I don't think we need this.

Do you think shield is better?

His check code doesn't really work for c++.

Yes I do think Rcpp::Shield<> is better. It is also super-simple. No loss here.

But I agree that his check gets tripped up. He also sent me a false positive for the (old !!) RcppClassic where things happen correctly in the destructor but he didn't see that.

The bigger story is of course that we could just as well

        ans = Rcpp::LogicalVector(n, NA_LOGICAL);

instead of

        ans = Rcpp::Shield<SEXP>(Rf_allocVector(LGLSXP,n));
        std::fill(LOGICAL(ans),LOGICAL(ans)+n,NA_LOGICAL);

Ie

R> library(Rcpp)
R> cppFunction("LogicalVector makeOne(int n) { return LogicalVector(n, NA_LOGICAL); }")
R> makeOne(3)
[1] NA NA NA
R> class(makeOne(3))
[1] "logical"
R> 

If you have a moment over the weekend I would appreciate it if you could glance at this.

Wil do. Just straight from work to dinner.

I'm looking at the source for shield.

It doesn't do out of order destruction. In this case, it doesn't really matter.

Why not use UNPROTECT_PTR ?

Anyway, his check is just bogus, so there's not really a need to make this change. but if you want to make the code more 'Rcppish' I'm ok with that.