[Rcpp-devel] Unable to build Package under Rcpp with Windows XP using canonical Example

[Dirk Eddelbuettel]

On 14 December 2010 at 14:41, Savitsky, Terrance wrote:
| Sorry for not being clear.  I'd already used verbose=T with inline to check the compile and build steps.  The scripts are identical, but inline works and R CMD does not.  I've used simpler and simpler examples and re-focused on Rcpp away from RcppArmadillo in an attempt to isolate the problem.  
| 
| So what I did next was to build a package skeleton under Rcpp with that included the canonical C++ code and R script (Rcpp_hello_world) included there.  I built it, but when I checked it; the compile step completed, but the build broke, with the following message:
| 

[...]

| I also attempted to build a package under RcppArmadillo using the fastLm C++ code.  This is the same code from which I attempted to create a SHLIB.  Not surprisingly, I receive the same compile error as before when checking the package:

[...]
 
| I'm unable to discover a pattern between these (and other) compile and build errors I'm achieving that would help me solve it.  As mentioned, the scripts compile and build fine with inline under the same flag texts.  The snprintf issue that appears when I employ RcppArmadillo is not fixable by including cstdio, cstring or locale.  Anyway, it doesn't make sense that an idiosyncratic tweak under one scenario will fix the other errors I receive.  Again, my Rtools installation is fine with nothing obvious there.  I'd appreciate your helping me think through what factors would cause inline to work and R CMD to not when the includes, namespace, wrappers and flag text are *identical* between the two.  Anyway, I'm at whit's end since I'm unable to build a package through Rcpp.

I am sorry but I do not have any further suggestions:

  - the tools to use are documented in Writing R Extension etc

  - it works for CRAN which creates Windows binaries of these packages from
    source; dito for R-Forge

  - I just rebuilt a package using Rcpp at work using the current Rtools and
    it also 'just works'.

Sorry,  Dirk

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