It turns out that this is in fact a regression -- it built with 0.10.6, but
failed unchanged with 0.11.0.
Notwithstanding the fact that we tested 164 CRAN packages, as well as 17
BioConductor packages, and all of the Rcpp Gallery -- as soon as we released
Rcpp, Tal found a new one. That's Murphy's Law for you.
The good news is that Kevin has already committed a fix to Rcpp at GitHub
which does address the regression, but Tal will now have to wait for a
subsequent CRAN release Rcpp 0.11.1.
Thanks, Dirk