[Rcpp-devel] Package checking using R-devel (3.1) and Rcpp

[Dirk Eddelbuettel]

Hi Robert.

On 24 September 2013 at 13:21, Robert Bruggner wrote:
| Hi Rcpp-devel,
| 
| I'm new to the Rcpp game so apologies if this question has be erroneously posted to this list. 

Yes, it is absolutely the correct list.
 
| I've recently taken over maintenance of a package (Rclusterpp) that makes
| heavy use of Rcpp & RcppEigen and have been trying to build and check the
| package for submission to CRAN. I'm currently working against R-devel (3.1,
|  Mac OS X 10.8, G++ 4.8 via macports) and am able to successfully run the R
| CMD BUILD of the package. However, when I run R CMD CHECK --as-cran
| <mypackage.tar.gz>, I encounter the following NOTEs:
| 
| * checking dependencies in R code ... NOTE
| Unexported object imported by a ‘:::’ call: ‘Rcpp:::RcppLdFlags’
|  See the note in ?`:::` about the use of this operator.
| See the information on DESCRIPTION files in the chapter ‘Creating R
| packages’ of the ‘Writing R Extensions’ manual.
| 
| To the best of knowledge, my package makes no calls to Rcpp:::RcppLdFlags
| so I'm not sure what's going on here.

It is the src/Makevars in your package which needs to call R so that R can
tell it where the Rcpp library is which your package needs.  [1]

We have recommended this very approach for years.  Now ':::' is outlawed, and
I changed the SVN version of Rcpp to export RcppLdFlags() so you can call it
via '::' which should not trigger the message.  But until the SVN version of
Rcpp is released [2], there is nothing you can do.  Just note in the submission
to CRAN that with the current CRAN version of Rcpp, there is nothing you can
do. 

We (as in Rcpp authors) are bystanders to changes beyond our control here.
 
| I do not get the same NOTE messages when building and checking using the
| current stable version of R (3.0.1), only the new development version. Does
| anyone have a suggestion as to how I might be able to debug the problem? I
| don't think this message causes any functional problems with the package
| but ask because the package was not accepted for hosting on CRAN due to
| these NOTE messages.

You are merely learning how CRAN "communicates" new requirements via checks
in unreleased versions of R :-/  There was a somewhat heated 

Dirk

[1] By the way, I just looked at your src/Makevars in the most recent CRAN
    release of Rclusterpp -- your version is way too complicated. See the
    Rcpp-package vignette, and any of the packages using RcppEigen (or
    RcppArmadillo).

[2] We are release-ready, but a regression test against all packages I can
    build on my machine (approximetely 100 out of 137) revealed that some
    changes by Romain broke the build for four packages. I fixed one such
    change; but three broken packages remain.  If you would like to see a new
    Rcpp release, it might help if you nudged Romain via a friendly email to
    fix the code to not break those three packages (all from the same issue,
    it seems).  Until those three packages build, CRAN would not admit our
    package so no CRAN release is possible until then.

-- 
Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com