>
> The function bstick.chclust listed below loads the package vegan which
contains bstick.default.
>
> bstick.chclust <- function(n, ng=10, plot=TRUE, ...) {
>   if (n$method != "coniss")
>      stop("bstick cannot display conslink results")
>   require(vegan)
>   disp <- rev(n$height)
>   tot.disp <- disp[1]
>   disp <- abs(diff(disp))
>   nobj <- length(n$height)
>   bs <- bstick(nobj, tot.disp)
>   yR <- range(disp[1:(ng-1)], bs[1:(ng-1)])
>   if (plot) {
>      plot(2:ng, disp[1:(ng-1)], type="o", ylim=yR, ylab="Sum
of Squares", xlab = "Number of groups")
>      lines(2:ng, bs[1:(ng-1)], col="red", type="o")
>   }
>   invisible(data.frame(nGroups = 2:(ng), dispersion=disp[1:(ng-1)], bstick
= bs[1:(ng-1)]))
> }
>
> vegan is included as a Suggests in the DESCRIPTION file and the 
> package checks OK on r-oldrel-macosx-ix86.  It also checks OK with R 
> 2.10.0 on OS X on my own Mac.  My code could be improved and I 
> should probably trap the warning issued by require, but I don't 
> understand why vegan is not loading when checking on CRAN.  Any help 
> will be gratefully received!
>
> Many thanks, Steve
>
> Steve Juggins
> School of Geography, Politics & Sociology
> Newcastle University   Tel: +44 (0)191 222 8799
> Newcastle upon Tyne    Fax: +44 (0)191 222 5421
> NE1 7RU, UK            Mobile:   +44 (0)7740054905
> http://www.staff.ncl.ac.uk/staff/stephen.juggins/
>
> ______________________________________________
> 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